Sign in

Jean-François Bonnefon

@jfbonnefon.bsky.social
3.8K followers 1.4K following 540 posts

Behavioral Science of AI @ Toulouse School of Economics @tse-fr.eu Director of @iast.fr Chair of Moral AI @ Artificial and Natural Intelligence Toulouse Institute jfbonnefon.github.io

PostsRepliesMedia
Reposted by Jean-François Bonnefon
Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 16/09/2026
Introducing: Time Machine Experiments? 🚀 ⏳ in which participants 'travel' to the past, and interact with a mind from the year 1930 (simulated by an AI with knowledge cut-off). Preprint: arxiv.org/pdf/2609.15468
2157
Jean-François Bonnefon @jfbonnefon.bsky.social · 09/09/2026
Something very strange happened indeed (From the 2018 tabletop RPG Invisible Sun)
010
Reposted by Jean-François Bonnefon
Jacob Oller @jacoboller.bsky.social · 04/09/2026
people reduce Ruby the elephant to just “the painter elephant” but never talk about her other great passion
companionship.!2! She would scatter grain in her small 20-by-30 foot enclosure to lure ducks into range, then she stomped them to death. This aberrant behavior was
1443382
Reposted by Jean-François Bonnefon
Sung Kim @sungkim.bsky.social · 06/09/2026
When building an agentic AI app, - you do not need memory, - you do not need a router, - you do not need an orchestrator, All you need is a message board!
420719
Reposted by Jean-François Bonnefon
Olympia L K Campbell @olkcampbell.bsky.social · 02/09/2026
New Paper out now in PNAS: Is cousin marriage associated with greater or lower exposure to physical intimate partner violence and coercive control? www.pnas.org/doi/10.1073/...
pnas.org
Cousin marriage and reduced exposure to intimate partner violence and coercive control | PNAS
Cousin marriage, practiced by over 10% of the world’s population, restructures kinship networks by overlapping blood and affinal ties. Theory makes...
1148
Reposted by Jean-François Bonnefon
Levin Brinkmann @levinbrinkmann.bsky.social · 18/08/2026
We worry about homogenisation, but can AI also expand human culture? In a multi-generational experiment we demonstrate AI-induced cultural shifts. Machines lastingly shift human behavior, beyond their presence, just as AlphaGo did to Go. www.nature.com/articles/s41...
nature.com
Propagation and preservation of AI-discovered problem-solving strategies in human culture - Nature Communications
AI can uncover problem-solving strategies beyond human discovery. Here, the authors show that AI-discovered strategies propagate and persist in human populations, producing cultural shifts when non-tr...
12213
Jean-François Bonnefon @jfbonnefon.bsky.social · 17/07/2026
Ouch 😅 I’m saving this here for re-reading every time I feel like complaining about my job
100
Reposted by Jean-François Bonnefon
Zoe Purcell 🕵🏻‍♀️ @zoepurcell.bsky.social · 17/07/2026
How can we collectively contain the threat of dishonest delegation to AI? Our paper "Whistleblowers can contain the unethical externalities of human–AI delegation" is out in PNAS, with @nckobis.bsky.social , Andrew Samuel, and @jfbonnefon.bsky.social. www.pnas.org/doi/10.1073/...
pnas.org
PNAS
Proceedings of the National Academy of Sciences (PNAS), a peer reviewed journal of the National Academy of Sciences (NAS) - an authoritative source of high-impact, original research that broadly spans...
165
Jean-François Bonnefon @jfbonnefon.bsky.social · 20/06/2026
I’m in the weights :)
031
Jean-François Bonnefon @jfbonnefon.bsky.social · 05/06/2026
Hmm how long before we get this on overleaf
businessinsider.com
This AI tool adds typos into your emails — to make it look like you didn't use AI
"Sinceerly" is a Chrome plugin that can make your emails sound like they're not AI-written. It adds typos, or goes full "CEO mode."
3143
Jean-François Bonnefon @jfbonnefon.bsky.social · 02/06/2026
I don't need AI to write my papers, but I desperately need it to submit them to manuscript central
0120
Reposted by Jean-François Bonnefon
Mary Elizabeth Sutherland @meharpist.bsky.social · 29/05/2026
@nature.com is expanding Registered Reports - an article type where the research plan is reviewed prior to conducting the study, and, if it passes peer review, we promise to publish it regardless of the results - to all fields we publish + a broader range of research: www.nature.com/articles/d41...
nature.com
Nature is expanding Registered Reports to all the fields in which we publish
Registered Reports improve the credibility of scientific claims by rewarding big questions, sound methods and solid analyses. They need to become a standard tool in research.
175
Jean-François Bonnefon @jfbonnefon.bsky.social · 27/05/2026
Did I just go to a coauthor’s google scholar profile and sort their publications by year, because I was not sure whether they were alive or dead? Yes I did and I’m happy to report they live.
170
Reposted by Jean-François Bonnefon
Hell Frasier: Bloodline @babadooknukem.bsky.social · 26/05/2026
Just found a subreddit that pairs people with fucked up balls of string with people who love detangling balls of string and it is *very* wholesome
The most jacked up pile of yarn barf you've ever seen in your life. It looks like pastel ramen noodlesSeveral replies from volunteers who want to help
7946791033
Jean-François Bonnefon @jfbonnefon.bsky.social · 25/05/2026
Just saw someone calling Bujold’s Vorkosigan books a cozy series about toxic masculinity. these books are many many things, but this is a truly mind boggling description.
000
Reposted by Jean-François Bonnefon
Sandra Wachter @swachter.bsky.social · 21/05/2026
Still time to apply!! Come & work with me at the Hasso Plattner Institute & @bmittelstadt.bsky.social @cruss.bsky.social @oii.ox.ac.uk If you care about ethical, legal, and social aspects of AI, generative AI and other emerging technologies and EU tech laws, then apply now! tinyurl.com/556jty8t
tinyurl.com
Postdoctoral Researcher (m/f/x) in Technology and Regulation
Postdoctoral Researcher (m/f/x) in Technology and Regulation
0410
Reposted by Jean-François Bonnefon
Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 18/05/2026
Kicking off the Machine+Behavior conference in Berlin today with @jfbonnefon.bsky.social , @azimshariff.bsky.social and 225 participants. Check out the program: machinebehavior.science
091
Reposted by Jean-François Bonnefon
anil oza @aniloza.bsky.social · 14/05/2026
wow — the preprint host, arxiv, is banning authors for a year if they submit papers with hallucinated citations 🤖
Twitter thread from Thomas Dietterich, reading "Attention 
@arxiv
 authors: Our Code of Conduct states that by signing your name as an author of a paper, each author takes full responsibility for all its contents, irrespective of how the contents were generated. 1/
3:03 PM · May 14, 2026
·
66.7K
 Views
Relevant
View quotes

Thomas G. Dietterich
@tdietterich
·
1h
If generative AI tools generate inappropriate language, plagiarized content, biased content, errors, mistakes, incorrect references, or misleading content, and that output is included in scientific works, it is the responsibility of the author(s). 2/
Thomas G. Dietterich
@tdietterich
·
1h
We have recently clarified our penalties for this. If a submission contains incontrovertible evidence that the authors did not check the results of LLM generation, this means we can't trust anything in the paper. 3/
Thomas G. Dietterich
@tdietterich
·
1h
The penalty is a 1-year ban from arXiv followed by the requirement that subsequent arXiv submissions must first be accepted at a reputable peer-reviewed venue. 4/
Thomas G. Dietterich
@tdietterich
·
1h
Examples of incontrovertible evidence: hallucinated references, meta-comments from the LLM ("here is a 200 word summary; would you like me to make any changes?"; "the data in this table is illustrative, fill it in with the real numbers from your experiments") end/"
9458111914
Jean-François Bonnefon @jfbonnefon.bsky.social · 15/05/2026
There is a German SF series which has been published weekly for 65 years, with more than 3,000 stories so far, and 3,000 years of internal chronology 😱
tvtropes.org
Perry Rhodan
Perry Rhodan is a weekly German Science Fiction Pulp Magazine series that has run uninterrupted since 1961, with over three thousand one hundred issues released so far not counting reprints, books, an...
000
Jean-François Bonnefon @jfbonnefon.bsky.social · 11/05/2026
ah yes, i have this problem all the time too
a social media post starting with "I'm thinking about writing a paper for Science Magazine, Nature Magazine, or PNAS (I haven't decided yet) on what"
030
Jean-François Bonnefon @jfbonnefon.bsky.social · 06/05/2026
One productive use of AI to review papers: asking the chatbot to scour through the 100-page supplementary materials to find places where my criticisms might in fact be partially or fully addressed, before submitting the review (assuming I'm allowed to do this per journal policies)
030
Reposted by Jean-François Bonnefon
Matt Novak @paleofuture.bsky.social · 04/05/2026
"On Polymarket, the Journal found, 67% of profits go to just 0.1% of accounts. That means less than 2,000 accounts netted a total of nearly half a billion dollars."
wsj.com
Why Almost Everyone Loses—Except a Few Sharks—on Prediction Markets
A WSJ analysis shows a small number of accounts on Polymarket and Kalshi—often pros using data-driven algorithmic trading—take home most of the winnings.
55935363
Jean-François Bonnefon @jfbonnefon.bsky.social · 03/05/2026
Georgette Heyer describing Sir Waldo’s hand in The Nonesuch : « a shapely member, bare of rings » 👀
020
Reposted by Jean-François Bonnefon
Izabela Skoczeń @izaskoczen.bsky.social · 30/04/2026
‼️Conference Norms in Language-based Human-AI Interaction 📆 5th of June📍Bracka 12, Kraków ☑️preliminary program: docs.google.com/document/d/1... @talkingtobots.bsky.social @edouardmachery.bsky.social @jfbonnefon.bsky.social @monasimion.bsky.social
073
Reposted by Jean-François Bonnefon
Ethan Mollick @emollick.bsky.social · 28/04/2026
This is an actual line that was added to the official system prompt for Codex for GPT-5.5 by OpenAI. Usually the system prompt is as minimal as possible, so I assume it would otherwise mention goblins a lot. AIs are weird.
571323266
Reposted by Jean-François Bonnefon
depths of wikipedia @depthsofwikipedia.bsky.social · 22/04/2026
The Wikipedia article “Meaning of life” has been edited 10,713 times in its 25 year existence, and 23% of those edits have been reverted as vandalism. One common pattern is for vandals to delete the entire article and replace it with a single line. I collected some examples from the old days:
1568897
Jean-François Bonnefon @jfbonnefon.bsky.social · 14/04/2026
it just struck me that in French the word for an adult woman is "wife" and the word for a young girl is "daughter", whereas we have different words for men, husbands, boys, and sons. Now I'm curious to know if other languages show the same pattern
010
Reposted by Jean-François Bonnefon
Oded Rechavi @odedrechavi.bsky.social · 03/04/2026
Every PI
731048
Jean-François Bonnefon @jfbonnefon.bsky.social · 01/04/2026
I don't understand why reviewers can send confidential comments to the editor about a paper, but authors cannot send confidential comments to the editor about the reviews. As in "look we did our best but this review reads like an entry in the dream diary of Samuel Beckett with a bad case of flu"
3243
Jean-François Bonnefon @jfbonnefon.bsky.social · 27/03/2026
As a companion piece to our 'Science Fiction Science Method' article, we explore in Nautilus how we could turn a Ted Chiang story into a behavioral experiment @azimshariff.bsky.social @iyadrahwan.bsky.social
nautil.us
How Science Fiction Can Save Us
How Science Fiction Can Save Us: Three scientists share their bold vision for turning stories into testable experiments
041
Jean-François Bonnefon @jfbonnefon.bsky.social · 12/03/2026
I just confirmed my participation to a conference and meant to write "looking forward to!" but typed instead "looking forward yo!" I have turned into the academic version of Jesse Pinkman
media.tenor.com
a man says i 'm a criminal yo in a netflix ad
ALT: a man says i 'm a criminal yo in a netflix ad
140
Jean-François Bonnefon @jfbonnefon.bsky.social · 11/03/2026
I like how typos in manuscripts are no longer considered evidence of sloppiness, and have turned instead into luxury signals of hand-crafted writing by a human master wordsmith who keeps to the traditional ways
412526
Reposted by Jean-François Bonnefon
Daniel E. Weeks @statgendan.bsky.social · 05/03/2026
The canvas() function from the ggview R package is very useful for previewing/tweaking a ggplot into publication-ready format: it renders a plot "as it would appear if saved to a file with the specified dimensions". cran.r-project.org/web/packages... #RStats #ggplot #ggview
cran.r-project.org
README
1539
Jean-François Bonnefon @jfbonnefon.bsky.social · 05/03/2026
[user] analyze and summarize the attached document [French LLM] *Bonjour*
050
Reposted by Jean-François Bonnefon
Suresh Venkatasubramanian @geomblog.bsky.social · 05/03/2026
My zombie Prof would complain bitterly about \textit vs \emph and the use of raw text in math mode latex instead of using \text or proper operator commands. 💀💀💀
1131
Jean-François Bonnefon @jfbonnefon.bsky.social · 05/03/2026
What a better way to start the day than reading the sentence “Grammarly is being accused of necromancy”. Now *that’s* my kind of news
150
Reposted by Jean-François Bonnefon
Experimental Philosophy @xphilosopher.bsky.social · 26/02/2026
This piece by psychologist/philosopher Brian Earp is about the differing attitudes toward AI in psychology vs. philosophy In psych, one common attitude is; The goal is to get the right answer to fundamental questions in psychology. If AI helps us do that, well then, all the better… 1/
1105
Jean-François Bonnefon @jfbonnefon.bsky.social · 26/02/2026
One month to wait for Trace Elements, a conversation on my favorite genres by the author of my favorite series @adapalmer.bsky.social and the person whose taste in books I admire the most @bluejo.bsky.social. I’ve not been that excited about a book for a very very long while
jowaltonbooks.com
Trace Elements | Jo Walton -- Science Fiction and Fantasy Author
045
Jean-François Bonnefon @jfbonnefon.bsky.social · 25/02/2026
Also: what a missed opportunity to make us Jonathan and Jennifer
Jonathan and Jennifer Hart, from "Hart to Hart". If you are French and ever watched this show I'll bet one million euros that your mind is currently going "L'aaaaaamour du risque... "
020
Jean-François Bonnefon @jfbonnefon.bsky.social · 24/02/2026
My phone decided to autocomplete “Je” (French “I”) with Jérôme and my wife’s phone decided on Jennifer. from our texts you’d think we are obsessed with these two - Jerome will be late tonight - it’s ok, Jennifer will be at yoga - Jerome will stop for food on the way - Jennifer won’t need anything
180
Reposted by Jean-François Bonnefon
Florian Keppeler @flokeppeler.bsky.social · 22/02/2026
#Wilmasreview ❄️ Wilma isn't looking for a new home in the snow — but she is wondering: should AI decide who gets welfare benefits like housing support? A new study by Dong, @jfbonnefon.bsky.social & @iyadrahwan.bsky.social looked at exactly this. 1/🧵
122
Reposted by Jean-François Bonnefon
Sung Kim @sungkim.bsky.social · 23/02/2026
Pixel Agents A VS Code extension that turns your AI coding agents like Claude Code into animated pixel art characters in a virtual office. It shows what the agent is currently doing with animations. You can also customize the virtual office however you want. Repo: github.com/pablodelucca...
89213
Reposted by Jean-François Bonnefon
Ryan Moulton @moultano.bsky.social · 23/02/2026
I went a little overboard with this and a little bit insane. moultano.wordpress.com/2026/02/22/t...
moultano.wordpress.com
The Hunt for Dark Breakfast
With a theoretical model of breakfast, can we derive the existence of “dark breakfasts,” breakfasts that we know must exist, but have never observed?
2127075
Jean-François Bonnefon @jfbonnefon.bsky.social · 18/02/2026
I do it when I'm stressed, these emails have such a soothing rhythm
150
Reposted by Jean-François Bonnefon
Pantelis P. Analytis @pantelispa.bsky.social · 18/02/2026
Join us in June at @iast.fr Toulouse for a 2-day workshop on social influence and its large scale implications!
074
Reposted by Jean-François Bonnefon
Elise Wang @elisewang.bsky.social · 16/02/2026
I propose to make universal the old policy of the Blackfriars conference at the American Shakespeare Center: If you do not end your paper on time, you will be forced to exit, pursued by a bear. Literally, a bear will come take your paper from you.
The organizer, introducing the bear on stageA tardy speaker, trying to fend off the bear as he/she leans forward patiently.
331624565
Reposted by Jean-François Bonnefon
Jo Walton @bluejo.eurosky.social · 14/02/2026
My January Reading List reactormag.com/jo-waltons-r...
reactormag.com
Jo Walton’s Reading List: January 2026 - Reactor
Compelling memoirs, Vikings, Heyer, and early experiments in authoritarianism
2253
Jean-François Bonnefon @jfbonnefon.bsky.social · 07/02/2026
“The revolutionary AI social network was largely humans operating fleets of bots.”
wiz.io
Hacking Moltbook: AI Social Network Reveals 1.5M API Keys | Wiz Blog
Learn how a misconfigured Supabase database at Moltbook exposed 1.5M API keys, private messages, and user emails, enabling full AI agent takeover.
121
Reposted by Jean-François Bonnefon
Gretchen McCulloch @gretchenmcculloch.com · 01/02/2026
I saw some people calling it Rectangular February
1517644
Reposted by Jean-François Bonnefon
Rachel Merrill Artist for Gil Thorp and Death To Pachuco! @ohhhaeee.bsky.social · 31/01/2026
NGL this description of Bsky in Vulture cracked me up
894404672