MeshUp #2 – Marianne Gunderson

Postdoctoral Researcher Marianne Gunderson (UiB) will present a short lecture on the topic Don’t Worry About Formalities: Prompting as Algorithmic Folklore.
Abstract
As large language models (LLMs) with chat interfaces such as ChatGPT have become available to the public, social media has become host to a proliferation of posts and dedicated communities featuring lay people’s experiments with prompt engineering. While much has been written about prompting as a specialized skill, there is currently very little research on the socio-cultural practices of prompting and the online communities dedicated to these activities and little is known about how non-experts are using prompts to shape their interactions with LLMs. This presentation contributes to the understanding of LLM prompting as a vernacular practice in digital culture. Through digital ethnography of the proliferation and modification of a prompt known as “the eigenprompt” on the social media platform X.com, this paper explores how the assumptions and ideals about LLM intelligence, interiority, and emotion are expressed through the practice of prompting, and how ideals and preferences about LLMs are negotiated and iteratively reformulated throughout online communities. Finally, I argue that the eigenprompt and similar vernacular prompts should be seen as examples of algorithmic creativity that contribute to the construction of new ways of being and relating to digital others.
Bio
Marianne Gunderson is a Postdoctoral Researcher with the ALGOFOLK project (Algorithmic folklore: The mutual shaping of vernacular creativity and automation) at the University of Bergen. Her research focuses on digital media practices, AI imaginaries, and monsters.
Access
MishMash MeetUps are short, informal meetings in the consortium where both early career and established researchers present ongoing projects. The events are open for everyone, but, for security reasons, Zoom links are only provided to people that are affiliated with a MishMash Work Package. If not, please ask for access.