Skip to main content Skip to navigation

CIM News

Show all news items

Seeing Like a Model: Exploring and Understanding Generative AI Methods

Drs. Janna Joceli Omena and Michael Castelle participated in the 3-day conference “Generative Methods — AI as Collaborator and Companion in the Social Sciences and Humanities” held at Aalborg University in Copenhagen, which aimed to explore the multifaceted impact of generative AI as both instruments and objects of research.

 

Dr Omena's contribution, the AI Methodology Map, is a pedagogical device designed to structure, visually represent, and explore Generative AI tools for digital methods-led research. Her research is part of the project "Designing With"* and involves collaboration with co-authors from interdisciplinary backgrounds, including Massimo Botta, Antonella Autuori, Matteo Subet, and Ginevra Terenghi from the University of Applied Sciences and Arts of Southern Switzerland (SUPSI), and Eduardo Leite from NOVA University Lisbon.

Dr. Castelle's contribution focused on generative Large Language Models (LLMs) and the extent to which they should or should not be dismissed by critics in the humanities and social sciences as “just statistics”; he suggests that a closer, on-hands familiarity with these models’ architectures and specific training processes will be necessary in order to help navigate between “revolutionary” hype and pragmatic adoption across various disciplines.

To read Omena et al. and Castelle’s abstracts or listen to songified abstracts transformed by generative AI methods themselves, visit this link: https://aau-generative-methods.vercel.app/.

 

Dr Michael Castelle (at left) presenting his paper, “A Feeling For the (Neural) Organism: What’s So Statistical About Large Language Models?” and (at right) providing further explanations on self-attention in Transformer architectures as requested by the section’s presenters and attendees.

 

The AI Methodology Map. Image source and teaching resources: https://genmap.designingwithai.ch/conference 

 

Below, you can explore the presentations and discover the authors who participated in the 'Seeing Like a Model' section, along with the complete conference program, by visiting this link: https://www.en.ssh.aau.dk/conference-generative-methods-e78062#ee4d31a4-d59b-44e6-9b35-88bf6bdbfe5d 

 

Section:

SEEING LIKE A MODEL: REPURPOSING AND OFF-LABEL USE

Chair: Kristine Bundgaard I Room 3.084a

AI Methodology Map. Bridging Concepts, Technicity, and Applications

Janna Joceli Omena & Eduardo Leite, University of Warwick; NOVA University Lisbon;

Massimo Botta, Antonella Autuori, Matteo Subet & Ginevra Terenghi,

University of Applied Sciences and Arts of Southern Switzerland (SUPSI)

Generative controversy mapping? Lessons from digital methods

Anders Kristian Munk, University of Aalborg

Prompt Compass: A Methodological Approach to Evaluating the Use of Large Language Models in SSH research

Erik Borra, University of Amsterdam

Exploring LLM self-consistency as confidence to build digital methods tools

Mathieu Jacomy, Aalborg University; Emillie de Keulenaar, University of Groningen; Erik Borra, UvA; Andrea Benedetti, Politecnico di Milano; Bernhard Rieder, UvA; Jelke Bloem,

UvA; Sarah Burkhardt, UvA

A Feeling for the (Neural) Organism: What’s so Statistical About

Generative LLMs? Michael Castelle, University of Warwick

*Designing With. A New Educational Module to Integrate ML, AI, and DV in Design Curricula, 2022-2024. Project Partners: SUPSI (project coordination), EPFL and NOVA. © 2022. This project is licensed under CC BY 4.0. Supported by Movetia. This project is financially supported by Movetia. Movetia promotes exchange, mobility and cooperation within the fields of education, training and youth work – in Switzerland, Europe and worldwide. www.movetia.ch

Mon 11 Dec 2023, 16:40