Eit eksperiment: Denne sida er omsett automatisk frå
bokmålsutgåva og kan innehalde feil.
MishMash vil prøve ut ny teknologi i det opne: teste, avdekkje problem og rette dei,
både i våre eigne og i underliggjande system. Det er ein del av MishMash-DNA-et.
Slik blir nynorsksidene laga.
Resultat
Forskingsresultat registrert under MishMash-prosjektet i NVA. Vil du at resultatet ditt skal stå her? Legg det inn i NVA og inkluder MishMash i finansieringsfeltet. Sjå også forskingspulsen for ei visuell oversikt. Datasett og programvare høyrer også heime i NVA, og kjem med i lista over når dei er registrerte der.
This article provides an overview of current AI-assisted tools used in music composition and production and demonstrate their application through research-based music production. By highlighting the diverse range of applications and levels of automation offered by AI tools, we nuance the discourse focused on AI-generated tracks and challenge the assumption that using AI necessarily equates to cheating. We demonstrate how musicians can harness AI to support their artistry without overshadowing human expression, ensuring that creativity and integrity remain at the forefront
This paper presents LightHearted, an open-source Python framework for mapping the heart’s electrical activity from electrocardiography (ECG) signals acquired from performers to stage lighting in a concert setting. The aim is to provide dynamic lighting coupled to the physiological processes occurring during performance. Drawing upon critical approaches towards using biosignals in interactive systems and live concert research, we outline the framework’s design and implementation and present a case study with the Aarhus Symphony Orchestra. This paper highlights both technical and conceptual challenges of integrating biosensor-based lighting into a large-scale orchestral context. Results from a post-concert audience survey (N = 324) suggest that while responses to the lighting were mixed, it was generally not perceived as distracting and would be welcomed for use in future concerts.
This paper presents a mixed-method approach to investigating the felt experience of a networked music–dance performance. Using this approach, we explore how a sense of togetherness emerges even when performers are distributed across distant spaces. Drawing on data from the Telematic@Popsenteret research concert, we propose a three-layered analytical framework that integrates computer-based audio-video analyses with qualitative observation, and post-performance interviews. By combining sound–motion synchrony and temporal coupling with performers’ and audience members’ experiential accounts, we examine how dancers and musicians negotiate coordination and shared presence in a telematic environment. Our findings suggest that togetherness in such settings is not merely a matter of temporal alignment, but a dynamically constructed, multimodal process involving perceptual adaptation, cross-modal attunement, and affective awareness.
RhyGlyph is a radial visualization system that represents rhythmic interaction as three pairwise spacetime trajectories within a compact three-wedge glyph, supporting both overview and detailed reading. Building on this representation, RhyDiff is a generative framework that learns aligned latent spaces for symbolic drum patterns and the encoded trajectory features through paired VQ-VAEs and FiLM-conditioned diffusion. Users edit relational trajectories and metadata to generate new drum patterns by treating visualization itself as a control surface.
The study presents key findings to evaluate and ensure the readiness of GenAI tools for use in the professional VFX, film and media industry. Building on previous insights, we present themes which will support rather than hinder artistic practices as well as considerations on ethical guidelines to ensure sustainable and fair use. The study suggest skills to develop in order to incorporate GenAI tools efficiently such as: adaptability of tools and workflow, knowledge in sufficient input quality, prompting, and to be able to apply human ideation and filtering as well as discipline specific competence to the GenAI tool assisted process throughout the whole production workflow. The study suggests that HAIC tools, while being highly cost- and time-effective, are less valuable to incorporate in post-production and final production as they lack sufficient control, high resolution output, nor an option for personalization.
This paper presents a live semi-autonomous system that co-improvises with a live musician using a model learned from the relationship observed in paired recordings of an improvising duo. This responsive system is based on a deep learning approach that models the relations between sequences of events produced by co-playing musicians in continuous time, using transformer models. We present the architecture for simultaneous sequences of sonic events and provide a quantitative evaluation on several representative tasks. Results are compared against a canonical transformer and a multichannel Factor Oracle, the latter being a widely used model for live sequence-based symbolic music generation. We detail a live implementation employing concatenative synthesis and introduce a customization procedure in which the generative model is iteratively retrained on curated, satisfactory sections from sound recordings of actual human-system co-improvisational interactions. This iterative fine-tuning enables the model’s stylistic output to diverge from the original training corpus. Two musical use cases demonstrate the application of this technique.
This study explores the accuracy and usefulness of OpenAI’s large language models (LLMs), particularly focusing on a customized GPT model trained on a specific dataset of musicology research. The research compares the performance of the custom model against the general ChatGPT-4.o model in an academic setting. The customized GPT, tailored to retrieve information from a controlled set of scholarly articles, was tested by a group of musicology experts, including authors whose works were part of the dataset. Results indicate that the custom GPT generally provides more precise and research-based responses compared to the general model. However, issues such as hallucinations, inconsistent performance, and ethical concerns regarding copyright and the use of LLMs in academic work were observed. The study concludes that while customized GPTs can enhance academic research, their use should be approached with caution due to ongoing challenges in accuracy and ethical implications.
Claudia Magdalena Sikora
;
Marieke van Otterdijk
;
Rebekka Soma-Jestilä
;
Jim Tørresen
;
Diana Saplacan Lindblom
Non-anthropomorphic robots are integrating into our homes. People
attribute human-like characteristics to these robots, although they do not
resemble humans. This raises the question of whether similar processes would
occur when people interact with robot furniture. Thus, the presented exploratory
pilot study investigates how users intuitively interpret and perceive the non
verbal communication of such robots. In an instrumental case study, we had
eleven participants interacting with the T-ABLE robot through expressive
behavior. Participants were interviewed about their experience. Our findings
indicate that the participants use their associations between behavior and prior to
similar experiences to attribute personality to the robot table. These findings
provide insights for theorizing the intuitive comprehension of the behavior of
non-anthropomorphic robots.
Human interaction with large language models (LLMs) is commonly framed as information exchange, optimized for efficiency and correctness. This paper proposes a different framing: keyboard-as-instrument, where typing to an LLM becomes a form of musical practice. Drawing on the lineage of keyboard instruments, the action-sound framework, and the aesthetics of failure in jazz improvisation and glitch music, the paper introduces vyping (typing without looking) as a technique that treats degraded input as an expressive gesture. When a person vypes, inter-onset intervals and keystroke dynamics carry embodied information that the LLM can integrate, in a process closer to musical comping (responsive accompaniment) than to improvisation proper. Three artifacts (ll00mtube, clix-vibe, and KeyMeter) illustrate this approach, making the temporal and sonic dimensions of typing visible and audible. The paper argues that typos can function as grace notes: productive deviations that open new directions. It proposes reverence (nuance, grace, and respect) as a design orientation for human-AI interaction, while acknowledging the discomfort of caring for a tool entangled with extractive labor and concentrated corporate power.
The maritime sector is both vast and crucial for society on a global scale. At the same time, it is extremely complex, relies on multiple information flows, operates in constantly changing conditions, integrates hazardous environments, and depends on critical human decision-making informed by all the previous elements. Such an environment arguably constitutes fertile ground for AI-powered optimisation, ranging from warning systems to co-pilot functions to autonomous vessels. These AI integrations would benefit from standard user interface design practices to ensure they are well integrated into mariners’ work. There are potential benefits, as well as risks and challenges, inherent in attempting such an integration. Reviewing these elements informs the requirements for designing the user interfaces for this purpose. This paper discusses the emerging picture of the qualities, conditions, and requirements, as well as the implications of the implemented design decisions.
Prompt engineering strategies emphasise clarity, specificity, and well-structured input, on the assumption that large language models (LLMs) require clean input to interpret user intent. But LLMs are pattern interpolators: they “absorb” noise through statistical reconstruction rather than strict symbol matching. Text-centric HCI has studied this tolerance for two decades without integrating it into prompt engineering as a resource. We introduce vyping as an intentional mode of prompting in which one types into LLMs at the speed of thought, without correction. We developed vyping through sustained first-person inquiry. It can be seen as a micro-practice inside a broader post-discursive turn in which meanings increasingly travel as rhythms and patterns through latent-space proximity rather than as precisely-formed symbols. The mismatch between current LLM interfaces and their (human) users has led to two common layers in implementation: (1) an orthographic residue at the interface (autocorrect, spellcheck, tidy-prompt guides), and (2) an impairment parse shadow at the training layer, a form of overalignment in which the model reads orthographic deviation as a sign that something is wrong with the user. Overall, we offer three attunements (ease, reach, and pulse) as alternatives to efficiency-first prompt engineering, and situate the practice within critical HCI traditions of seamful, frictional, and embodied interaction. The critique is structured around a seed prompt that suspends the impairment parse and makes vyping livable; its section headings are taken from the prompt’s own rules. A copy of the seed prompt is provided at the end.
Pixasonics is a new Python library for interactive image analysis and exploration through image sonification. It uses real-time audio and visualization to help uncover patterns in image data. With Pixasonics, users can launch one or more small web applications (running in a Jupyter Notebook), probe image data using various feature extraction methods, and map those feature vectors to synthesis parameters. The target users are researchers interested in exploring image and volumetric data and creative users who want an intuitive tool for experimental sound design. Pixasonics’ design aims to strike a balance between an easy-to-use web application with minimal boilerplate code necessary and a library that can be integrated into more advanced workflows. Real-time exploration is at the heart, but it can also be used to script non-real-time sonifications of large datasets. This paper presents Pixasonics, its structure, interface, and advanced features, and discusses preliminary feedback from biology researchers and music technologists.
Kunstfagene spiller en viktig rolle i å utvide måten vi jobber med og forstår komplekse samfunnsproblemer. Likevel blir kunstfagene stadig oversett og nedprioritert i forskningspolitikken. Vi spør derfor: hva er, bør og kan kunstens rolle være i det norske forskningslandskapet?
In this chapter, the colonial intervention of the technologies of capture are examined in the Global South regions, especially in South Asia in the beginning of the twentieth century. I study the ramifications of this entry in the form of mutating everyday sounds and musical practices into commodified sound objects for industrialisation and sale. From a critical decolonial position, this introduction of the technologies of capture is read here as a precarious intrusion into the sounding and listening cultures of the Global Souths, and a moment of emergence of an extractive capitalism through the colonial routes that devastate the sonic symbiosis of the land.
This paper explores how musicking technologies—interactive systems with musical properties—can enhance everyday public environments. We are particularly interested in investigating the effects of musical interactions in non-musical settings, such as offices, meeting rooms, and social work areas. Traditional music technologies (such as instruments) are built for goal-directed, conscious, and voluntary interactions. We propose a new perspective on embodied AI through systems that utilize indirect, inverse, unconscious, and, at times, involuntary interactions. Four different sound/music systems are examined and discussed with regard to their activity level: a reactive “birdbox,” a reactive painting, active self-playing guitars, and interactive music balls. All these systems are multimodal, containing sensors that detect various physical inputs to produce sound and light, and having varying levels of perceived agency. The paper explores differences between direct/indirect and regular/inverse embodied AI paradigms. This study demonstrates how minimalistic interactions have the potential to yield complex and engaging musicking experiences, challenging the norms of overly intricate AI implementations.
Musicologists, psychologists, and computer scientists study relationships between auditory and visual stimuli from very different perspectives and using various terminologies and methodologies. This article aims to bridge the gap between phenomenological sound theory, auditory–visual theory, and audio–video processing and machine learning. We introduce the SoundActions dataset, a collection of 365 audio–video recordings of (primarily) short sound actions. Each recording has been human‑labeled and annotated according to Pierre Schaeffer’s theory of reduced listening, which describes the property of the sound itself (e.g., ‘an impulsive sound’) instead of the source (e.g., ‘a bird sound’). With these reduced‑type labels in the audio–video dataset, we conducted two experiments: (1) fine‑tuning the latest audio–video transformer model on the reduced‑type labels in the SoundActions dataset, proving that the model can recognize reduced‑type labels, and observing that the modality‑imbalance phenomenon is similar to the added value theory by Michel Chion and (2) proposing the Ensemble of Perception Mode Adapters method inspired by Pierre Schaeffer’s three listening modes, improving the audio–video model also on reduced‑type tasks.
Under-resourced languages (and musics) pose a challenge to machine translation (MT). The challenge is greater when the content of the collected dataset is a varied sample taken from a data population that is even more diverse and dynamic. This is the challenge of Arab music vocal improvisation (mawwal). Here, we present the development of AMICOR, a parallel dataset consisting of vocal improvisatory phrases and their corresponding instrumental responses (or tarjamat in Arabic, which literally means “translations”) in the mawwal tradition. These melodic phrases are handled as “sentences” from the viewpoint of natural language. When developing the dataset, we integrated musicological insights in order to evaluate music theoretical differences between sub-datasets, primarily regarding their size, sentence length, performance quality, and shared musical identity. We then experimented with MT to generate instrumental responses to new vocal sentences, comparing several translation modeling configurations that differ (1) in translation approach (Neural MT (NMT) versus Statistical MT (SMT)), and (2) in the dataset handling approach in respect to the maqam (an Arabic musical term referring roughly to a melodic mode), comparing an individual model for each maqam versus a unified model for all maqamat. We found that merging related sub-datasets does not necessarily lead to better results, and may even favor simpler and shorter sentences with lower performance quality and less sophisticated patterns. This issue applies to both NMT and SMT; however, it is greater for NMT. A comparison of confusion matrices of individual-maqam models suggested that, in such a small dataset, the gap between SMT and NMT performance increases further if the styles, or skills, of potential users differ from those who built the dataset used in the training. Our discussion asserts that key factors in system design are the musical background and performance decisions of vocalists who may use such responsive generative models, as well as dataset size and performance quality.
In his recent critique of well-established simplified comparisons between the human mind and AI, Alva Noë encourages us to 'rage against the machine' (Noë, 2024). In this article, I build on Noë's timely encouragement, and sketch perspectives from both music studies and enactivist philosophy in order to interrogate what musicking might add to the debate. As a case study, I explore the recent public discourse surrounding the AI-band The Velvet Sundown and consider the implications — mental health related and ethical — of engaging with AI-generated music. I draw on current scholarship on musical creativity and AI to discuss the possibilities and limitations of creative agency in AI, laying the foundation for why human creative agency should be regarded more valuable than the artificial creativity that AI offers. I also discuss some ethical implications of sharing agency with AI, suggesting that we should continue to foster attunement to our creative agency and capabilities, as empowered but also vulnerable musicking human agents. I argue that musicking offers an enactive entanglement through which we are granted an opportunity to embrace risk, resist rules, and learn new modes of orientation and reorientation that serve as empowering tools for navigating our way through the resistant path to being human.
Studies on large language models (LLMs) have shown that these technologies can reproduce stereotype-based biases present in their training data. In this paper, we hypothesized that an OpenAI-based model would more frequently attribute first authorship to women in masculine-typed research fields than to men in feminine-typed fields, an asymmetry consistent with gender-role beliefs. Across three studies, we analyzed how the model’s generative outputs reflect gendered associations. Study 1 replicated Fulgu and Capraro’s phrase-association paradigm, replicating the study’s findings that feminine-stereotyped phrases were more consistently linked to female characters, whereas masculine-stereotyped phrases were less reliably linked to male characters. The results suggest that the training data contains a stronger and more stable association between femininity and communal traits than between masculinity and agency. Study 2 examined 4,585 randomly generated hypothetical studies and identified a substantial skew toward female first authors (>82%), with male authors overrepresented primarily in engineering, technology, and theoretical contexts. Study 3 extended these findings across nine specified research fields, where the results of the first stage were consistent with Study 2. When prompted to classify rather than generate the gender of the first author, the model produced a lower overall proportion of women, but retained the same relative distribution across fields observed at the first stage. Taken together, the results indicate that the model’s outputs correspond more to feminine than masculine gender-role beliefs. While this tendency may appear to increase women’s representation in masculine-typed domains, it does not afford men equivalent visibility in feminine-typed domains. These results underscore how LLMs, as statistical systems shaped by training corpora and reinforcement learning, can reflect and reproduce cultural and historical asymmetries in gender representation, challenging assumptions of epistemic neutrality.
This article examines how performative togetherness is constructed in a technologically mediated music–dance performance environment. While togetherness is often associated with co-presence and effortless flow, networked contexts introduce perceptual reconfiguration that makes it fragile and effortful. We present the research concert Telematic@PopSenteret, a networked performance involving dancers and musicians distributed across two floors in a museum, connected through audio–video transmission. Using interviews and audio-video recordings, we explore how latency, fractured perception, and the absence of bodily cues transformed the performers’ coordination and induced cognitive strain. We discuss how performative togetherness – defined as the sense of shared meaning-making that emerges through attentional, sensory, and adaptive negotiation – was maintained during the performance.
‘It isn’t real music!’ As I write this review there is an ongoing debate in Norwegian media about a band that has been booked to play in a festival next summer. The band is composed of human performers, but they are playing songs generated by AI. Or, more precisely, they have written the lyrics themselves and then fed them into a commercial AI-based music service that has created a melody and accompaniment. Is this ‘AI music’? Is it ‘real music’? Is it more or less ‘real’ whether they play acoustic instruments or laptops on stage? And, ultimately, does it matter who made what if people want to listen? To understand today’s disruptive introduction of AI-based commercial services better, it may help to look more closely at history. This is where Deirdre Loughridge’s book is a timely and valuable contribution. It reveals that human fascination with—and harsh scepticism of—machine-generated music is nothing new.
This article investigates the intersection of generative Artificial Intelligence (AI) and the book industry, focusing on the use of copyrighted literary works for training large language models (LLMs). As LLMs require enormous volumes of content for training, published books and other media have become valuable sources of data. Many training datasets appear to have been assembled without authorization or transparency, raising acute legal, ethical and cultural questions. In this article, we provide a detailed empirical analysis of the Scandinavian content within the Books3 dataset, used to train multiple major AI systems, characterizing it by publication year, publisher, authorship, genre, and language. As comparatively small language markets situated outside of the main hubs for AI development, the Scandinavian countries provide a relevant context for exploring the local implications of global AI deployment. The findings are discussed through a theoretical lens drawing on Michel Serres’s concept of the “parasite.” We argue that the inclusion of Scandinavian books in Books3 is part of a multidimensional ecology of parasitic exchange. This framework interprets the relationship between AI developers, shadow libraries, and publishers as a cascade of information appropriation, signifying a reconfiguration of cultural production and ownership in the age of generative AI.
Explainable document visual question answering systems improve transparency by visualizing document regions relevant to a query, but they can unintentionally expose sensitive information through explanation outputs, even when textual answers are restricted. We analyze privacy risks arising from coarse explanations and adversarial jailbreaking queries. To mitigate this, we propose a policy-aware visual explanation sanitization framework based on a Role-centric Attribute-Based Access Control (RABAC) model, combining document structure analysis and line-level localization to mask restricted regions. Our method is model-agnostic, supports dynamic policies, and significantly reduces sensitive information leakage while preserving interpretability.
Jiin Han
;
Suyoung Bae
;
Yerim Choi
;
Sangyun Lee
;
Kyutae Kim
;
Ryunho Kim
;
Jee-Hyong Lee
;
Changkyu Choi
;
Yunseok Choi
Tables are central to scientific communication, yet their interpretation depends critically on layout, continuity, and surrounding context. While existing multi-page and multi-table benchmarks formally require document-level reasoning, models can often succeed through stochastic shortcut search—locating local answer-bearing regions without navigating table structure. To address this gap, we introduce PubTables-QA, a multi-page TableVQA benchmark designed to audit responsible cross-page visual reasoning in academic documents. Built on PubTables-v2’s structural annotations, PubTables-QA provides 2,106 QA pairs that are unanswerable from any single page, compelling models to reconstruct table continuity across pages or integrate evidence across multiple tables. We further introduce a three-level QA taxonomy spanning Document, Table, and Cell/column levels. Experiments on recent MLLMs, including GPT-4o, Gemini-2.5-Pro, and Gemma-4, show that even the strongest model reaches only 32.6% accuracy. Oracle-setting analysis further shows that performance depends strongly on access to the correct page, table region, and cell-level evidence, highlighting persistent failures in faithful cross-page evidence integration.
Automated code documentation is essential for modern software development, providing the contextual grounding that both human developers and coding agents rely on to navigate large codebases. Existing repository-level approaches process components independently, causing redundant retrieval and conflicting descriptions across documents while producing outputs that lack hierarchical structure. Therefore, we propose MemDocAgent, a long-horizon agentic framework that generates documentation within a single, integrated context spanning the entire repository. It combines two components: (i) Dependency-Aware Traversal Guiding that predetermines a traversal order respecting dependency and granularity hierarchies; (ii) Memory-Guided Agentic Interaction, in which the agent interacts with RepoMemory, a shared memory accumulating prior work traces through read, write, and verify operations. Through an in-depth multi-criteria evaluation, MemDocAgent achieves the best performance over both open and closed-source baselines and demonstrates practical applicability in real software development workflows.
Martine Hjelkrem Tan
;
Marius Aasan
;
Rwiddhi Chakraborty
;
Gabriel Yanci Arteaga
;
Changkyu Choi
;
Adín Ramírez Rivera
Masked Image Modeling (MIM) has become a ubiquitous self-supervised vision paradigm. In this work, we show that MIM objectives cause the learned representations to retain non-semantic information, which ultimately hurts performance during inference. We introduce a model-agnostic score for semantic invariance using Principal Component Analysis (PCA) on real and synthetic non-semantic images. Based on this score, we propose a simple method, Semantically Orthogonal Artifact Projection (SOAP), to directly suppress non-semantic information in patch representations, leading to consistent improvements in zero-shot performance across various MIM-based models. SOAP is a post-hoc suppression method, requires zero training, and can be attached to any model as a single linear head.
Researchers have argued that temporal microdeviations from the metric grid, such as those producedby musicians in performance, are crucial to making a musical rhythm groovy and danceable. It is curious, then, that the music currently dominating the dance floor, “electronic dance music” or EDM, is typically characterized by grid-based rhythms. But is such a “mechanistic,” grid-based aesthetic necessarily devoid of microrhythmic nuance? In this article, we aim to show that the microrhythmic component of an engaging groove involves the manipulation of more than simply the onset locations of rhythmic events—the sonic features fundamentally contribute to shaping the groove as well. In particular, we seek to demonstrate that EDM producers, with their preference for a grid-based microtiming aesthetic, are very sensitive to and adept at manipulating such sonic features for expressive effect. Drawing on interviews with EDM producers, we show that producers are often concerned with both sonic and temporal features, as well as their interactions. We argue that sonic features are crucial to shaping groove and feel at the micro level of rhythm. Moreover, such features also tend to introduce an indirect microtiming aspect to the grid-based aesthetic of EDM through the ways in which they shape timing at the perceptual level.
Atomspheres is a series of performances on elemental listening engagement and meditative environmental attunement, leading to a participatory performance with the co-listeners. The performances attune to the macro-scale crises of the climate and ecology in a planetary context and bridges its intersection with micro-level coping strategies to the crises in a regenerative gathering. The performances resonate with the tradition of free improvisation as well as communal and participatory sonic gatherings for affective and discursive engagement with crises and conflict, practiced in many Global Souths cultures. These performance gatherings involve field recordings of climate variables like wind, water, and woods, comprovised by chance generative processes, accompanied by radio interferences transducing situated weather patterns, weaved together by meditative instrumental interventions of traditional instruments, as well as live electronics and algorithmic interpretation of data and assemblages embracing free improvisational techniques and ethos. As an invocation, the series invites the co-listeners into an exercise of slow listening and slow breathing that try to suspend time, control, ego, and individuality to tap into a humbler mode of fatalism and a distributed self-hood. This philosophical positioning stresses the subjugation of events or actions to fate or destiny and is commonly associated with an attitude of acceptance in the face of future events, which are thought to be inevitable. This approach manifests into yielding control by an emergent form of togetherness, intersubjectivity, and kinship when listening transcends an overt causality, and the psychosomatic condition embraces slow breathing to contemplate, and affective attunement. The project is inspired by the micro-level entanglements of human, more-than-human, and other life forms in a survival context, and explores sensory transduction of environmental conditions to figure how to make them post-immersive and attuning experiences in an openly circular, collaborative performance setting.
An improvisation with Muzziballs, ball-shaped musical instruments developed in collaboration between Muzziball AS and RITMO, in which sound is shaped by how each ball is moved and handled. Performed at the Department of Musicology's department concert in the ZEB building on 13 April 2026 by Bilge Serdar, Maham Riaz, Arthur Guo, Alexander Refsum Jensenius, Katerina Teleli and Margarida Leal.
Trenger vi egentlig musikere, forfattere og filmprodusenter nå som KI kan gjøre jobben? Vi tar diskusjonen om hvordan KI påvirker kreativ næring og kunstneriske prosesser. Debatten raser rundt bruken av kunstig intelligens i Kultur-Norge.
Forlaget Bonnier brukte nylig et KI-generert bilde som forside på en bok, og "KI-bandet" Vestavind er booket til flere norske musikkfestivaler.
Flere kjente kunstnere og artister tar til orde for tydeligere regulering eller forbud av KI for å beskytte kunstnernes virke og opphavsrettigheter.
Hva er egentlig kreativ KI? Hvordan påvirker kunsten KI? Hvordan kan vi beskytte kunst og kunstnere, samtidig som vi tar i bruk ny teknologi?
EPARM 2026: Alone Together: Exploring Dimensions of Interdependence and Cooperation. The EPARM Working Group is in charge of organising the AEC Annual EPARM meeting. The Working Group is formed by active individuals in the relevant field of specialization within AEC member institutions.
AI Mission Hub and BI Startup Village co-host the Build what matters event
AI Mission Hub and BI Startup Village bring together organisations, AI professionals, researchers, founders, students and technology experts to transform real-world challenges into practical AI solutions.
The event will feature expert talks, panel discussions, real-world AI case studies, networking opportunities and the official presentation of AI Mission Hub’s founding cohort.
The Great Human Filter er et KI-generert eksperimentelt funk- og nu-jazz-prosjekt. Prosjektet kombinerer generativ kunstig intelligens med menneskelig kuratering, og utforsker en tematiske kobling mellom teknologi, kreativitet og menneskelig dømmekraft.
Over the past few months, students in IMV's Computer Music course (MUS2850) have learned how to create their own instruments using the audio programming language Pure Data.
At this event, the students will perform together in groups using their newly programmed instruments. The concert will also feature performances by guest artists Victoria Johnson and Vincenzo Madaghiele.
The research exhibition Archival Intelligence examines the roles of today’s all-pervasive digital technologies, such as Artificial Intelligence, in extracting data from archives and museums for profit.
Dhvāni means “resonance” in Sanskrit. Resonances between sounding objects and between people are the starting point for this installation by Indian-born artist and thinker Budhaditya Chattopadhyay. In STUK’s courtyard, a web is stretched with hundreds of Indian ceremonial bells and other instruments, such as wind chimes and ghungroos. By means of a self-built artificial intelligence system, this resounding network responds to human presence: footsteps, voices, hand clapping. Through machine learning, the interconnected system improves its performance over the six weeks of Hear Here, creating a beyond-human resonant organism. In this way, Chattopadhyay links ancient musical traditions with contemporary technologies, inviting visitors to listen empathetically and step outside their judgmental “ego”.
This booklet contains the program and abstracts for the Opening Conference of MishMash Centre for AI and Creativity held on 14–15 September 2026 at Kilden Performing Arts Centre, Kristiansand, Norway.
This report presents extended summaries of the presentations and discussions at the DAIM 2026 Symposium on Digital and AI Musicology, held on 8–9 June 2026 at the National Library of Norway and the University of Oslo. Organized by RITMO, the MishMash Centre for AI and Creativity, and the National Library of Norway, the symposium brought together the COST Action EarlyMuse and MishMash Work Package 6 on AI for Cultural Heritage.
The contributions examine how computational methods and artificial intelligence can support the preservation, transcription, analysis, discovery, and interconnection of musical heritage. Topics include digital music collections, Optical Music Recognition, audio transcription, chant and Sámi yoik, musical similarity and pattern discovery, Linked Open Data, metadata standards, and sustainable research infrastructures. Discussions address scholarly validation, provenance, cultural context, and the rights of creators and communities.
Compiled and edited by Olivier Lartillot, the report synthesizes presentation abstracts, slides, and recorded talks. It includes introductions to both days, presentation and panel summaries, links to related resources, contributor biographies, and the symposium programme, alongside a tribute to Eleanor Selfridge-Field.
Innlegget presenterer forskningssenteret MishMash – Centre for AI & Creativity, et av Norges nasjonale forskningssentre for kunstig intelligens. Presentasjonen diskuterer hvordan KI kan forstås ikke bare som et verktøy, men som en aktiv medspiller i kreative prosesser. Med utgangspunkt i kunstnerisk praksis og forskning i skjæringspunktet mellom teknologi og kultur, belyses nye muligheter, utfordringer og etiske spørsmål som oppstår når KI inngår i kunstnerisk arbeid og kreative næringer.
First, I will present MishMash Centre for AI and Creativity, a new Norwegian research centre advancing AI to explore and shape human-machine creativity. Funded by the Research Council of Norway in 2025, the centre brings together artists, engineers, and scholars across disciplines to create, explore, and reflect on AI systems that augment creative practice while foregrounding agency, inclusion, and sustainability. Organized into seven work packages, the centre addresses technical challenges such as real-time multi-agent systems and hybrid symbolic learning methods, societal concerns including bias, copyright, and equitable value distribution, and applied domains spanning health, education, cultural heritage, and the creative industries.Second, I will present some of my own ongoing explorations into embodied AI, including inverse control, small language models, and distributed computing. This approach challenges current monolithic, generative models, suggesting an alternative future for creative AI.
Motion is at the core of how we experience and understand the world, express ourselves, and interact with others. At RITMO, we explore motion that is artistic, expressive, and interactive within a highly interdisciplinary research environment that integrates musicology, psychology, and technology. Increasingly, we are moving our research out of the lab and into the real world, where we can study people behaving as they normally do and capture unique occurrences that cannot be simulated in laboratory conditions. This transition is illustrated by the MusicLab and Bodies in Concert projects, which investigate bodily activity and experiences in performers and audience members during live concerts. From data collected during a series of symphony orchestra concerts, we are learning how musicians coordinate their expressive motion and how audiences synchronize in their bodily responses to the music. We are also learning how to adapt new technologies for real-world data capture, and exploring how captured data can be used for artistic purposes. In this talk, we will discuss some of the concepts, methods, and findings from this line of research.
When individuals use generative AI, idea diversity decreases sharply; when whole teams rely on the same tools, convergence compounds. No prior studies have examined how distributing different AI tools across team roles affects creative collaboration. This paper introduces the "AI as Fifth Team Member Model", which defines four roles: Research Coordinator, Visual Design Specialist, Space Planning Expert, and Presentation Coordinator. Each uses a different tool, through text-based analysis, image generation, 3D modelling and layout design, forcing heterogeneity into the team's creative process. A pilot showed that daily synchronisation protocols are essential for cross-role coordination. Future studies should test larger groups and measure diversity formally.
A frequently used perspective for theological analysis of AI is that of the Imago Dei, an application pioneered by Noreen Herzfeld. The Imago Dei concept highlights the “artificiality” so to speak (the very createdness), also of the human person. AI is then created in the imago hominis, while, in a third step, AI systems generate artificial objects: increasingly in the imago automati, as the Internet is taken over by agentic AI. To put AI into the context of creation theology and art, is, I think, a promising path to thematize also the aesthetically and religiously important aspects of the current worldwide project striving to simulate intelligence, but also life. In this paper presentation, I will investigate to what extent the thinking of Thomas Aquinas can be a resource for such an investigation into the dignity of artificial intelligence. Of course, Aquinas had no idea of digital technology and modern neurophysiology, but maybe his taxonomies and principles can still be of value for an intellectual engagement with the nature and worth of AI.iv
The rise of increasingly sophisticated simulations of human reasoning, creativity, and language presents many challenges to individuals, institutions, and societies. For religious and spiritual traditions navigating the tension between adoption and rejection, AI revitalizes long-standing debates on human nature, the relationship between science and faith, and the natural and supernatural. Simultaneously, it fuels both secular and religious radical forms of techno-criticism against the industrialization and rationalization of culture. Within the Catholic tradition specifically, AI is often met with skepticism, as computational processes are seen to lack human qualities such as consciousness, emotion, and understanding, a stance exemplified by the Vatican document Antiqua et Nova. While many current Christian responses to AI development focus on the ethical, environmental, and social risks of large-scale AI adoption, this paper investigates whether a Thomist framework can offer an alternative evaluation that avoids a binary antagonism between humanism and automation. In a discussion with Edward Feser's critique of AI in Immortal Souls (2024), the paper argues for understanding AI through the lens of art (ars) rather than mere mechanics. As biotechnology further blurs the boundary between the organic and the mechanical, a theology of creation and art may better balance human nature in relation to artificial simulations than a humanism based on the privilege of interiority. This paper suggests that Thomism provides intellectual resources helpful for developing such a perspective.
These days' almost manic attention to AI should also trigger a response in the field of art. Not necessarily a denial of AI, but an "celebration" of what it means to be human. Can reflecting AI through art lead us to 'rediscover' our human potential? Rediscover our unique abilities, what we take for granted: like being able to create community through music and dance, like dealing with complicated ethical and moral questions through theater and film, like understanding through the visual arts what we do not see in our daily reality?
These abilities are something we have cultivated over thousands of years, a process of formation that is both biology AND culture: Music is a human surplus phenomenon, most likely sprung from our language ability (Patel, A. 2010). Therefore, music learning has many common features with language learning (Solli, M., Aksdal, E. and Inderberg, J. P. 2021). Just as we learn a native language through being surrounded by it, and through participating in social communities where it is used, music has traditionally been learned through imitation and through participation in (musical) communities.
How will AI affect these uniquely human forms of forming community/common consciousness? Time as a prerequisite for process, process as a prerequisite for exploration, exploration as a prerequisite for reflexive judgment/creativity. We will share our artistic research from a studio recording session where we focused on the idea of 'Try and Fail' or 'Not knowing'
Special Focus: Modeling Life Systems: Art, Algorithms, Ecologies
Presentation of the National Library of Norway's collection of audiovisual materials, it's legal framework, provenance, availability, as well as it's possibilities for dissemination and computational analysis (MIR).
Legal interpretation has traditionally presupposed an interpreter: a situated subject
capable not only of giving meaning to legal texts but also of judgment, justification
and responsibility. Generative AI unsettles this assumption. Large language models
can now produce persuasive legal interpretations without intention, lived experience
or normative commitment. This paper asks what that capacity reveals about legal
interpretation itself. Drawing on more than a decade of research and teaching in
human rights and law and technology across the Global South and Global North, the
paper argues that generative AI is best understood not as the end of legal
hermeneutics, but as a test of its underlying assumptions. The ability of LLMs to
reproduce recognizable forms of legal argument does not establish that legal
reasoning is reducible to statistical prediction. It does, however, expose the extent to
which legal interpretation relies on inherited linguistic patterns, doctrinal conventions
and institutional practices. The argument becomes particularly consequential in
human rights law, where interpretation claims sensitivity to dignity, vulnerability and
lived experience. Because legal languages and traditions are unevenly represented
in computational systems, AI may also influence whose interpretations become
easier to reproduce and whose experiences remain difficult to render legally visible.
The paper ultimately distinguishes interpretive capacity from interpretive authority
and responsibility. Its central claim is that the legitimacy of legal interpretation lies not
simply in intelligence or persuasive reasoning, but in answerability: the possibility of
being called upon to justify, defend and bear responsibility for meanings that acquire
authority over human lives.
MishMash is one of the six national AI research centres funded by the Research Council of Norway from 2025, and the only one that joins technology with creativity. In this presentation, in the showcase of the national centres, I introduce the centre: seven work packages spanning music, visual art, literature, games, performance and human-centred AI; a consortium of more than 20 partners across Norway; and the argument that creative practice is both a testbed for AI systems and a driver of new AI research.
Music production and songwriting tools based on machine learning and artificial intelligence are part of every producer’s toolkit. Whether it is the use of stem splitters to clean up tracks, algorithms for mastering, or generative systems for sound creation and lyric manipulation, these tools have found a place in the modern music studio.
However, the rapid introduction of such tools raises many questions. Do they improve the creative process? Are algorithms trained with sufficient awareness of juridical and ethical challenges? Is the use of training data in accordance with existing laws and practices regarding copyright and patents? Do these tools affect artistic output, making music better—or less diverse? Is AI making music creation more accessible by lowering technical and artistic thresholds and enabling non-experts to perform tasks that previously required specialist skills?
Based on interviews, experimental work, and a literature review, this paper explores which AI-based tools are used in contemporary music production, how they are employed, and how they influence the creative process. Technical, juridical, and ethical aspects of different types of AI-driven tools are discussed.
The paper concludes with an analysis of how new workflows may affect revenue streams for music producers, songwriters, and artists.
Mixing in immersive formats, designed to place the listener inside a three-dimensional sound environment rather than in front of a left–right stereo image, allows for many creative opportunities beyond traditional stereo mixing. However, there are different immersive formats, each with its own strength and challenges.
This paper/performance/workshop explores the creative differences between mixing in Atmos, created by Dolby Laboratories and used in movie theatres and commercial 3D audio distribution, and ambisonics, a format that is used extensively in electro-acoustic music and VR.
The music used in the paper/performance/workshop is a sonification of Aurora Borealis/Northern Lights. With the use of positional signals from an antenna array positioned in Kilpisjärvi in Northern Finland, soundscapes and composed music are projected into an 3D audio dome and create a real-time sonic experience of the movement of the Aurora on the sky over the arctic parts of Norway, Sweden and Finland.
These compositions exist in both ambisonics and Atmos, and in the paper/performance/workshop the workflow and creative opportunities inherent in the formats are explored, along with a discussion of the challenges encountered when working in the different 3D arrays.
Participants are invited to explore the immersive mixes and remix the performances in real-time.
Aim / key questions: As generative AI becomes embedded in creative practice, the
speed of creation has never been higher and the temptation to skip the slow parts of
the work never greater. We ask: what does creativity actually require, and what is at
risk when AI is allowed to bypass those requirements? We approach this as an
empirical question about creative process — and as a starting point for a comparison
between human and machine modes of musical thinking.
Methods: This artistic research project documents a two-day recording session at
Grieghallen in Bergen, where we deliberately worked from a position of "try and fail"
and "not knowing" — refusing pre-planned outcomes, making sketches on the spot,
and staying with uncertainty long enough for something unexpected to emerge. We
treat the recording process as a case through which the preconditions of creative work
can be examined.
Results: The work surfaced a chain of preconditions: time enables process, process
enables exploration, exploration enables reflexive judgment — what we identify as
creativity itself. Each link depends on the one before. Generative AI tends to compress
this chain — offering averaged ideas, weakening ownership, and removing the
productive resistance the not-knowing depends on.
Implications: In a next phase, we will examine our findings in dialogue with the Robotic
Musicianship group at Georgia Tech (Weinberg et al.), whose improvising robots such
as Shimon raise the question from the opposite direction: what does machine
improvisation share with — and lack from — human musical thinking?
Value / originality: A practice-grounded contribution to MishMash, opening a
comparison between embodied, communal music-making and AI-driven musical
generation.
Legal education is increasingly oriented towards “futureproofing” graduates for a rapidly
transforming legal profession. Much of this discourse, however, remains grounded in
narratives of optimisation, innovation and adaptation, implicitly treating technological futures
as inevitable and desirable. This paper challenges that assumption by asking a different
pedagogical question: what if legal education should prepare students not only for the futures
that are coming, but also for the futures that must be actively resisted?
Drawing on critical legal theory, human rights scholarship, and teaching experience across
the Global North and Global South, the paper introduces dark scenario pedagogy as an
experimental approach to teaching law. Dark scenarios depict deliberately dystopian, yet
plausible legal futures shaped by unchecked legal technologies, algorithmic governance,
climate breakdown, authoritarian data regimes, and the erosion of the rule of law. Rather than
treating such futures as speculative fiction, students are invited to analyse them as legally
actionable realities.
The paper demonstrates how dark scenario pedagogy enables students to: (1) identify early
warning signs of legal and human rights collapse embedded within seemingly neutral
technologies; (2) interrogate the limits of professional ethics and legal responsibility under
conditions of technological coercion; and (3) cultivate skills of resistance, refusal and legal
imagination that are rarely foregrounded in mainstream legal curricula.
By shifting the focus from “innovation” to prevention, this approach reframes future-oriented
legal education as a human-centred, rights-based and politically conscious practice. The
paper argues that teaching for futures we must prevent is not pessimistic, but pedagogically
necessary in a moment when law is increasingly mobilised to legitimise technological harm.
Dark scenario pedagogy offers a concrete method for reimagining legal education as a site of
critical foresight, ethical courage and collective responsibility.
Throughout my artistic career, I have created collaborative site-specific installations combining immersive audio with video, painting, drawings, and objects. These works are experienced only in specific locations for limited durations; afterwards, they exist only as documentation, typically a stereo video that struggles to capture their temporal, spatial, and immersive qualities.
Recent technological advances offer new possibilities. Dolby Atmos support from Apple and other providers now enables the distribution of spatial audio media for home consumption. Additionally, Swift and SwiftUI allow developers to build feature-rich media applications across Apple's ecosystem (such as iOS, macOS and tvOS). For artists with limited coding skills, agentic coding tools in Xcode open unprecedented opportunities to develop artworks as applications.
I have applied these technologies to recreate the site-specific installation "Lento o Tempo Muda em Poço do Canto" (first shown in Portugal in 2022) as a "home edition." Using Apple TV, the application presents four videos on a split screen with an immersive spatial audio track, with all media files playing and looping asynchronously.
This project demonstrates how recent technologies open possibilities for developing artistic works to be experienced in domestic surroundings that, from a programming standpoint, might appear simple but, depending on the quality and composition of the media content, can still offer rich experiences.
Poster publisched in https://mishmash.no/assets/events/kristiansand2026/mishmash-opening-conference-2026-abstracts.pdf
This presentation discusses the National Library of Norway's migration from MAVIS, a legacy audiovisual collection management system, to Axiell Collections. While initially driven by the announced end of support for MAVIS, the project evolved into a broader exercise in metadata modelling and collection management for audiovisual heritage.
Particular attention is given to the challenges of representing recorded music within frameworks based on RDA and the Library Reference Model (LRM). The presentation examines different understandings of musical works in library, archival, and music industry contexts, and discusses how compositions, recordings, albums, and production materials can be modelled and linked within a shared system.
The talk outlines the implementation adopted by the National Library, including parallel work structures for compositions, recordings, albums, and music production archives. Rather than enforcing a single hierarchy, these entities are connected through relationships that better reflect the complexity of music creation, publication, and preservation.
Finally, the presentation reflects on the broader implications of metadata standardization, interoperability, and shared infrastructure for music heritage collections, and considers how these issues relate to future developments in digital musicology and AI-driven research.
AI has been part of games since ghosts started chasing Pac-Man. This talk traces the shift from traditional game AI to generative AI, explores how it is reshaping different areas of game development, and considers its impact on teaching and learning. How should game development education respond to a technology that remains contested among developers, students and educators?
For centuries, knowledge has typically begun with curiosity. Questions came before answers, and discovery depended on uncertainty, exploration, and dialogue. That sequence is now beginning to change. As generative AI becomes an increasingly common starting point for learning, research, and creative work, answers often arrive before curiosity has the opportunity to develop. This raises a broader question than whether AI can generate knowledge: what happens when curiosity is no longer needed to begin creating it? This study explores how the widespread use of generative AI may reshape the human processes through which knowledge is produced. Rather than focusing on the capabilities or limitations of AI systems themselves, it examines how immediate access to AI-generated responses may influence questioning, exploration, collaboration, and intellectual risk-taking. Drawing on perspectives from education, law, philosophy, and human-centred technology, the project develops an interdisciplinary framework for understanding curiosity not simply as a personal characteristic but as an essential condition for creativity and knowledge creation. The central argument is that the most significant transformation brought about by AI may not be the automation of knowledge production, but the gradual reconfiguration of how knowledge begins. If curiosity becomes less necessary, the practices that sustain original thinking, collaborative inquiry, and meaningful discovery may also begin to change in subtle but important ways. By shifting attention from what AI produces to how it reshapes the conditions under which humans create knowledge, the project offers a new perspective on the future of learning, innovation, and humancentred technological development.
This chapter explores the persistent gap between human rights law as a formal normative order and lived truth as it is experienced, expressed, and contested in public space, taking street art as a critical point of entry. While human rights frameworks promise protection for expression, participation, and access to truth, they often remain ill-equipped to account for the ways in which truth becomes publicly present-or fails to do so-in everyday life. Street art emerges within this gap as a spatial and visual practice through which lived experiences are rendered visible when institutional channels prove insufficient or unresponsive. Rather than treating such practices as marginal or exceptional, the chapter argues that they illuminate a structural tension at the heart of contemporary democratic systems. By situating street art within debates on lived truth, expressive inequality, and democratic legitimacy, the chapter calls for a rethinking of how human rights law engages with forms of truth that unfold beyond authorised institutional frameworks.
Posteren presenterer AI as Environment, et teoretisk rammeverk som undersøker kunstig intelligens som et komplekst miljø av sammenvevde materielle, energetiske og algoritmiske systemer, snarere enn som et avgrenset verktøy eller medium. Med utgangspunkt i økologisk og prosessorientert medieteori utforskes tre gjensidig forbundne økologier: den infrastrukturelle, den epistemiske og den operasjonelle. Rammeverket undersøker hvordan KI former betingelsene for persepsjon, kunnskapsproduksjon og kulturell praksis, og hvordan kunstnerisk forskning kan bidra til kritisk undersøkelse og intervensjon i disse systemene. Posteren ble presentert på MishMash sin åpningskonferanse i Kristiansand i september 2026.
Mette Nyseter
;
Andrew Michael Davidson
;
Christian Schüssler
This artistic research project explores whether AI-based audiovisual transformation can offer a viable and ethically responsible way to anonymise real people in documentary film, protecting participants’ identity while preserving perceived authenticity in their human presence on screen.
Conventional anonymisation methods such as heavy blurring, pixelation and strong voice distortion often strip away facial expressions, body language, emotional intensity and relational dynamics that give the material its meaning, and can introduce stigma or associations with crime, danger or shame. The project investigates whether AI can protect identity without erasing the embodied expressions, emotional nuances and relational dynamics that make documentary material meaningful.
Through a pilot study, we test selected AI tools and approaches, examine how well they retain expression, gesture, interaction, voice and environmental context, and identify key technical, ethical and legal challenges. The outcomes include proof-of-concept examples, a preliminary anonymisation workflow, and a framework for a larger multidisciplinary research project on ethical, cinematic and technological approaches to anonymisation in documentary practice.
The project draws on extensive video recordings from an ongoing artistic research project on role-play and embodied methods in documentary practice and education at the TV School in Lillehammer, providing a concrete use case through which the possibilities and limitations of AI-mediated anonymisation can be explored. You can read more about this project here: https://www.inn.no/english/research/research-projects/Role-playing-for-documentarists/
RESEARCH TEAM
Mette Nyseter – Project leader, University of Inland Norway, the TV School
Andrew Michael Davidson – Research advisor, University of Inland Norway
Christian Schüssler – Industry collaborator, Reimagine AS
Artificial intelligence is changing not only how journalism is produced and distributed, but also how journalists experience risk. Traditional approaches to journalist safety have focused mainly on visible threats such as violence, legal pressure, surveillance, and censorship. Yet digital platforms introduce a quieter form of vulnerability. Algorithms can reduce the visibility of reporting, misclassify sensitive content, expose journalists to coordinated harassment, or influence professional choices without journalists knowing how or why these decisions were made. This study develops the idea of “algorithmic safe havens” and “digital traps” to examine this tension. AI can provide protection by detecting threats, supporting secure communication, and helping journalists reach audiences. At the same time, opaque ranking and moderation systems can create new forms of control and encourage journalists to adapt or censor themselves in anticipation of algorithmic decisions. The paper focuses particularly on invisible censorship, algorithmic self-censorship, and unequal protection across languages and contexts. It argues that journalist safety can no longer be treated only as protection from external threats. It must also include the digital environments in which journalism now takes place, with greater transparency, meaningful appeal mechanisms, and accountability built into their governance.
The growing institutionalization of grassroots esports within schools and universities raises a fundamental normative question: should competitive digital play be treated merely as extracurricular activity, or as an emergent site for the realization of human rights in the digital society? This study advances the latter claim. Drawing on international human rights law-particularly the right to education and the right to participate in cultural life under the International Covenant on Economic, Social and Cultural Rights- it argues that grassroots esports constitute a contemporary arena in which these rights are enacted, contested, and reshaped.
First, the study reconceptualizes the right to education as extending beyond formal curricula to include digital literacy, collaborative competence, and algorithmic awareness. Structured esports environments cultivate strategic reasoning, collective problem-solving, and socio-digital skills essential for meaningful participation in technologically mediated societies. When embedded in academic institutions, esports can function as pedagogical infrastructure aligned with states’ obligations to ensure accessible, adaptable, and relevant education.
Second, the study develops a doctrinal account of esports as cultural participation. Competitive digital play represents contemporary cultural production and community formation. Exclusion from these spaces- through gender-based harassment, disability inaccessibility, algorithmic bias, or digital divides- should be understood not merely as social disadvantage but as differential access to cultural rights, engaging principles of equality and non-discrimination.
Third, the study identifies governance risks, including opaque platform rulemaking, data extraction, and infrastructural inequality. It proposes a “human rights by design” framework for academic esports, positioning universities as duty-bearers responsible for safeguarding dignity, equality, and procedural fairness in digitally mediated competition.
By reframing grassroots esports as a laboratory for digital human rights realization, the paper offers a normative foundation for human-centered governance in emerging competitive digital ecosystems.
Murtaza Mohiqi
;
Gagandeep Kaur
;
Mohammad Kazem Mohaqei
The EU Pact on Migration and Asylum introduces interlinked legal and technological reforms aimed at enhancing efficiency, harmonisation, and control in European migration governance, with biometric identification, data retention, and system interoperability at their core. While the Pact has been widely analysed in relation to border management and asylum procedures, its implications for children-particularly as subjects of biometric data processing-remain under-theorised. Focusing on the reformed Eurodac Regulation and the Screening Regulation, this study offers a doctrinal and analytical examination of how the Pact reconfigures childhood within EU migration law through the mandatory collection, long-term retention, and cross-system exchange of children’s biometric data from the age of six. Situating these measures within international and European children’s rights law, data protection law, and EU administrative law, the paper assesses whether the Pact’s regulatory design adequately reflects core principles such as the best interests of the child, necessity and proportionality, purpose limitation, and individualised assessment in migration-related decision-making. The study adopts a comparative and socio-legal perspective on the governance effects of digital infrastructures. This perspective underscores how technological systems introduced to address specific regulatory objectives tend to acquire a degree of institutional permanence and functional expansion, thereby reshaping legal subjectivities over time. In the context of the Pact, this raises the question of whether migrant children are increasingly constructed not primarily as rights-holders requiring enhanced protection, but as data subjects integrated into a security- oriented information architecture.
Methodologically, the paper combines close doctrinal analysis of the Pact’s legislative instruments with insights from legal technology scholarship on datafication and automation
in public administration. It argues that, while the Pact acknowledges children as a category of vulnerability, it provides limited guidance on how heightened safeguards should operate in practice within biometric and digital systems. The paper concludes that without a more explicit and operational integration of children’s rights norms into the design and implementation of biometric infrastructures, the Pact risks normalising a technocratic approach to childhood in migration law that is difficult to recalibrate once entrenched.
As artificial intelligence (AI) systems become increasingly embedded in migration governance, asylum procedures, visa processing, and public service delivery, questions surrounding trust, legitimacy, and accountability have become central to contemporary debates on responsible AI. Existing frameworks for trustworthy AI, however, remain predominantly shaped by top down technical and regulatory approaches that prioritize abstract principles such as fairness, transparency, and explain ability, often without adequately engaging with the lived realities of migrants and other vulnerable communities. This paper argues that such models risk producing institutional forms of trust that fail to resonate with those most directly affected by automated decision-making systems.
The study develops the concept of “trust from below” as a human rights-oriented framework for understanding and evaluating trustworthy AI in migration contexts. Rather than treating trust as a purely technical outcome or regulatory benchmark, the paper conceptualizes trust as a socially situated and relational process shaped by procedural justice, historical marginalization, linguistic accessibility, and the ability of affected individuals to challenge or contest automated decisions. In this respect, distrust toward AI systems is understood not merely as a matter of perception, but as a reflection of deeper structural inequalities embedded within digital governance infrastructures.
To ground this argument in contemporary practice, the paper examines emerging uses of AI assisted risk assessment systems, biometric border technologies, and automated asylum processing tools within European and transnational migration governance contexts. These developments illustrate how algorithmic systems are increasingly shaping administrative decision-making processes and raise important questions regarding procedural fairness, transparency, accessibility, and the ability of affected individuals to meaningfully understand and challenge automated outcomes in legally and technologically complex environments. Building on interdisciplinary scholarship in human rights law, migration governance, critical AI studies, participatory design, and digital constitutionalism, the paper argues that existing universalized models of trustworthy AI frequently overlook context-specific vulnerabilities across different socio-legal settings.
In response, the paper proposes a human rights-based framework grounded in three interrelated principles: participatory inclusion of affected communities throughout the AI lifecycle; context-sensitive fairness and accountability mechanisms; and accessible avenues for contestation, explanation, and redress. Particular attention is given to the importance of culturally and linguistically responsive design practices, especially in migration settings characterized by legal precarity, informational asymmetries, and unequal access to digital literacy.
By shifting the focus from formal compliance toward socially embedded trust-building, the paper contributes to ongoing debates on trustworthy and responsible AI by advancing a
normative and practice-oriented reorientation: from designing systems presumed to be trustworthy in principle toward fostering trust as a lived, participatory, and institutionally
accountable outcome. Ultimately, the paper argues that trustworthy AI in migration governance requires not only technical safeguards and regulatory oversight, but also sustained engagement with human dignity, social legitimacy, and justice in high-stakes decision-making environments.
Vulnerability has become a structuring principle of contemporary EU asylum law, shaping access to procedural guarantees, reception standards, exemptions from border procedures, and administrative prioritisation. At the same time, asylum systems are increasingly digitalised through standardised screening, interoperable databases, and decision-support tools. This paper examines how these infrastructures reshape the legal meaning of vulnerability and asks: what happens when suffering becomes administratively legible through data? he study conceptualises vulnerability as a governance interface between human rights law and migration management. In digitalised systems, experiences of harm, trauma, dependency, and risk must be translated into structured indicators. This translation changes how vulnerability is recognised: from narrative and contextual interpretation toward standardisation, comparability, and calibrated categorisation. The study then considers the consequences for individualised assessment, transparency, evidentiary standards, and meaningful human oversight. Drawing on doctrinal analysis of EU asylum instruments and emerging Pact mechanisms, alongside comparative insights from digital governance practices, it introduces the concept of the “datafied vulnerability paradigm” to capture the shift from experiential recognition to structured classification. It argues that the central challenge is not digitalisation itself, but ensuring that technological mediation does not narrow the normative horizon of asylum protection.
The traditional Norwegian Hardanger fiddle repertoire is a rich oral heritage. Existing
score transcriptions are scarce and rarely capture essential performance features,
particularly ornamentation, microtonal intonation, two-string textures, and systematic
beat unevenness. We develop an innovative AI-assisted transcription framework
producing detailed, performance-aligned MIDI representations that preserve fine-
grained timing and pitch.
The neural front-end adapts a state-of-the-art deep-learning model for polyphonic
music transcription to our small but growing dataset of expert annotations. It
progressively learns from expert corrections, providing increasingly robust results, as
reflected by a rising F1 score. The model is complemented by signal processing,
musicological knowledge, and cognitive principles addressing repertoire-specific
challenges such as non-standard tuning, sympathetic-string resonances, sustained
drones, and symbolic note offsets. Acoustic analysis refines pitch beyond the
conventional twelve-tone MIDI grid, while tune-specific pitch distributions organise the
performed notes into musically meaningful categories.
These results are transformed into a new score representation proportionally aligned
with the recording. Colours make stable “micromodal” pitch categories visible and
highlight ambiguous notes or pitches lying outside those categories. This makes the
score useful not only for reading and performance, but also for checking AI output,
exploring intonation and texture, and identifying recurring patterns.
This expert-in-the-loop workflow drastically reduces transcription correction time, with
the aim of progressing towards full automation. The resulting transcriptions and scores
will support performance, education, archive exploration, corpus research, and future
AI methods for musical heritage. This original hybrid AI approach also demonstrates
the limitations of purely machine-learning approaches to music transcription.
The Norwegian Collection of Folk Music was founded not simply to
preserve tradition, but to enable scientific study based on sound rather than manuscripts
alone. Its first director, Olav Gurvin, even collaborated with the Institute of Physics on
an early melody-writing device. Field collecting later documented regional practices, while
editorial work produced seven volumes of Hardanger fiddle tunes followed by volumes
for standard fiddle. Since 2014, the original reel-to-reel tapes have been digitized and a
searchable catalogue placed online.
The collaboration with RITMO first addressed long tapes containing sequences of tunes
separated by spoken announcements. Existing metadata were combined with energy
measurements and Apple’s estimates of speech, music, violin, and singing. Rule-based
processing proposed boundaries, which a research assistant verified and associated with
catalogue information. This divided 343 one-to-two-hour tapes into individual tunes,
currently disseminated, as a first prototype, through the Open Science Framework.
19
The second part focused on automatic transcription of orally transmitted Hardanger fiddle
music. Rather than treating a conventional score as the sole endpoint, the team aims
to produce time-aligned representations retaining note onsets, offsets, and continuous-
frequency pitch, eventually aligned with notation or displayed as a proportionally spaced
score. This matters because flexible beat durations and microtonal intonation can be ob-
scured by conventional notation. A compact model adapted from the piano-transcription
system Transkun proposes polyphonic events. Transfer from piano transcription is facil-
itated by the instrument’s stable pitch and frequent two-string textures, but the small
dataset—119 manually annotated tunes—requires a reduced architecture and dedicated
augmentation. Psychoacoustic post-processing refines pitches beyond the 440-Hz equal-
tempered grid and reconstructs symbolic note endings masked by resonance, for instance
by linking melody notes above a sustained drone. The transcriptions can support studies
of intonation, rhythm, and recurrent structure; metadata generation and search; and
learning and performance. Beat tracking and large-scale deployment remain ongoing
work.
Alexander Refsum Jensenius opened Day 2 by welcoming participants to the University of
Oslo and introducing MishMash—the Centre for AI & Creativity, a national consortium led
by RITMO and bringing together more than 200 researchers and practitioners. The centre
examines the possibilities, limitations, and wider consequences of creative AI through seven
thematic work packages. Its activities follow three interconnected approaches: creating
new AI systems and practices, exploring their uses in different contexts, and reflecting
critically on their cultural and societal implications. Jensenius also outlined MishMash’s
role, alongside the National Library of Norway and the Department of Musicology, in
shaping the symposium and situating its focus on music and cultural heritage within this
broader interdisciplinary programme.
Ingrid Romarheim Haugen, Arnulf Mattes, and Olivier Lartillot then introduced Work
Package 6: AI for Cultural Heritage. WP6 starts from the central role of archives, libraries,
and museums in preserving and communicating human culture, and from the rapidly
growing scale of their digital collections. The National Library provides an important
foundationthroughitsextensivedigitisationprogrammeandDigitalHumanitiesLaboratory,
butmanyresourcesremaindistributedacrossnationalandlocalinstitutions. WP6therefore
seeks to improve their discovery, organisation, interpretation, and presentation, while
contributing to a more coherent digital framework connected to initiatives such as the
European Cultural Heritage Cloud.
The work package proposes hybrid AI methods combining machine learning with mu-
sicological knowledge and cognition-based symbolic approaches. Its current activities
include the transcription and segmentation of audio collections, computational analysis
of Norwegian folk music and Hardanger fiddle repertoire, metadata inference, and the
development and interlinking of digital work catalogues. These activities also connect
with the National Library’s digitisation programme and the proposed MusikkData research
infrastructure. Particular attention is given to minority cultural expressions, including
Sámi joik and world chant traditions.
The speakers also proposed several new topics that could be introduced into ongoing
discussions within WP6. These included generating richer metadata from basic music
catalogues through computational analysis; linking music transcription and analysis with
metadata and machine-readable musicological and editorial discourse through semantic-
web technologies; and automatically detecting the reuse of musical material across a
repertoire. Grieg’s Peer Gynt, in connection with the National Ibsen Year in 2028, was
mentioned as one possible setting in which some of these ideas might be developed.
27
The presentation illustrated the broader scope of WP6 through projects involving different
cultural materials and institutions. These included the reconstruction of provenance
and circulation networks behind the fragmented Schøyen Collection; semantic image
search across the collections of the National Museum and other institutions; the Phygital
Lab’s integration of physical and digital museum experiences; and the preservation and
dissemination of craft knowledge from the archive of the Women’s School of Applied Arts
and Crafts. Together, these examples showed how methods developed for one type of
collection may inform work in other areas of cultural heritage.
Finally, the speakers stressed that technological development must be accompanied by
ethical, legal, and institutional reflection. Public cultural institutions were presented
as having a leading role in preventing unchecked or exploitative uses of AI, protecting
minority cultural expressions, and supporting rights management, respectful attribution,
and responsible reuse. The three co-leaders brought complementary perspectives to these
questions: Lartillot through signal processing and computational music analysis, Mattes
through musicological and archival research, and Romarheim Haugen through library
infrastructure and policy. As an opening overview, the session established several of the
themes developed throughout the remainder of Day 2.
Olivier Lartillot presented preliminary experiments in the computational transcription of
chant and yoik. Conceived as a workshop contribution rather than a report on a completed
study, the presentation explored how signal-processing methods might support the analysis
introduced in the preceding talks, while identifying questions requiring further discussion
with musical and cultural experts.
Using the same Hungarian lament recording analysed by Biró and van Kranenburg, Lartillot
first considered how the beginning and end of individual notes might be detected. A
fundamental-frequency (F0) contour provides information about pitch, but does not by
itself reveal the precise temporal organisation of a vocal performance. Note boundaries
are affected by consonants, vowels, low-energy vocal gestures and changes distributed
across the harmonic spectrum. Lartillot therefore proposed tracking the energy of the
fundamental and its successive partials. Coordinated increases or decreases across these
partials can indicate the onset or offset of a vocal sound, including a brief gesture preceding
the principal note. The approach draws on earlier work in which changes in spectral energy
were used to detect onsets in complex musical recordings.
These experiments used simple signal-processing rules, but Lartillot emphasized that
the eventual choice between rule-based and machine-learning methods would depend on
the available annotated data. If sufficiently large training collections were not available,
signal processing could provide an initial segmentation from which annotations could be
corrected and subsequently used to train a model. Such a hybrid workflow would combine
interpretable preliminary rules with iterative expert validation.
The proposed segmentation also complements the global pitch-histogram method pre-
sented by Biró and van Kranenburg. Whereas their approach derives the principal scale
positions from peaks in the distribution of F0 values, Lartillot considered first detecting
the temporal extent of each note and then estimating its representative frequency locally
from the stable portions of its pitch contour. Each estimate could subsequently be related
to the scale degrees obtained from the recording as a whole. For long performances
in which tuning changes progressively, a single global histogram might be insufficient;
the pitch reference would then need to adapt over time. This extension remained to be
implemented.
Lartillot next presented an exploratory analysis of a yoik recording supplied by Camilla
Brattland. Its spectrum showed substantial activity beyond the fundamental, including
prominent partial and formant regions suggestive of a diphonic vocal quality. These timbral
features would ultimately need to be represented alongside pitch. Even the F0 contour
36
alone revealed rapid movement and extensive oscillation around apparently recurrent pitch
levels. Some events appeared comparatively stable, whereas others approached or circled
a level without settling on it. The neutral letter labels used in the demonstration were
therefore provisional analytical markers rather than Western scale degrees. Slowing down
the recording made these details easier to hear and inspect, but Lartillot refrained from
proposing a definitive segmentation or categorization before discussing the results with
yoik experts.
The ensuing discussion broadened the notion of transcription beyond conventional staff
notation. Continuous transitions, approaches from above or below, oscillations and the
presence or absence of clear note boundaries may carry information that a sequence of
discrete pitches cannot preserve. Participants also asked how computational results could
be made accessible without relying on visual representation, as well as to tradition bearers
who do not use Western notation. Lartillot suggested synchronizing analytical information
with the original recording, enabling rapid auditory comparison between passages, and
sonifying aspects of the analysis. The presentation thus framed transcription as an
iterative, expert-guided process in which computational representations should remain
connected to both the sounding material and the ways in which the tradition itself
distinguishes musically meaningful features.
Olivier Lartillot’s second talk on Day 2 addressed music analysis at two complementary
scales: intensive, investigating the internal structure of individual pieces, and extensive,
comparing large collections to identify related tunes and characterise musical styles.
Sequential patterns provided the connection between these scales, linking detailed motivic
analysis to corpus organisation, retrieval, and metadata inference.
The talk opened with a visualisation of a Hardanger fiddle performance presented at
the MishMash opening event. Conventional notation indicates repeated bars, but the
boundariesofmusicalpatternsdonotnecessarilycoincidewithbarlines, andshortermotifs
can recur within larger repeated passages. A paradigmatic representation rearranged
the score so that corresponding patterns aligned vertically, making these relationships
explicit. Applied to Bach’s Contrapunctus XIV from the Art of Fugue, the same approach
revealed how the counter-subject draws on recurring fragments whose arrangement varies
between appearances. A demonstration from the MusicLab Copenhagen concert with
the Danish String Quartet showed how live pitch tracking could animate this analysis
during performance, while also illustrating the difficulty of following several simultaneous
contrapuntal relationships.
Lartillot emphasised that transcription and analysis are closely interconnected. Moving
from audio to notation requires note detection, pitch estimation, beat tracking, and
rhythmic quantisation, each of which already involves interpreting musical organisation.
The opening of Mozart’s G-minor symphony illustrated how repeated accompaniment
figures support the inference of beats and metrical groupings. Patterns can themselves
consist of smaller patterns, requiring models that accommodate several levels of or-
ganisation. Debussy’s Syrinx illustrated exact repetition, altered endings, and rhythmic
correspondences despite changes in pitch. It also showed how a new motif can emerge
within the ending of a repeated passage. Such relationships suggest that stylistic analysis
should consider not only which motifs recur, but also how repetitions are organised and
connected.
Earlier work with Mondher Ayari on Tunisian istikhbâr improvisation provided a connection
to the morning’s discussion of recitation. Here, modal identity depends on characteristic
melodicfiguresaswellasscalestructure,makingtherecognitionofmotifsthroughvariation
an important analytical task. For Norwegian folk music, Lartillot proposed extending
this approach to the construction of phrases and larger forms, the relation between
ornamentation and core melodic structure, and the development of an ornamentation
taxonomy. Further dimensions included modal organisation, rhythmic figures, expressive
timing, and pitch intonation in relation to melodic context.
42
At the corpus scale, the proposed aim was to discover correspondences between pieces
without requiring predefined categories, identifying shared patterns, possible quotations,
and tune families. These relationships could help characterise genres, dances, musicians,
regions, or periods. An interconnected catalogue could offer an overview placing related
tunes near one another, links between corresponding passages while browsing individual
recordings, and searches formulated around musicological features. Collaboration with
Sámi experts at the University of Tromsø raised related possibilities for segmenting yoiks
within long recordings, identifying distinct yoiks and performers, and mapping collections
throughcontentanalysis. Bothcontinuousaudio-derivedrepresentationsandmorediscrete
transcriptions were considered.
These proposals formed part of the research agenda for MishMash WP6. Pattern-
based comparison could also support metadata inference, using labelled items to suggest
attributesforrecordingswhosedocumentationisincomplete. Lartillotenvisagedcombining
rule-based modelling with machine learning in a framework of interdependent transcription
and analysis processes.
Discussion emphasised that the intended users must guide the choice of representations
and analytical categories, through close collaboration with experts and musicians. Po-
tential uses included research, teaching, and public engagement. A concrete example
concerned fiddle tunes in which successive variants retain an ending while changing their
beginnings: conventional repeat notation handles this awkwardly, whereas alternative
graphical representations could make the shared structure clearer.
MishMash, the Centre for AI and Creativity, is Norway's national research centre for AI in the creative fields, funded by the Research Council of Norway from 2025 to 2030. In this presentation for the Association of Nordic Music Academies I introduce the centre and what it can offer higher music education: research on AI in composition, performance and music education, tools and datasets, and ways for conservatoires to take part.
Human interaction with large language models (LLMs) is commonly framed as information exchange, optimized for efficiency and correctness. This paper proposes a different framing: keyboard-as-instrument, where typing to an LLM becomes a form of musical practice. Drawing on the lineage of keyboard instruments, the action-sound framework, and the aesthetics of failure in jazz improvisation and glitch music, the paper introduces vyping (typing without looking) as a technique that treats degraded input as an expressive gesture. When a person vypes, inter-onset intervals and keystroke dynamics carry embodied information that the LLM can integrate, in a process closer to musical comping (responsive accompaniment) than to improvisation proper. Three artifacts (ll00mtube, clix-vibe, and KeyMeter) illustrate this approach, making the temporal and sonic dimensions of typing visible and audible. The paper argues that typos can function as grace notes: productive deviations that open new directions. It proposes reverence (nuance, grace, and respect) as a design orientation for human-AI interaction, while acknowledging the discomfort of caring for a tool entangled with extractive labor and concentrated corporate power.
Dual-encoder vision-language models systematically fail at negation, scoring "a cat without a dog"' nearly identically to "a cat with a dog.'' We show that this failure is inherent to the single-embedding framework. Under contrastive training, single-embedding methods can at best ignore the negated concept (dilution) but never penalize its presence (rejection). We propose Know-Thy-Not (KTN), a training-free method that introduces an image-side rejection dimension to penalize the negated concept directly in the visual representation. We also introduce COCO-Confusion, the first benchmark requiring genuine rejection to succeed. Across 9 backbones, KTN improves COCO-Confusion mAP by up to +25.6 over vanilla VLMs and +9.6 over the strongest existing method.
The lecture focuses on how AI systems are built, what they reflect from human data, and why their outputs require critical judgment. In addition, it explains the development of LLMs and VLMs through large-scale data collection, pre-training, fine-tuning, and human-feedback-based alignment. It highlights both the strengths of AI, such as summarizing well-established knowledge and generating standard code, and its limitations, including biases, hallucinations, and sycophancy. The main takeaway is to use AI as an augmentative tool rather than a replacement for human judgment, while remaining aware of bias, questioning plausible but false information, and developing AI literacy as an essential skill.
Changkyu Choi
;
Martine Hjelkrem Tan
;
Marius Aasan
;
Adín Ramírez Rivera
AURoRA researchers co-hosted a public seminar and panel discussion with the Embassy of the Republic of Korea to Norway, held at Litteraturhuset in Oslo as part of the K-Initiative campaign. The evening opened with the talk "Mirrors of Intelligence" by Adín Ramírez Rivera (University of Oslo). The panel then discussed trust, fairness and the societal impact of AI with the audience.
Recent advances in artificial intelligence (AI) have enabled the creation of music that closely resembles human-composed works, raising questions about how listeners perceive and evaluate music.
While prior research suggests listeners often exhibit a bias against AI-generated music, most studies have focused on aesthetic judgments such as liking or quality, leaving the social-emotional dimensions largely unexplored. This is a notable gap, given that music listening is fundamentally a social experience, deeply tied to feelings of connection with both the music and its creator.
The present study investigated how purported composer identity influences listeners' appreciation, social-emotional responses, and behavioral intentions. Using a within-subjects experimental design, 75 adult participants listened to identical human-composed classical music excerpts labeled as either human-composed, AI-composed, or human–AI co-composed. By holding the musical material constant while manipulating composer identity, the study isolated the effects of cognitive and social biases related to AI involvement. Participants evaluated each excerpt on measures assessing appreciation, social connectedness to the composer and the music, emotional responses, and behavioral intentions. Data were analyzed using linear mixed-effects models.
Results showed that perceived composer identity significantly affected most measures, with ratings decreasing as AI involvement increased. Crucially, feelings of connectedness—both to the composer and to the music—were affected, with AI-composed and human–AI co-composed music eliciting significantly lower connectedness to the composer and music than human-composed music. These findings highlight the deeply social nature of music and suggest that AI involvement disrupts not only aesthetic appreciation but also the human connection that makes music meaningful.
Songwriting supports creative and emotional expression and social and emotional development through emotion regulation and wellbeing, self-esteem, and a stronger sense of identity. Recent advances in AI technology have provided access to music creation even for people without musical expertise, through AI song generation. We conducted an exploratory, single session study to examine these outcomes in participants attending public outreach events. 45 participants with varying musical proficiency were instructed to express themselves creatively and emotionally by `composing' a song with an AI music generator. They listened to their generated songs and completed a questionnaire measuring song evaluation, task engagement, emotional expression, agency, authenticity, creativity, and connectedness. These exploratory results suggest that more positive attitudes towards AI are related to greater connectedness, while more musical proficiency is related to lower authenticity and evaluation. Agency partially mediated the effects of emotional expression and engagement on authenticity, creativity, and connectedness. These exploratory results can be leveraged for future confirmatory research.
Through a few real-life and imagined scenarios, we want to focus on the questions of authorship and provenience in metadata for musical works. AI-assisted and mixed music, where the role of the machine can be hard to document, challenges classical model of authorship. Indigenous and traditional music, where oral tradition is often the norm and composers and authors can be unknown, also forces us to reconsider how works are catalogued.
AI and algorithmic musicological analysis promises to enrich our metadata descriptions and make it possible to interlink objects in our collections. For the sake of accountability, it becomes important to document the provenience of these new automated analyses.
These questions are of particular interest for The National Library of Norway, which aims at building a new infrastructure for its music collections, in order to unite the different types of material (printed and manuscript scores, archive recordings, and objects obtained via legal deposit) and make it available for research, as much as legally permitted.
Designing a single ontology that will allow for rich and precise metadata across a wide range of musical genres and traditions is a challenging task. Document-based cataloguing practices can fail to fully capture relationships between objects, actors, and more abstract entities. FRBR entities of work/expression/manifestation are heavily influenced by the book-cataloguing practice and are not well suited for music outside of the western classical music canon. We hope to find a solution flexible enough to cover all scenarios, while remaining in line with international standards for interoperability.
The 1951 Refugee Convention is grounded in a legal imagination that conceives refugee status determination as individualized, interpretive, and rooted in human judgment, centred on narrative-based credibility assessment and the ex post legal recognition of risk following flight and arrival. This study argues that these assumptions are increasingly unsettled by the growing use of digital and algorithmic systems in contemporary asylum governance. Informed by scholarship in legal technology and critical data studies, it uses the notion of the computational refugee as a heuristic to examine how algorithmic practices are reshaping refugee protection. It contends that refugeehood is increasingly constituted through anticipatory calculations of future risk, credibility, and governability, as asylum seekers are rendered legible through data extraction, profiling, and probabilistic modelling, often prior to formal legal recognition.
Prompt engineering strategies emphasise clarity, specificity, and well-structured input, on the assumption that large language models (LLMs) require clean input to interpret user intent. But LLMs are pattern interpolators: they “absorb” noise through statistical reconstruction rather than strict symbol matching. Text-centric HCI has studied this tolerance for two decades without integrating it into prompt engineering as a resource. We introduce vyping as an intentional mode of prompting in which one types into LLMs at the speed of thought, without correction. We developed vyping through sustained first-person inquiry. It can be seen as a micro-practice inside a broader post-discursive turn in which meanings increasingly travel as rhythms and patterns through latent-space proximity rather than as precisely-formed symbols. The mismatch between current LLM interfaces and their (human) users has led to two common layers in implementation: (1) an orthographic residue at the interface (autocorrect, spellcheck, tidy-prompt guides), and (2) an impairment parse shadow at the training layer, a form of overalignment in which the model reads orthographic deviation as a sign that something is wrong with the user. Overall, we offer three attunements (ease, reach, and pulse) as alternatives to efficiency-first prompt engineering, and situate the practice within critical HCI traditions of seamful, frictional, and embodied interaction. The critique is structured around a seed prompt that suspends the impairment parse and makes vyping livable; its section headings are taken from the prompt’s own rules. A copy of the seed prompt is provided at the end.
Poster presented at the opening conference of MishMash Centre for AI and Creativity. The poster presents TAO, a persistent artistic-research instrument assembled from open-weight models for multimodal language, music generation and sound generation, running on a single self-hosted machine. Developed with MishMash seed funding as research infrastructure, TAO addresses what AI infrastructure artists and cultural institutions need in order to work critically, experimentally and sustainably.
This article investigates how Artificial Intelligence (AI) assistive music tools influence the creative processes in the production of contemporary popular music. By employing a qualitative research method with semi-structured interviews and auto-ethnography, the paper investigates the role of the music producer in a progressively AI-integrated studio setting. The findings suggest that while AI tools might enhance efficiency and reshape the technical practices of music producers, they often fail to foster genuine creative trust as a collaborator. Through the empirical findings, the research demonstrates that these tools can only operate within their pre-determined parameters, which causes limited capability to engage with sound, context, or artistic expression. Additionally, these tools pose a risk of homogenization and dismantling the artistic identity and decision-making. Ultimately, although AI can efficiently automate tasks, it provides commodities rather than being a creative partner, since making authentic music is tightly linked with human agency, emotional investments, and domain expertise, the dimension that current AI systems cannot replicate.
Globally, in the arena of everyday practices, AI systems have increasingly been causing a profound effect on contemporary everyday life across the globe. From surveillance to the Internet of Things and domestic consumption, from social media interactions to arts and culture, AI is taking over the processing of information to provide new perceptions and augmented experiences. At the core of this reconfiguration of technologically mediated everyday practice are datasets, used to drive and produce new narratives across wider ranges: consumer behavior and patterns, mobility, and immigration, post-human thinking, scientific in(ter)ventions, media production and propaganda, online and offline social formations, etc. In this context, the proposed panel presents the collective research of the newly formed Center for AI Ethics, Aesthetics and Creative Human Operations (CAIEAC) at The Art Academy – Department of Contemporary Art (KMD), University of Bergen, Norway. The centre aims to reimagine the integration of artificial intelligence in creative human operations and to challenge hegemonial AI tropes. By moving beyond the binary of utopian and dystopian paradigms, the centre embraces a “Prototopian" model that envisions continuous improvement in adjusting AI technologies to society. The centre fosters innovative and critical approaches to AI, focusing on computational aesthetics, the expansion of creativity through human-machine collaboration, and the development of new conceptual frameworks informed by art, media theory, and interdisciplinary research based on a reflective and inclusive AI landscape that integrates ethical considerations, addresses biases, and promotes decolonial and Indigenous knowledge into different AI infrastructures for a diverse society. The proposed panel will address the ethical lapses in AI with a regenerative approach to consciousness. Especially engaging with applied AI in the arts, the panel explores the creativity-induced resistance against a monolithic world of AI that populates contemporary everyday practices, as mentioned earlier. The panel will advocate for an ethical and inclusive AI system for the future that is emergent, self-reflective, and self-regulating, with an aesthetic inclination towards artistic experimentation and radical innovations by being susceptible to the questions of social justice and political agency.
This presentation explores the emergence of cloud-based sample platforms like Splice and Loopcloud and consider their potential decline in the age of AI-generated music. Sabrina Carpenter’s hit single “Espresso,” released on April 11th, 2024, put the spotlight on the ubiquitous use of online sample platforms in popular music. The songs’ main intrumental elements-guitars and drums-stem from two loops of producer Oliver’s popular pack “Power Tools Sample Pack III” on the sample platform Splice. This caused a surge of social media posts demonstrating how the instrumental track could be recreated in little under a minute. This form of sampling—both cheap and legal, fast and convenient—represents a distinctly different take on sampling than that commonly associated with “golden age hip hopproductions”, where samples are extracted from existing recordings (Brøvig 2023; McLeod and Dicola 2011; Schloss 2004). Besides being more efficient, these sample platforms’ royalty-free models allow producers to use samples freely without fear of legal action. Our presentation will begin by examining the factors leading to the emergence of cloudbased sample platforms. While the legal aspects might have contributed to the development of sampling platforms, we will also trace the platforms’ rootedness in less explored areas ranging from “library music,” to synthesizer preset cassettes and sample CDs bundled with music tech magazines (Cameron 2020). We will then discuss how these platforms, partly due to their mitigation of copyright issues, have contributed to transforming the sampling practices and aesthetics established by early hip-hop productions while also encouraging other forms of practices (Arrieta 2021). Lastly, we will discuss whether the advent of AIgenerated music technology has the potential to disrupt the sample industry altogether. This work will draw on research on sampling, copyright and musical platforms, and include insights gained from our interviews with producers of sample packs.
Generative AI is changing legal education in ways that are not confined to research, writing or analytical efficiency. It also affects how students encounter the people and experiences from which legal problems emerge. Drawing on work in law and emotion, human rights pedagogy, clinical legal education and critical AI studies, this chapter considers what these changes mean for legal empathy. Generative systems may help students rehearse perspectives, question assumptions and reflect on difficult material. At the same time, a fluent representation of another person’s experience is not equivalent to understanding that experience, still less to assuming responsibility for a professional judgment about it. The chapter uses the idea of emotional anaesthesia to identify a particular educational risk: that technological mediation may make ambiguity, vulnerability, cultural difference and narrative difficulty easier to process by making them less demanding to encounter. It argues for an approach in which AI remains supplementary to, rather than a substitute for, the human encounters through which legal judgment is formed.
As a scholar of human rights and legal technologies with over a decade of teaching and research experience across the Global South and North, I have witnessed how digital tools both expand and constrain access to justice. Legal technologies-ranging from algorithmic decision-making to online courts and predictive analytics-promise efficiency, yet they often reproduce structural inequalities and silence marginalized voices. This paper explores the frontiers where Legal Tech intersects with legal anthropology. Focusing on the lived experiences of individuals and communities subject to digital legal systems, I examine how technology reshapes fundamental notions of fairness, authority, and accountability. Drawing on human rights frameworks and ethnographic insights, I highlight the tension between automation and dignity, standardization and contextual nuance. I argue that anthropology’s interpretive and human-cantered approaches are indispensable for understanding justice in the digital age. By tracing stories from diverse contexts across the Global South and North, this contribution shows how ethnography can illuminate what technical discourse obscures, ensuring that the future of digital justice remains grounded in human rights and lived realities.
As Legal Tech tools increasingly mediate intellectual property (IP) registration, enforcement,
and adjudication, the cultural biases embedded in these systems pose serious risks. Most AI-
driven IP tools are built on Western legal logics, linguistic norms, and data sources. This
limits their capacity to recognize, protect, or fairly evaluate culturally specific forms of
expression and innovation—particularly from the Global South or marginalized communities.
Drawing on my experience as a Legal Tech specialist and professor of human rights who has
taught for over a decade across both the Global South and North, this paper offers a grounded
critique of current AI applications in IP. I focus on how automated trademark classifiers,
copyright enforcement bots, and patent search engines can unintentionally reproduce
epistemic injustice, cultural erasure, and exclusion from national and global IP regimes. The
paper argues for the development of multicultural AI frameworks in IP Legal Tech that
integrate three core principles: localized and multilingual data training sets to ensure cultural
relevance; participatory co-design processes involving underrepresented creators to
incorporate diverse perspectives; and the incorporation of diverse legal traditions, including
indigenous and customary law, to broaden the epistemic foundations of AI tools in
intellectual property. This approach recognizes AI not merely as a tool for legal automation
but as a mechanism that shapes cultural legitimacy, national identity, and access to economic
empowerment. By embedding cultural competence and ethical pluralism into Legal Tech, we
can move toward IP systems that reflect the full fabric of global nations and narratives. By
embedding cultural competence and ethical pluralism into Legal Tech, we can move toward
IP systems that reflect the full fabric of global nations and narratives.
This paper develops the concept of “code-switching for survival” to explain how journalists adapt to contemporary forms of repression. Moving beyond its sociolinguistic meaning, the concept is used to capture linguistic, legal, digital, and strategic practices that help journalists navigate legal harassment, surveillance, censorship, and algorithmic suppression. Drawing on examples from the Global South, the paper examines the use of metaphor, satire, coded language, encryption, pseudonymous publishing, and legal strategies as forms of journalistic resilience. It also introduces the idea of “algorithmic code-switching,” whereby journalists adapt their content to opaque moderation and ranking systems. The paper proposes an integrated legal-tech-ethical framework linking these practices to international human rights standards and responsible AI governance. It argues that journalists should be understood not only as targets of repression, but also as active agents who design strategies of resilience in increasingly complex media environments.
AI literacy is becoming part of everyday legal education as students and teachers increasingly encounter generative AI in research, writing, and classroom work. This presentation looks at how law schools are responding to these changes across different educational settings, with particular attention to experiences from the Global South. Drawing on published work, institutional materials, course information, and student-led initiatives, it considers how local resources, languages, teaching cultures, and access to technology shape the way AI is introduced into legal education. The presentation highlights practical forms of engagement, including classroom experimentation, interdisciplinary projects, and uses of AI connected to access to justice. Rather than proposing a single model, it argues for approaches that are flexible enough to reflect different educational realities. Experiences from the Global South are especially valuable in this discussion, not as isolated examples, but as sources of ideas that can enrich wider conversations about responsible, practical, and socially aware AI literacy in legal education.
Across the Global South, migration governance is increasingly shaped by digital
infrastructures- from biometric identification and facial recognition to algorithmic profiling.
In these systems, the migrant body is transformed into a datafied entity: tracked, categorized,
and governed not only by law but by code. This paper examines the legal and conceptual
implications of this shift, where digital infrastructures fragment the migrant body into
dispersed data points subject to classification and control beyond traditional legal oversight.
Drawing on research in legal technology and human rights, the paper argues that datafication
produces a new mode of governance-embodied algorithmic control-that challenges
foundational legal notions of autonomy, privacy, and personhood. Existing rights
frameworks, premised on a coherent, rights-bearing subject, are ill-equipped to address the
opacity and transnational reach of algorithmic systems. In response, the paper proposes a
multidimensional approach: reframing the right to the body to include its digital dimension;
extending accountability to private and transnational actors; fostering South–South legal
collaboration; centering migrant voices in technological design; and rethinking legal
personhood in data-driven contexts. By confronting the digital fragmentation of the migrant
body, the paper calls for a rights-based and ethically grounded approach to migration
governance- one that recognizes data as part of human dignity and seeks a more just and
inclusive digital order.
The growing reliance on data and machine learning models as core components of military
capability poses a fundamental challenge to the weapon-centric logic of arms control law,
particularly in the context of accelerated rearmament and strategic autonomy initiatives
within the European Union. Traditional disarmament and non-proliferation regimes are
premised on the regulation of material weapons, identifiable platforms, and observable
capabilities. By contrast, contemporary algorithmic warfare increasingly depends on largely
intangible assets-training datasets, model architectures, parameters, and continuous model
updates-that blur the boundaries between civilian and military domains and between public
and private actors.
This article argues that the primary regulatory difficulty raised by military artificial
intelligence lies not merely in normative gaps, but in a categorical mismatch between existing
arms control frameworks and the data- and model-centric nature of emerging military power.
As the EU invests heavily in artificial intelligence to enhance defence preparedness and
technological sovereignty, data and models function as strategic force multipliers without
qualifying as “arms” in a conventional legal sense. This undermines the capacity of existing
regimes to identify regulable objects, define meaningful thresholds of control, and develop
credible verification mechanisms consistent with both security and transparency
requirements.
Adopting an interdisciplinary legal and technological approach, the article explores how data
and machine learning models could be reconceptualised as objects of regulation within arms
control and export control regimes, with particular attention to the EU’s strategic autonomy
agenda. It further examines how legal technology tools-such as data provenance mechanisms,
model auditing, and compliance-by-design architectures-could support novel forms of
verification and accountability without constraining legitimate defence innovation. The
article concludes that sustaining arms control in an era of EU rearmament requires a shift
from regulating weapons to governing the foundational digital resources that increasingly
determine military capability.
The National Library of Norway hosts a range of digital music collections that are currently not readily
accessible and are dispersed across various databases, encompassing formats such as music
recordings, sheet music, and related materials like texts, videos, and interviews. We envision creating
a centralised infrastructure that brings these collections together on a high-standard platform,
adhering to the latest metadata standards to enhance search capabilities and facilitate integration
with international networks. This would allow users, even those querying from other systems abroad,
to access music data and related information seamlessly.
Additionally, we aim to extract musical information from diverse formats—such as audio recordings
and sheet music images—promoting interconnectivity between these modalities. This foundational
work will also pave the way for incorporating emerging AI technologies for music analysis and search
in the future.
To inform our vision, we are surveying music researchers in Norway to gain insights into the needs of
potential users. This infrastructure represents the initial step toward developing more user-friendly
interfaces targeting a broader audience, ultimately extending the impact of music research
A controlled cross-device consistency test of the accelerometers in four Samsung phones (Galaxy A52s 5G, S21, S23 Ultra, S26 Ultra), logged with Physics Toolbox Sensor Suite at a nominal 10 Hz. Two recordings per phone: Still on the floor with triple-clap sync markers - the sensor noise floor Exercises -> standstill -> shaking - working range and saturation.
Software. One implementation of quantity of motion — the average speed of a body part, band-limited to 0.2-5 Hz, in millimetres per second — for optical marker data, body-worn accelerometers and force-plate centre of pressure, together with the readers, resampling rules, scaling measures and alignment methods that surround it.
Foredraget handler om hvordan kunstig intelligens kan brukes i kreative prosesser, spesielt innen kunst og utdanning. KI utfordrer tradisjonelle forståelser av kreativitet, som tidligere har vært definert gjennom originalitet og verdi, og at nye perspektiver også inkluderer intensjon og autentisitet.
Foredraget understreker forskjellen mellom produktivitetsøkning og kreativitet: KI er svært effektiv til å effektivisere produksjon, men dette er ikke det samme som å styrke den kreative prosessen. Når kreative oppgaver automatiseres, kan også verdien og eksklusiviteten av resultatet reduseres.
KI kan brukes i flere deler av kreative prosesser, som idéutvikling, eksperimentering, produksjon og kuratering, og kan fungere både som verktøy, samarbeidspartner eller materiale. Samtidig vil KI ofte trekke mot det generelle og statistisk sannsynlige, noe som gjør menneskelig perspektiv og kunstnerisk intensjon fortsatt viktig. Foredraget fremhever også at virkelig kreativ bruk av KI ofte krever teknologisk kompetanse og programmering, og at uten dette blir bruken lett overfladisk. Hovedpoenget er at KI ikke bare bør brukes til effektivisering, men til å utvide det kreative handlingsrommet og muliggjøre nye kunstneriske prosesser og uttrykk.
This lecture presents how Artificial Intelligence (AI) assistive tools are fundamentally altering the creative workflows within contemporary popular music. By utilizing qualitative data from semi-structured interviews and first-hand auto-ethnographic research, we will explore the evolving role of the music producer in an increasingly automated studio environment.
Core Areas of Investigation:
Efficiency vs. Creative Partnership: The research indicates that while AI significantly enhances technical speed and reshapes production practices, it consistently fails to earn the "creative trust" required for a true partnership. We will discuss why these tools are often viewed as utilities rather than collaborators.
The Limitations of Parameters: A central theme of this lecture is the boundary of AI capability. We will analyse how these systems are constrained by predetermin mathematical parameters, leaving them unable to truly "hear" sound, understand cultural context, or engage in genuine artistic expression.
The Threat of Homogenization: We will address the risk of AI-driven tools dismantling individual artistic identity. Relying on predictive algorithms can lead to a "sameness" in music, where the human element of decision-making is sidelined by automated suggestions.
Human Agency and Authenticity: Finally, we will argue that making authentic music is deeply rooted in human agency, emotional investment, and domain expertise. While AI can efficiently produce musical commodities, it remains unable to replicate the soulful dimension that defines human-led art.d
In this presentation, Alexander Refsum Jensenius explores the practical and theoretical dimensions of running radically interdisciplinary research centres, focusing on his experiences with RITMO and the newly established MISHMASH Centre for AI and Creativity. He introduces the "coffee machine" philosophy, which advocates for physical colocation and social meeting points as essential tools to overcome institutional silos and bridge the diverse research motivations of fields like musicology, informatics, and psychology. Jensenius highlights the innovative potential of this approach through projects that translate artistic research into medical applications, such as using dance analysis software to screen infants for cerebral palsy and investigating the impact of musical stimuli on biological cells. He further details the MusicLab initiative, which scales data collection to full symphony orchestras to study embodied music cognition and human behaviour in real-life concert settings. The talk concludes by introducing MISHMASH, a national consortium dedicated to fostering human-centric AI that integrates artistic practice with technological development while navigating ethical challenges such as copyright and the preservation of cultural heritage.
Trenger vi egentlig musikere, forfattere og filmprodusenter nå som KI kan gjøre jobben? Vi tar diskusjonen om hvordan KI påvirker kreativ næring og kunstneriske prosesser.
Innlegget undersøker begrepet «kunstnerisk intelligens» i lys av dagens KI-utvikling, med utgangspunkt i forståelsen av kunstig intelligens som infrastrukturelle systemer heller enn autonome, intelligente aktører. Det drøfter hvordan kunstneriske praksiser kan synliggjøre og utfordre slike systemer gjennom eksperimentering, deltakelse og kritisk refleksjon. Med eksempler fra samtidskunst vises hvordan kunst kan åpne «den svarte boksen» og bidra til nye innsikter i forholdet mellom teknologi, samfunn og kunnskapsproduksjon. Innlegget posisjonerer kunstnerisk utviklingsarbeid som en form for kunnskapsutvikling med relevans for en KI-drevet offentlighet.
Panelsamtalen «Risiko og motstandskraft i kulturens infrastruktur – det neste tiåret» samlet Camara Lundestad Joof (Dramatikkens hus), Kirsti Mathiesen Hjemdahl (Cultiva), Sveinung Rudjord Unneland (Den uferdige institusjonen), Øystein Strand (Kulturdirektoratet) og Synne Tollerud Bull (Høyskolen Kristiania), moderert av Hild Borchgrevink. Samtalen tok utgangspunkt i hvordan kultur i økende grad formes av infrastrukturer, systemer og distribusjonslogikker, og drøftet risiko og handlingsrom i møte med plattformbaserte medieformer. Det ble særlig lagt vekt på behovet for å utvikle robuste og fleksible kulturinfrastrukturer, samt rollen til offentlig finansierte institusjoner og høyere kunstutdanning i å legge til rette for eksperimentering, kunnskapsutvikling og langsiktig motstandskraft.
Arrangementet Design Agentics: Creative R&D x AI undersøker kunstig intelligens som en integrert del av design- og kunstpraksis, ikke bare som verktøy, men som infrastrukturell og agentisk aktør. Gjennom en studentpresentasjon av egenutviklede AI-baserte verktøy og en påfølgende panel- og rundebordsdiskusjon med forskere og bransjeaktører, tematiserer arrangementet hvordan kreative praktikere kan forholde seg kritisk og operativt til eksisterende AI-systemer. Fokus ligger på spørsmål om handlingsrom, makt og design i møte med ferdigdefinerte teknologiske infrastrukturer, samt muligheter for å omforme, tilpasse eller motsette seg disse. Arrangementet avsluttes med lanseringen av forskningsgruppen CRAITR ved AHO.
Artist and researcher Budhaditya Chattopadhyay will discuss his body of work around sounding and listening to sculpt and give shape to two broad theoretical frameworks, namely Auditory Situationism and Unmedia, within which his works can be contextualised; both ideas are hinting at manifestations of decolonial thought intervening into existing sound studies canon, which Chattopadhyay enters with an outsider, nomadic, and polemical impulse.
Hva skjer med vår opplevelse av virkelighet, kropp, natur og kreativitet når grensene mellom det menneskelige og det syntetiske blir mer flytende?
Samtalen har ikke til hensikt å konkludere om kunstig intelligens er bra eller dårlig for kunsten. Sammen skal vi heller undersøke hva som skjer med oss som skapende mennesker, som publikum, og kanskje også som naturvesener, når teknologien begynner å produsere bilder, idéer og identiteter som tidligere bare kom fra menneskelig erfaring.
Fotografiens Hus ønsker velkommen til en samtale som trekker på kunst, økologi og filosofi for å undersøke hvordan kunstig intelligens påvirker hvordan vi skaper, sanser og møter verden rundt oss på.
Alexander Refsum Jensenius
;
Jesper Mardahl
;
Simon Kimmel
;
Martin Eyerer
;
Roland Sillmann
;
Philipp Grefer
Future-facing cultural ecosystems require new forms of collaboration across public, private, and civic actors. This panel examines best practices in building intersectional institutions and clusters that integrate diverse disciplines, and leaders from art, music, science, policy, to communities, and funding models.
This project explains how I got into biology-related research and the potential of using musical, artistic, and creative methods to get obtain both artistic and scientific results.
AI keeps faking it until it makes it as infrastructure. Vibing effortlessly with wearing language as interface for doomstyle hype, timewarping efficiency, subscription-based fate and a long generative etc.
So, let's talk about something else…
The roundtable moves between creative practice and research, speculation and error, embodiment and idiocy. Beyond ritual denunciation or loving embrace, we open a conversation about the pleasures, traps, politics, and odd possibilities of working with these tools to learn (and act on) how they set the terms. As the slogans dry out: how to resist, how to misuse, how to play inside systems built to smooth everything into an average slop(py) shape. After the refusal, after the fatigue, after the sales pitch and the creative ditch, what else?
Organized by AHO's CRAITR research group
Denne presentasjonen forteller om det nye norske KI-senteret Mishmash. Rundt 200 forskere vil i årene fremover jobbe med både utfordringer og muligheter ved bruk av KI i kunst og kultur, særlig knyttet til opphavsrett, personvern og fremtiden for kreative fag. Målet er å utforske og utvikle ansvarlig bruk av KI i blant annet musikk, utdanning, helse, kulturvern og kreativ næring.
On April 8, 2026, the University of Oslo (UiO) hosted a launch event in the University Aula for MishMash, one of Norway's six new national research centres for artificial intelligence. The primary objective of MishMash is to create, explore, and reflect on AI for, through, and in creative practices.
A presentation held at the annual NORA AI conference about MishMash. The centre will conduct research on how AI promotes and challenges creativity, including issues related to copyright and regulation. The research results are relevant for many actors, and especially for creative industries such as film, music, design and literature, and for public and private actors in health, education and cultural heritage.
Alexander Refsum Jensenius
;
Nils Petter Mørland
;
Ingjerd Egeberg
;
Christopher Pahle
;
Hans Ole Rian
;
Åsmund Færavaag
I likhet med en rekke bransjer, står kunst- og kulturfeltet overfor store endringer preget av digitalisering og fremveksten av kunstig intelligens (KI). Representanter fra statlige teaterinstitusjoner, selvstendige scenekunstnere og forskere møtes til samtale om hverdagen for scenekunstnere og framtida for vår bransje i lys av den teknologiske utviklingen.
The closing session of the SDG conference will also represent an opening – towards provoking thoughts and emotions, plural perspectives and anticipation of futures simultaneously too far and too near to grasp. The evening will bring exhilarating possibilities of ontological shifts, deep listening and embodied perception through collaborative human-technology movement and critical conversations. But let’s not forget, there will also be the comfort of food, laughter and relaxation after two full, invigorating conference days where we have aimed, and hopefully succeeded, to hold complexity without slipping into solutionism or despair.
Alexander Refsum Jensenius presenterer om sin forskning som omfatter blant annet kroppslige dimensjoner ved musikkopplevelse og -utøvelse, som kombinerer perspektiver fra musikkvitenskap, psykologi og teknologi. Hans arbeid har som mål å frembringe både vitenskapelige og kunstneriske resultater.
Fagdagen «KI og musikk» ved NLA Høgskolen i Oslo satte søkelys på hvordan kunstig intelligens påvirker musikkfeltet, med særlig vekt på kreativitet, kunstneriske prosesser, opphavsrett og musikkbransjens utvikling. Gjennom faglige innlegg, samtaler og musikalske eksempler ble blant annet bruk av generativ KI, eierskap og rettigheter, algoritmiske komposisjonsprosesser og forholdet mellom menneskelig og maskinell kreativitet diskutert. Fagdagen var primært rettet mot studenter på masterprogrammet i utøvende musikk, men var åpen for alle ansatte og studenter ved NLA i Oslo. Medvirkende var Ole Fredrik Norbye, Njål Ølnes, Torstein Slåen og Mathieu Lacroix.
Can AI be creative? The talk starts from Boden's definition, novel and meaningful, and shows why the first half is cheap for machines and the second is where the human work remains. It sorts AI into rules, learning, and evolution, traces seventy years of artists working with all three, and argues that the interesting questions concern co-creation rather than replacement: control, intention, and ownership. It presents MishMash, Norway's national centre for AI and creativity, with its seven challenges across performance, health, education, industry, and heritage, and closes with what creative practice teaches scientists: small models, embodied agents, reproducible workflows, and provenance.
Panelsamtale som undersøker hvilken rolle kunst og kunstnerisk forskning kan spille i utviklingen av framtidens kunstige intelligens. Med utgangspunkt i MishMash, nasjonalt forskningssenter for KI og kreativitet, diskuteres hvordan KI ikke bare påvirker kunst og kultur, men hvordan kunstneriske og humanistiske perspektiver kan bidra til å forme teknologien. Samtalen retter særlig oppmerksomhet mot kreativitet, menneskelig erfaring, demokratiske verdier, kunstnerisk frihet, opphavsrett og kulturelt mangfold. Kunstnere, forskere, teknologer og representanter fra kultur og arbeidsliv diskuterer hvordan kunstnerisk og tverrfaglig kunnskap kan bidra til mer ansvarlig, mangfoldig og menneskeorientert utvikling av KI.
Dette samsvarer godt med MishMash sitt uttalte mål om å undersøke KI for, gjennom og i kreative praksiser, og å bruke kunstnerisk utforskning som inngang til kritiske spørsmål om KI og samfunn.
NE #00: New Embeddings of/for Artistic Intelligence was a participatory workshop presented at Ars Electronica 2026 that explored how new forms of artistic intelligence might emerge when artists, institutions, scholars, scientists, and technologists organise around shared needs rather than disciplinary boundaries. The workshop asked how AI might support, rather than determine, such collaborations, and how alternative infrastructures for artistic and interdisciplinary practice might be developed.
Structured as an exchange rather than a conventional presentation, participants articulated what they need in relation to generative AI, what they can contribute, and what they refuse. These contributions included conceptual approaches, technical and institutional infrastructures, artistic methods, knowledge, and resources. Participants clustered around points of convergence and translated these into concrete offers, requests, and possible collaborations recorded in a shared ledger.
The workshop initiated the New Embeddings series as an experimental format for building interdisciplinary connections around artistic intelligence and generative AI through collective exchange and the formation of new “neighborhoods” of practice.
Synne Tollerud Bull
;
Jakob Margit Wirth
;
Gwen Rakotovao
;
Sarah Kordecki
;
Merel Visse
;
Ryan Woodring
;
Janine Davidson
This session invites contributors to share their artistic research approaches and strategies for navigating, resisting, or intervening in times of societal transition. Against a backdrop of rapid policy shifts, the impacts of AI and surveillance capitalism, the deprioritization of the climate crisis, rising authoritarianism, and threats to democracy, the discussion will explore: What can artistic research do in response to the challenges of our time?
Recent breakthroughs in AI—foundation models, ChatGPT, large language models, and vision and audio systems—are often framed as software revolutions. Yet they depend fundamentally on advances in nanofabrication: the extreme miniaturization of logic and memory enabled by EUV lithography and foundries such as TSMC. Large language models are economically and technically viable only because computation, memory, and energy efficiency have been radically reconfigured at the level of silicon. This lecture reframes recent advances in AI as an intensification of mineral memory within the archive. The demand for comprehensive, real‑time, AI‑mediated access to cultural heritage across all media and art forms increasingly binds the archive to mineral and infrastructural regimes of computation. It expands the archive’s material footprint—chips, energy, water, and carbon—posing new ethical, operational, and sustainability challenges for cultural institutions, researchers, and artists. New cross‑disciplinary research alliances—linking archival and humanities research with semiconductor engineering and life‑cycle sustainability modeling—are needed to understand the metallurgical and ecological consequences of contemporary memory systems.
Kunstig intelligens påvirker det meste om dagen, også musikklivet og musikkbransjen. Hva skjer når KI topper hitlistene og alle kan lage sin egen musikk? Vil KI ta over alt, eller blir vi lei og vil ha mer menneskeskapt musikk? Hva er mulighetene for at KI-generert musikk kan bidra positivt til å hjelpe folk med f.eks. ADHD til å konsentrere seg og Parkinson-pasienter til å bevege seg?
An update on the development of MishMash. The centre aims to create, explore, and reflect on AI for, through, and in creative practices. More than 200 researchers investigate AI’s impact on creative processes, develop innovative co-creative AI systems and educational strategies, and address AI’s ethical, cultural, legal, and societal implications in creative domains.
Hvordan ble MishMash-søknaden til, og hva kan konsortiet bak senteret bety for kunstnerisk utviklingsarbeid i Norge? I dette innlegget forteller jeg om arbeidet med søknaden til Forskningsrådets utlysning av nasjonale KI-sentre, om hvordan konsortiet med mer enn 20 partnere ble satt sammen, og om mulighetene for at kunstneriske institusjoner kan bruke senteret til å utvikle egne prosjekter og søknader videre.
I will talk about how developing AI systems over the last 25 years has helped me better understand how humans move, act, think, and feel. Now more than ever, we need to incorporate human perspectives into new AI developments, while also critically exploring new AI-based systems and reflecting on how they change humanity in different ways.
In this workshop, I will lead the participants into exploring their own micromotion through standstill and slow-movement exercises. The focus will be on both physical and mental stillness. In the end, we will explore how mobile phone-based apps can be used for sonic microinteraction.
Hva er forskjellen på en KI-assistent og en KI-agent, og hva betyr den for administrativ databehandling ved et universitet? Med utgangspunkt i egne erfaringer med agenter i forskning og administrasjon går jeg gjennom hva agentene trenger for å virke, hva som er god praksis, og hvordan samskaping med en agent ser ut i praksis, med eksempler fra reiseregninger, lederstøtte, CV-er, personaldata, styrepapirer, nett og arkiv. Innlegget ender i tre ønsker: tilgang til egne data gjennom API og nedlasting, systemer som kan føre dialog, og rom for agenter.
Hvordan kan KI-agenter brukes i undervisning, og hva bør vi holde fast ved? Med utgangspunkt i en åpen lærebok med web-apper, en wiki og et åpent verktøy for videoanalyse i emnet MUS2640 forteller jeg om drømmen fra 2003 om et åpent økosystem for læremidler, og hvordan agenter nå gjør den mulig. Prinsippet er KI utenfor timen og mennesker i timen: agentene forbereder, studentene og underviserne møtes.
South Korea is one of Norway's priority partner countries outside Europe, and few Norwegian research groups collaborate with Korean AI research. In this presentation for the Ministry of Education's special envoy at the Norwegian embassy in Seoul, HK-dir and the Research Council of Norway, I introduce RITMO, the Centre for Interdisciplinary Studies in Rhythm, Time and Motion, and MishMash, the national Centre for AI and Creativity, and outline the centre's plans for international collaboration and the existing contacts among MishMash partners with Korean institutions, research organisations and artists.
Hva kan kunstig intelligens bety for en musikkinstitusjon som setter klassisk musikk i sentrum? Med utgangspunkt i forskningen ved RITMO og MishMash-senteret går jeg gjennom muligheter i klassisk musikundervisning, fra forberedelse og samspill til vurdering: digitale akkompagnatører, analyse av øvingsvideo for skadeforebygging, transkripsjon og notasjon, og KI-modeller trent på komponister og utøvere. Mer maskin kan også bety mer menneske.
When AI enters a creative process, authorship does not disappear, it redistributes. Who is actually doing the creating? In this keynote I present MishMash, the Norwegian Centre for AI and Creativity, show examples of AI in creative work from our research, and argue that the risk is not that AI replaces human creativity but that organisations stop noticing whose judgement a tool encodes and whose is quietly displaced.
Panel discussion during the MusicLab research concert with the hip-hop group Sinsenfist, in which musicians and audience members were recorded with audio, video and motion sensors. Alexander Refsum Jensenius moderated the conversation with researchers Finn Upham and Kjell Andreas Oddekalv about what happens in the bodies of those who play and those who listen during a live performance, and what the concert measurements can tell us.
Is artistic research research? And if so, of what kind? In this short talk, I propose the Artistic Readiness Level (ARL), from an initial artistic impulse to a consolidated and reproducible artistic work with demonstrated impact, alongside the technology and societal readiness levels used in research funding, and ask whether music in a given project is a subject, object or method.
Kunstig intelligens er allerede i musikken, men hva gjør den med oss som lytter, spiller og lager? I dette innlegget viser jeg eksempler fra forskningen på RITMO og det nye MishMash-senteret: forskningskonsertene i MusicLab, små musikkapper som Tap Bloom, Pinch Bass og Firefly, og hva som skjer når KI blir noe mer enn et verktøy.
Foredraget tar utgangspunkt i musikkproduksjonsarkivet som både praksisfelt i musikkindustrien og som kunnskapskilde for forskning. De siste årene har inntektene fra og interessen for eldre musikk, den såkalte katalogmusikken, økt betydelig. I første del av foredraget diskuteres hvordan musikkproduksjonsarkiv inngår aktivt i prosesser knyttet til denne delen av musikkindustrien, gjennom blant annet arbeid med reutgivelser, remastering og forvaltning av kataloger.
I andre del rettes oppmerksomheten mot selve materialet i musikkproduksjonsarkivet i Norge. Gjennom eksempler fra ulike teknologiske perioder vises det hvordan spor av praksis, som merking, formater og arbeidsprosesser, kan gi innsikt i hvordan innspilt musikk har blitt forstått og organisert over tid, og også hvordan materialet kan bidra til å klargjøre den teknologiske utviklingen i landet. Kunnskap hentet direkte fra materialet kan gjennom dette også bidra til å belyse sentrale problemstillinger i praksisfeltet rundt katalogmusikk.
Avslutningsvis inviteres det til diskusjon om hvordan slike perspektiver kan kaste nytt lys over forholdet mellom arkiv, musikkproduksjon og fremveksten av AI-basert musikk.
Current research on the use of computational methods and AI in music libraries and musicology.
This event brings together key members of the European consortium EarlyMuse, in coordination with MishMash – Centre for AI & Creativity, particularly its thematic area on cultural heritage. It presents recent work on digital archives, music information retrieval, and AI-based approaches to musicological and heritage research. Participants will also have the opportunity to engage with new and insightful reviews in computational musicology. The programme introduces and discusses a range of core technologies—including music transcription (from both score images and audio), music analysis, and the transformative potential of rich metadata—while highlighting strategies for interconnecting disparate music collections so that their metadata can meaningfully interact.
I presented my research on embodied music cognition at the first DHN conference in Oslo in 2016. Ten years later, I direct MishMash Centre for AI and Creativity, where the arts and humanities sit inside a national AI programme for the first time. This short provocation asks whether there is still a need for "digital humanities". My claim is that AI is a catalyst for more humanities research, and that the analogue humanities matter more than ever.
What can a concert tell us that a laboratory cannot? Drawing on the Sound Spaces work at RITMO, I reflect on live concert research as a meeting point for acoustics, physics, psychology and physiology, and on what interdisciplinarity costs and gives. I contrast exploratory research with hypothesis-driven designs, and technology readiness with what one might call artistic readiness, using our research concerts as examples.
En innspilt introduksjon på 20 minutter til kunstig intelligens, kreativitet og kreativ KI, laget for emnet FIL1010 Redelig bruk av kunstig intelligens i høyere utdanning ved Universitetet i Oslo. Videoen gir et felles begrepsapparat uten krav om forkunnskaper i programmering eller maskinlæring: hva vi mener med KI i 2026, hva kreativitet er, og hvordan de to møtes i forskning og praksis.
An introduction to MishMash, the Centre for AI and Creativity: a national research centre that explores, creates and reflects on AI through creative practices in music, visual art, literature, games and performance, with more than 20 partners across Norway.
This roundtable discussion focused on the AI infrastructure that Norwegian higher education, students, teachers, freelancers, and the public and private sectors have access to today, and how MishMash can help improve the situation.
In this presentation, I will tell about how I spent the summer on, at least partly, sitting still while working on data of myself and others standing still. This is part of completing my upcoming book Still Standing, which tells the story about how much a human body moves when it is trying not to. The empirical material behind the book spans fifteen years of research: a three-person artistic project begun in 2011, seven Norwegian Championships of Standstill with more than 700 participants, a laboratory study of headphones against loudspeakers, a cross-cultural listening study, a year of daily standstills in 2023, and a group performance at a public library. One of the research questions is how much people move when they try to stand still. My updated number is about 5 millimetres per second, measured at the head. Most of the book is about explaining how I reached that number and what it takes to make people move more (or sometimes less). After my summer adventure, I now have most of the data deposited online in the Oslo Standstill Database, scripts and reports documenting every number in the book, and a new Python toolbox, micromotion, to support such analyses.
TKAI Meetup #4 - Deepfakes
Wed, Mar 25 · 5:30 PM CET
Digs - Mesh Community, Krambugata 2, Trondheim, NO
Deepfakes. Ordet alene vekker både fascinasjon og uro.
Vi har alle sett dem – videoer der Jonas Gahr Støre plutselig holder nyttårstale hjemme hos familien, eller der ansikter og stemmer manipuleres så overbevisende at det blir vanskelig å skille falsk fra ekte.
Men hva er egentlig deepfakes? Hvordan lages de? Og hvor utsatt er vi i Norge for konsekvensene av at hvem som helst kan lage slike videoer?
På vår fjerde Kreativ AI Meetup går vi tett på fenomenet deepfakes – både fra et forskningsperspektiv og fra innsiden av hvordan de faktisk lages.
🍔 Sulten etter jobb? DIGS Café er åpen og selger nydelige sliders både før og under arrangementet!
Viggo Tellefsen Wivestad
Forsker og data scientist, SINTEF Digital
Viggo jobber i skjæringspunktet mellom mennesker og kunstig intelligens, og forsker blant annet på hvordan deepfakes påvirker samfunnet vårt. Han kombinerer teknisk innsikt med en dyp forståelse for de samfunnsmessige konsekvensene av AI-teknologi.
Foredrag: "Hva er deepfake, hvordan påvirker det samfunnet vårt, og hvor utsatt er vi i Norge?"
Viggo tar oss med inn i forskningsverdenen og gir oss et grundig blikk på:
Hva deepfakes egentlig er, teknologisk sett
Hvordan deepfakes påvirker samfunnet – fra politikk til personvern
Norges sårbarhet: Hvor godt er vi rustet til å håndtere dette?
Et foredrag for deg som vil forstå både teknologien, truslene og mulighetene ved deepfakes.
John-Ivar Eidsmo
Innholdsskaper og AI-rådgiver, Smidesang & Lyng
John-Ivar har jobbet med en lang rekke kunnskapsbedrifter i både offentlig og privat sektor, og har blant annet arbeidet mye med internasjonale forskningsprosjekter. Han fungerer som både rådgiver og innholdsprodusent for nasjonale og internasjonale kunder og leder Smidesang & Lyng's satsing på AI-støttet innhold.
Foredrag: "Fra idé til viral deepfake – erfaringer fra nyttårsvors hos Jonas"
John-Ivar tar oss med bak kulissene på hvordan han laget den virale "nyttårsvors hos Jonas"-videoen. Du får se:
Hvordan han faktisk gjorde det – fra prompt til ferdig video
Hvilke verktøy han bruker og hvorfor
Triksene for å komme forbi sensurering og få til det du vil
Utfordringer med flere kjente ansikter i samme bilde
Reaksjonene på videoen – og hva dette betyr for veien fremover
Dette er en praktisk gjennomgang for alle som lurer på hvordan man egentlig lager slike videoer – og hva mulighetene (og farene) er når verktøyene blir tilgjengelige for alle.
Program
Onsdag 25. mars – DIGS, Trondheim
17:30–17:55 — Dørene åpner, barsalg
18:00–18:05 — Velkommen
18:05–18:30 — Viggo Tellefsen Wivestad
18:30–18:40 — Pause
18:40–19:05 — John-Ivar Eidsmo
19:05–19:15 — Q&A
19:15 → — Barsalg & mingling
Participatory workshop and installation at the opening conference of MishMash Centre for AI and Creativity. Thirty-eight participants were grouped by a machine-learning model on the basis of voice notes recorded in four languages; the installation displayed the model's representation of the audience and documented participants' refusals, making the limits of algorithmic grouping a subject of collective discussion.
Invited contribution to the MACHINES panel at the opening conference of MishMash Centre for AI and Creativity, addressing the current use of AI infrastructure in art and design education and research, and the infrastructural needs of creative practitioners in Norwegian institutions.
Presentation of reflections and thoughts around the use of Impsy AI model as an interplay partner in song-performance and improvisation. Part of the session machine agency, shared autonomy, and expressive generative models: perspectives on creative performance with AI
Morten Qvenild: Performer
Charles Martin: Programmer
Ivar Grydeland: Moderator
Ny teknologi har alltid hatt evnen til å forenkle og effektivisere etablerte arbeidsoppgaver, men med det følger også frykten for at funksjoner og individer blir overflødige. Hva skjer når teknologien strekker seg forbi det kjente og utfordrer oss til å utforske noe nytt?
Kunstig intelligens (KI) gir oss muligheten til å stille disse spørsmålene på en ny og spennende måte. Hvilke nye kreative uttrykksmuligheter kan KI åpne for? Hvordan kan kunsten utfordre og subvertere KI? Hvordan kan vi forstå lyd, bilde, video og litterær tekst generert av KI i forhold til autentiske kunstneriske uttrykk? Hvor overlapper og utfyller de hverandre, og hvor er de fundamentalt forskjellige?
Kunsten har et ansvar for å gå i kritisk dialog med KI, for å bidra til en dypere forståelse og debatt om KIs natur og samfunnspåvirkning. Hvordan skal kunstutdanningene ta dette ansvaret, gjennom kunstnerisk utviklingsarbeid, forskning og utdanning?
Vi inviterer til en debatt der vi sammen utforsker disse spørsmålene og mer. Bli med oss i en diskusjon om kreativitet, subversjon og autentisitet i møte med kunstig intelligens. Velkommen!
This workshop is targeted at students and researchers working with video recordings You will learn to use MG Toolbox, a Python package with numerous tools for visualizing and analyzing video files. This includes visualization techniques such as motion videos, motion history images, and motiongrams; techniques that, in different ways, allow for looking at video recordings from different temporal and spatial perspectives. It also includes some basic computer vision analysis, such as extracting quantity and centroid of motion, and using such features in analysis. MG Toolbox for Python is a collection of high-level modules that generate all of the above-mentioned visualizations.The toolbox is relevant for everyone working with video recordings of humans, such as in linguistics, psychology, medicine, human-computer interaction, and educational sciences.
As researchers, we are increasingly using emerging technologies, such as multiple mobile eye tracking, virtual reality, and physiological indicators (e.g., heart rate and respiration) to study professionals’ individual and collaborative work practices. In this workshop, we will demonstrate how these technologies can be provided to professionals in various fields (e.g., education, healthcare, business, engineering, the arts) as a resource for self-reflection, enabling them to study and improve their own practices.
The goal of this workshop is to introduce and facilitate participants to experience novel approaches that use these emerging technologies and tools to help practitioners study their own skills and understand their learning processes. We will also show how focus groups and stimulated recall interviews can encourage and guide professionals to discover ways to incorporate these new technologies into their practice as resources for reflection and growth.
The workshop’s theme is educational practice and research, with a focus on showing how we can offer teachers theoretically driven and empirically validated methodologies for witnessing the micro-processes of collaborative mathematics learning. We will show and discuss how multiple mobile eye-tracking and virtual reality can be used in educational practice and for teacher training and professional development.
This approach and these emerging technologies are applicable not only in education, but also in all other fields of research that aim to study individual and collective practices, as well as professional learning, during the process of acquiring new skills or improving existing ones.
Kunstig intelligens påvirker det meste om dagen også musikklivet og musikkbransjen. Men hva er egentlig KI og hva er utfordringer og muligheter innenfor kunst og kultur? Presentasjonen diskuterer ulike pedagogiske tilnærminger og gir eksempler på hvordan det nye KI-senteret MishMash skal angripe problemstillingene.
A presentation of MishMash, a large Norwegian consortium dedicated to exploring the intersection of AI and creativity. Our primary objective is to create, explore, and reflect on AI for, through, and in creative practices. We will investigate AI’s impact on creative processes, develop innovative CoCreative AI systems, and address AI’s ethical, cultural, and societal implications in creative domains.
A presentation of MishMash, a large Norwegian consortium dedicated to exploring the intersection of AI and creativity. Our primary objective is to create, explore, and reflect on AI for, through, and in creative practices. We will investigate AI’s impact on creative processes, develop innovative CoCreative AI systems, and address AI’s ethical, cultural, and societal implications in creative domains.
Fra leksikon til psykolog og kunstner – på rekordtid har kunstig intelligens blitt en stor del av livene våre. Hvordan bør kunstfeltet møte denne nye teknologien? Vil KI gjøre tegnere og illustratører overflødige, eller er dette snarere en enorm mulighet for utøvende kunstnere?
Den kunstige intelligensens frammarsj fortsetter å være den største kulturelle og samfunnsmessige omveltningen siden den industrielle revolusjonen. Siden fjorårets konferanse har det skjedd vanvittig mye, derfor inviterer vi igjen til en dag fylt med internasjonale nøkkelpersoner, banebrytende prosjekter og nye perspektiver på hvordan KI endrer måten vi utvikler, produserer og opplever film.
Alexander Refsum Jensenius
;
Kristin Bergtora Sandvik
;
Birgitte Grimstad
;
Andreas Hoem Røysum
;
Lars Klevstrand
På konsert føler vi samhold med fremmede, viser forskning. I et kort øyeblikk samler musikken oss. Hvordan kan musikk også samle oss i urolige tider? Flere artister jobber for fred, på ulike måter. Møt noen av dem på Scene Domus Bibliotheca! Hva er det med akkurat musikk som forener oss? Bli med på musikksnakk med artistene Birgitte Grimstad, Lars Klevstrand og Andreas Røysum. Du møter også fredsforsker Kristin Bergtora Sandvik. Her vil musikkprofessor Alexander Refsum Jensenius lede samtalen med ulike spørsmål knyttet til tematikken – kanskje svarer de på ditt spørsmål også? Samtalen er beregnet for et publikum uten faglig bakgrunn i temaet.
I forbindelse med utstillingen Conscientia på Gamle Munch arrangeres en samtale om bevissthet og det å skape.
Utgangspunktet for tema til utstillingen er bevissthet og forhold knyttet til det å skape.
En del av utstillingen tar utgangspunkt i en serie med selvportretter av Randi Wøien basert på MR bilder tatt av hennes eget hode. Bildene representerer et sett av intrikate mønstre i ulike størrelser og sammensetninger og kan refereres til ulike organer, andre vesener og ren natur. Samtidig er de en representasjon av kunstneren slik hun er satt sammen i sitt eget hode. Keramiker Jorid Krosse lager objekter med form og mønster som tar inspirasjon fra naturen og kan relateres til organiske strukturer, hoder og andre vesener. I samspill med maleriene vil de keramiske objektene settes i en relasjon til det kroppslige.
Samtalen om bevissthet og det å skape bruker utstillingen som et utgangspunkt til å få belyst hva den skapende prosessen kan bety for vår egen utvikling og hvordan hjernen fungerer og responderer på skapende prosesser. Tema for samtalen vil være forholdet mellom kunst og bevissthet, om relasjonen mellom maleri og objekt og om hvordan det å skape kunst kan påvirke vår forståelse av oss selv og omgivelsene.
For tiden forskes det mye på hva som faktisk skjer i hjernen når man skaper noe. Vi har fått med oss to av de fremste forskerne på temaet fra universitetet i Oslo.
Alexander Refsum Jensenius er professor i musikkteknologi ved Universitetet i Oslo, hvor han også leder RITMO Senter for tverrfaglige studier av rytme, tid og bevegelse og MishMash Senter for KI og kreativitet. Han forsker på hvordan lyd og musikk påvirker kropp og sinn, bevisst og ubevisst.
Tor Endestad er førsteamanuensis i kognitiv- og nevropsykologi på univeristetet i Oslo og er tilknyttet Ritmo. Han leder FRONT neurolab og forsker på kognitiv psykologi og kognitiv nevrovitenskap med fokus på hjerneavbildningsmetodikk. Pågående forskningsprosjekter omfatter studier av basale mekanismer i oppfattelse av rytme og tid, oppmerksomhet og hukommelse.
Til å moderere samtalen har vi fått med oss Per Snaprud. Han er vitenskapsjournalist og før det hjerneforsker. Han arbeider i det Stockholm baserte magasinet Forskning og Framsteg og har tidligere vært virksom ved Dagens Nyheters og Sveriges Radios vitenskapsredaksjoner. Han er også forfatter av boken «Medvetandets återkomst, om hjärnan, kroppen och universum».
Victoria Johnson er fiolinist, underviser ved Institutt for musikkvitenskap og deltar i ulike forskningsprosjekter ved UiO. Hun har hatt solokonserter blant annet under Festspillene i Bergen, Ultima, Borealisfestivalen og Soundwaves i London. Hennes lidenskap for samtidsmusikk har resultert i flere bestillingsverk og plateinnspillinger. I denne sammenhengen vil hun spille musikk som er direkte komponert til bildene og objektene i utstillingen.
Panel discussion with Robertina Šebjanič, Anetta Mona Chişa, Alexander Refsum Jensenius. Moderated by Benedetta D'Ettorre.
This panel examines the boundary zones where artistic and scientific approaches intersect, entangle, and permeate into one another. Bringing together practitioners working across more-than-human ecologies, technological imaginaries, and embodied research, the conversation will explore how meaningful collaboration can emerge from inter- and trans-disciplinary exchange.
Rather than framing art and science as opposites, we ask how their methods can become mutually generative; how artistic mindsets can expand scientific inquiry, and how scientific perspectives can deepen artistic experimentation. The session aims to discuss questions related to collaborative ethics, shared vocabularies, and the value of embracing “noise” as a catalyst for new forms of knowledge-making.
Kreativ AI Meetup #2 - Vibecoding & fremtidens kreative AI-arbeidsflyt
Wed, Nov 26, 2025 · 5:30 PM CET
Digs - Mesh Community, Krambugata 2, Trondheim, NO
🔥 Oppdatert program for Vibecoding-meetupen – Onsdag 26. november, DIGS Trondheim
Hva er egentlig Vibecoding?
Tenk deg å programmere, designe og bygge prototyper – ikke ved å skrive nøyaktig syntaks linje for linje, men ved å ha en flytende, intuitiv dialog med en AI. Vibecoding handler om å bruke AI som en kreativ partner – du guider den med naturlig språk og raske tilbakemeldinger for å iterere frem løsninger i et tempo som var utenkelig for kort tid siden.
Dette er en game-changer som senker terskelen mellom en god idé og en fungerende applikasjon. Det er en ny arbeidsflyt som endrer måten både utviklere, designere og kreative tenkere jobber på.
🧠 Viggo Tellefsen Wivestad – Forsker, data scientist, SINTEF Digital
15 min: “Hva er vibecoding?”
Viggo jobber i skjæringspunktet mellom mennesker og kunstig intelligens, og forsker blant annet på hvordan KI påvirker utviklingen av programvare.
I dette lynforedraget får du:
En komprimert, engasjerende reise gjennom hvordan “programmering” har utviklet seg
Hvorfor dagens verktøy gjør at vi står midt i et paradigmeskifte
Hvor vibecoding passer inn akkurat nå – og hva det egentlig betyr i praksis
Perfekt for deg som vil forstå vibecoding fra bunnen av.
💡 Kristian Fagerlie – AI-pioner og skaperen av YouTube-kanalen All About AI
Kristian har drevet med Vibecoding helt siden det teknisk ble mulig, og er en sann pioner på feltet.
Han er hjernen bak den populære YouTube-kanalen “All About AI”, hvor han jevnlig deler sylskarpe tips, triks og innsikter om generativ AI – ofte med et ekstra fokus på nettopp Vibecoding.
På meetupen vil Kristian gi oss:
En innføring i hva Vibecoding egentlig er (perfekt for deg som er ny på temaet).
Et eksklusivt innblikk i hans personlige arbeidsflyt og hvordan han bruker verktøyene i praksis.
Eksempler på spennende og kreative prosjekter han selv har bygget med denne metoden.
Dette er en unik sjanse til å lære av en av de fremste på feltet!
📅 Program Onsdag 26. november – Digs, Trondheim
17:30–17:55 Dørene åpner, barsalg
18:00–18:05 Velkommen
18:05–18:20 Viggo Tellefsen Wivestad
18:20–18:30 Pause
18:30–19:20 Kristian Fagerlie
19:20–19:30 Q&A
19:30 → Barsalg & mingling
This thesis investigates how synchronized whole-body movement shapes the experience of listening
to music. The two research questions ask whether floor movement changes the listening
experience, and whether the specific character of the movement matters for that experience.
To address these questions, a one-degree-of-freedom robotic floor was designed and built from
scratch. The system, named Harmonic Audio-Reactive Platform (HARP), lifts a person standing
on it in a vertical rolling motion derived from features extracted from the music. Two motion patterns
were implemented for use in the study: a simple pattern oscillating at the beat frequency,
and a complex pattern that combines two sinusoids whose amplitudes vary with the spectral content
of the music. The hardware, software, and trajectory generation are described in detail in the
thesis, and form a methodological contribution alongside the empirical findings. Six participants
took part in a within-group user study. Each experienced three songs from contrasting genres
(classical, EDM, and metal) under three motion conditions (no motion, simple, and complex),
giving nine conditions per participant. A qualitatively-driven mixed methods design was used.
Semi-structured interviews were analyzed with reflexive thematic analysis, alongside Likert ratings,
amplitude adjustment logs, synchronization measurements, and video observations. Three
themes were constructed from the interviews: Enjoyment, Emotion, and Congruence. Across
all data sources, the most consistent finding is that motion pattern preference is personal. No
single pattern worked for everyone, and the same pattern could feel mismatched or stressful to
one participant and meaningful to another. The movement was mostly received as adding to the
listening experience, with Likert ratings, qualitative themes, and participants’ own descriptions
of real-world use pointing the same way. Synchronization between music and motion shaped the
experience but its specific role could not be settled within this study.
Hører du på musikk i 528 Hz, 432 Hz eller 8D kan det gjøre underverker for den mentale helsen din, ifølge en rekke videoer på sosiale medier. Er det sant – eller bare tull?
Mase på folk om å danse sjølv om dei ikkje vil? Ja, - eller nei takk til KI-musikk? Skal 12-åringen få vite at han syng falskt? Og: Satse på musikken - eller safe? Med Tine Thing Helseth, Alexander Jensenius og Ragnhild Folkestad. Programleiar Kjersti Anderdal Bakken.
Lars Løberg Monstad
;
Olivier Lartillot
;
Karstein Grønnesby
;
Martin Fuglseth Kolden
Bots For Music vil skape et bibliotek over musikknoter, både for musikk som ikke er laget ennå, eller noter som ikke finnes. Appen skal kunne brukes av profesjonelle musikere, til læring og hobby.
I kveld var en rekke personligheter fra musikkmiljøet samlet til debatt ved Universitetet i Oslo. Temaet var «Kan musikk skape fred?». Blant deltakerne var Birgitte Grimstad og Lars Klevstrand , som har underholdt med musikk i flere tiår. Debatten ble ledet av blant annet professor i musikkvitenskap ved UiO, Alexander Refsum Jensenius.
Et stort, nytt forskningssenter er bevilget 173 millioner. Det skal undersøke skjæringspunktet mellom kunstig intelligens (KI) og kreativitet, blant annet ved å se på bruk av KI i kunstneriske prosesser.
This report provides an overview of Artificial Intelligence in European musicological research, addressed to musicologists, librarians, archivists, curators, educators, research-policy officers, infrastructure coordinators, and tool developers. Moving from conceptual foundations to institutional implications, it maps the affordances and limitations of AI-based models and tools for the discipline, while situating them within the European normative landscape defined by the ALLEA European Code of Conduct for Research Integrity, the EU Artificial Intelligence Act, and the UNESCO Recommendation on the Ethics of Artificial Intelligence. The six complementary sections successively address: (1) methodological foundations, clarifying notions of model, tool, and critical posture, and surveying audio-, image-, encoding-, and language-model-based applications such as automatic music transcription, optical music recognition, score-audio alignment, automated analysis, reconstruction, and LLM prompting; (2) data infrastructure, covering FAIR/CARE governance, ontologies and persistent identifiers, European federated infrastructures, auditory digital twins, and risks of fragmentation, representational bias, quality variability, and copyright; (3) a continuation of the disciplinary stakes involved in source and corpus processing; (4) musicological education in the age of AI, including curricular redesign, AI literacy, and design models such as the Two-Lane Approach; (5) professional practices and responsibilities, examining scholarly credibility, intellectual property, academic integrity frameworks, cultural and creative industry collaborations, institutional and legal risks, gender equality, and lessons from organised labour; and (6) best practices, articulating methodological pluralism against technological determinism and formalising principles of ethical reflexivity, human oversight, explainability, reproducibility, and accountability.
Augustin Braud
;
Alexandra Büttner
;
Olivier Lartillot
;
Anna Katharina Laux
;
Laurent Pugin
;
Kristina Richts-Matthaei
;
Torsten Schrade
;
Martha Stellmacher
;
Linnaea Charlotte Söhn
;
Barbara Wiermann
The report presents the results and the outcome of an international musicology workshop held at the Academy of Sciences and Literature | Mainz on 7th- 8th May 2026. The workshop brought together for the first time members of musica* – Consortium en musicologie numérique from the French digital humanities infrastructure project Huma-Num and musicologists from NFDI4Culture, the German Consortium for Research Data on Material and Immaterial Cultural Heritage, as well as experts from the RISM Digital Centre of the Répertoire International des Sources Musicales (RISM), based at the Swiss National Library in Basel, and from the MishMash Centre for AI and Creativity, a consortium of Norwegian research institutions focusing on the application of artificial intelligence in the cultural sector. Furthermore, musicologists and developers from the Specialized Information Service (FID) for Musicology (musiconn) and the Centre for Digital Music Documentation (CDMD) were present.
At the workshop, the group established a new Special Interest Group on Linked Metadata and Interoperability in Musicology. This group will lay the groundwork for future collaborative efforts with a focus on improving the interoperability and alignment of metadata between the aforementioned projects and consortia. The first steps in this process are documented in this report.
This volume gathers the six reports produced between 2025 and 2026 within COST Action CA21161, A New Ecosystem of Early Music Studies (EarlyMuse), the first COST Action ever devoted to historical musicology. Running from September 2022 to October 2026, the Action brought together close to three hundred researchers from some forty countries in five working groups: Education, Sources, Publications, Performances and Policies.
Each report was drafted, discussed and published separately, with its own editorial team, corpus, method and primary audience. Assembled here, they describe a single object from six angles: not "early music" as a repertory, but the ecosystem in which musical scholarship is now produced: the departments and conservatoires that train and employ musicologists, the doctoral frameworks that certify their work, the libraries and archives that hold the sources, the digital infrastructures and computational tools that mediate access to them, the European funding architectures that decide which questions become fundable, and the professional and public worlds in which musical knowledge circulates.
The six reports cover: musicology as data science and heritage science; the mapping of musicology in European higher education and research institutions; musicology and artificial intelligence; historical musicology and European research infrastructures and initiatives (EOSC, ECCCH/ECHOES, the Data Space for Cultural Heritage, DARIAH, E-RIHS); doctoral-level artistic research in music; and musicology within Horizon Europe co-funded partnerships, notably Resilient Cultural Heritage and the emerging Virtual Worlds partnership.
Taken together, they document a growing mismatch between the intellectual vitality of the field and the fragility of the structures through which it is taught, documented, funded and made visible. Recurrent findings include the statistical invisibility of musicology in OECD, Eurostat and UNESCO classifications; the institutional cost of methodological pluralism; and the claim that the field is under-positioned rather than under-qualified for the European agendas now being written. The volume is neither a state of the art nor a synthesis nor a manifesto, but a handover document, a description of the terrain, closing with recommendations addressed to institutions, national funders, European bodies, learned societies and individual researchers.