Benoît de Courson

@bdecourson.bsky.social

Postdoc at @lsemethodology.bsky.social. Formal models, ABMs & NLP on newspaper archives. I program gallicagram.com

IP law sucks, but there is a lot we could do within the confines of the law if University Counsel and the people who run libraries decided it was worth it. For instance, you would think it’s a no brainer that faculty can use books *owned by their own library* for nonprofit research.

Why do societies rely on rigid moral rules? Don't lie. Keep your promises. Don't steal. We all recognize there are good reasons to make exceptions. So why are moral rules so often categorical? Our new paper proposes an answer. 🧵

Screenshot of article title page, "Why human societies adopt rigid moral rules"

We've seen this over and over again. Why don't women negotiate more? Because they're punished when they do. Why don't they network with more people? Because they're punished for it. Women don't need to change the way they act. They're acting rationally based on the way the world reacts to them. +

David Ho@davidho.bsky.social · last mo.

Women’s use of AI is perceived as a sign of incompetence, while men’s use of AI is seen as an indication of initiative. 🤦🏻‍♂️

C'est toujours cassé, mais on a rendez-vous avec les équipes de la BnF demain après-midi pour négocier une exemption au cloudflare. Envoyez-nous plein de force !

Gallicagram@gallicagram.bsky.social · last mo.

On a de petits soucis avec l'API @gallicabnf.bsky.social qui s'est mise à bloquer notre IP, ça endommage l'affichage du contexte, les modes par document et cooccurrences... On leur écrit, on cherche une solution, ça risque de finir en rustine genre VPN, mais on va trouver !

The notion that this displaces technical knowledge & makes it useless is I think upside-down It has never been more useful to understand data structures, objects, bias/variance, discriminative/generative, &c

1. Static word embeddings might model culture. But they're fixed. Boo. 2. Dynamic embeddings are contextual! 3. But encoding is computationally expensive. Boo. 4. Decoder only models are efficient! 5. But they seem to collapse all perspectives. Boo. What, then, ought the sociology of culture do? 👇

Tom Einhorn@tomeinhorn.bsky.social · 2mo ago

New essay with @lauraknelson.bsky.social on AI and cultural sociology. Revisiting what made word embeddings so useful for cultural sociology, we argue that contemporary LLMs risk collapsing perspectival differences and discuss provisional solutions that might help recover cultural variation.

So, naive question about these campuses that pay OpenAI $13mil/yr. Why don’t they pour that money into a cluster running the strongest available open-weight model, with API and interactive options. And if the answer is “overhead,” why aren’t we collaborating?

If human editors can’t control who reviews science, it’s no longer peer review — it’s a rubber-stamp machine designed for volume and profit, not quality. I have no intention of attaching my name to it. So I’m out.