Event Date
AI Large-Corpus Interface Imagining Workshop
Maxime Le Calve (Humboldt University, Berlin)
Joe Dumit (UC Davis)
This workshop will run Friday April 11: 10-12pm, 1-3:30pm PST
9:30am (doors open welcome)
10am-12p Intro to large corpus interactions and ideas
12-1pm Lunch (in person)
1-3:30pm Workshopping Interfaces
In the STS Room (SSH 1246)
It will be on Zoom here
Tools:
The workshop is designed for anyone who has way too much text (or audio) data to deal with and is interested in exploring additional means of provocative encounters with it. Those who have large amounts of interviews, fieldnotes, documents, qualitative surveys, etc. are especially encouraged. You do not need to know anything about AI.
The aim is to share some ongoing work developing workflows that combine Large Language Modes (LLMs like ChatGPT, DeepSeek, Claude, Gemini, Grok) with large corpuses of texts to create ways to send prompts that retrieve potentially important paragraphs of the texts and share them back to you in the form of synthetic coherent-sounding responses. BUT with the additional benefit of footnotes that take you directly back to the paragraph in the document in came from. Ideally this allows you to find paragraphs you had forgotten about, didn’t know existed, or hadn’t thought about seeing connected to your current prompt interest. Even if you have 500 or 5000 interviews and documents.
Ways of doing this exist right now. See NotebookLM for example. Yet their interfaces are intriguing but often annoying. They were not designed with you and your data and your process and your desires in mind. This workshop is to learn about and imagine what those interfaces might look like.
In the morning we will
Introduce and explain these possibilities, generally called Retrieval-Augmented Generation (RAG - AI frameworks that enhances language models by integrating an external retrieval system)
Introduce ways of thinking about and with them in data inquiry, analysis, provocation, and speculation
Live Demonstration of the AI RAG System: Showcasing how AI can engage with a curated corpus (e.g., interdisciplinary publications, prior conversations), dynamically retrieving and generating responses that expand rather than fix knowledge
Hands-on Learning Session, basics of RAG technology, learning how to curate a corpus, frame prompts, and fine-tune AI, interactions for their own research applications.
In the afternoon, we will
Facilitate co-creative thinking with some RAG possibilities. Encourage small groups to work with, brainstorm, and speculatively draw together how they might actually want to use them.
Everyone will have access to some current interfaces to a large dataset of oral histories to think with.
Share these ideas in multiple ways, and start figuring out how to bring these desires into reality.
Rather than positioning AI as a tool of automation, we will engage it as a generative, performative, and participatory device—a means of opening rather than closing experimental futures.