The 2nd AI Lab for Human and Social Sciences conference

On September 17–18, the Centre d’Économie de la Sorbonne was delighted to host the second conference of the AI Lab for Human and Social Sciences, bringing together researchers from economics, sociology, political science and management to discuss how AI is transforming both our societies and the way we study them.

The conference opened with a keynote by Janet Pierrehumbert (University of Oxford) on the opportunities and hazards of using LLMs in the social sciences. 

The program then covered a wide range of questions:
🔹 Grazia Cecere (Institut Mines-Télécom Business School) asked whether AI image generation services produce a "whitening effect" in advertising
🔹 Donato Ricci (Sciences Po) explored how AI reconfigures ordinary work practices
🔹 Etienne Ollion (CNRS) delivered our second keynote on machine bias and how generative LLMs respond to opinion polls
🔹 Hugo Subtil (University of Zurich) presented neural ideal point models
🔹 Philine Widmer (Paris School of Economics) shared evidence on the political slant of LLMs from a persuasion experiment

One common thread: LLMs are powerful new tools for social science research, but they also carry biases that we need to measure and understand.

A big thank you to the organizers, Fabian Aguirre Lopez, Clément Gorin and Arianna Novaro, to all speakers and participants, and to the Université Paris 1 Panthéon-Sorbonne for its support. See you at the next edition!