 | , - |
 | Lecture Halls 2 and 3 - Paris Dauphine-PSL University |
From Letter Shapes to Large Language Models: Measuring Information and Meaning
Organized by ACSS-PSL Institute with the financial support of PSL Global Seed Fund
This workshop explores how digital technologies and language models are reshaping systems of representation and analysis. The session brings together leading researchers to discuss how computational methods are transforming our understanding of language, text, and information across multiple scales and disciplines. The first presentation examines the use of Large Language Models (LLMs) in social sciences, focusing on large-scale content analysis and discussing opportunities and epistemological risks linked to this “augmentation” of researchers. The second talk addresses the transition from traditional lexicons to LLMs in economic measurement, showing how these tools transform the production and interpretation of indicators derived from text. The third presentation investigates the informational efficiency of writing systems compared to spoken languages, using measures of compressibility and symmetry applied to typologically diverse scripts. Together, these contributions shed light on the socio-technical and cognitive challenges posed by codes and models that structure scientific and social practices.
Programme:
8.30 - Welcome coffee
9.00-10.15 - Introduction
9.10-10.00 - Etienne Ollion (Ecole Polytechnique) - Augmenting ourselves. How LLMs change Content Analysis, and Social Sciences altogether
10.00-10.50 - Thomas Renault (Paris-Saclay) - Turning text into economic measurement - From Lexicons to LLMs.
10.50-11.20 - Coffee Break
11.20-12.10 - Olivier Morin (Institut Jean Nicod) - Letter shapes do not encode information as efficiently as the sounds of languages do.
12.10-12.15 - Conclusions
12.15-1.30pm - Buffet lunch
