Max Puelma Touzel
@mptouzel.bsky.social
Staff Research Scientist@Mila/complexdatalab.com
Getting at the psycho-social in our digital spaces
with models and data with the aim to make better ones
mptouzel.github.io
correlated diffusion over AI/ML/(MA)RL/psych/soc/media/pol/econ/energy
Getting at the psycho-social in our digital spaces
with models and data with the aim to make better ones
mptouzel.github.io
correlated diffusion over AI/ML/(MA)RL/psych/soc/media/pol/econ/energy
Pinned
Agent Societies are generating significant attention—but we lack consensus for what they mean for us.
This confusion isn't surprising—even researchers designing these systems struggle to evaluate what's real, as evidenced in recent scientific papers.
This confusion isn't surprising—even researchers designing these systems struggle to evaluate what's real, as evidenced in recent scientific papers.
Reposted by Max Puelma Touzel
*Please repost* @sjgreenwood.bsky.social and I just launched a new personalized feed (*please pin*) that we hope will become a "must use" for #academicsky. The feed shows posts about papers filtered by *your* follower network. It's become my default Bluesky experience bsky.app/profile/pape...
March 10, 2025 at 6:14 PM
*Please repost* @sjgreenwood.bsky.social and I just launched a new personalized feed (*please pin*) that we hope will become a "must use" for #academicsky. The feed shows posts about papers filtered by *your* follower network. It's become my default Bluesky experience bsky.app/profile/pape...
Reposted by Max Puelma Touzel
Never have I felt more like my job will soon by taken by AI. Statistical learning theory in Lean: concentration inequalities, Dudley's entropy integral, and local Gaussian complexity bounds.
30000 lines of code, over 1000 lemmas, formalizing Wainwright and Boucheron et al arxiv.org/abs/2602.02285
30000 lines of code, over 1000 lemmas, formalizing Wainwright and Boucheron et al arxiv.org/abs/2602.02285
February 3, 2026 at 7:20 PM
Never have I felt more like my job will soon by taken by AI. Statistical learning theory in Lean: concentration inequalities, Dudley's entropy integral, and local Gaussian complexity bounds.
30000 lines of code, over 1000 lemmas, formalizing Wainwright and Boucheron et al arxiv.org/abs/2602.02285
30000 lines of code, over 1000 lemmas, formalizing Wainwright and Boucheron et al arxiv.org/abs/2602.02285
Agent Societies are generating significant attention—but we lack consensus for what they mean for us.
This confusion isn't surprising—even researchers designing these systems struggle to evaluate what's real, as evidenced in recent scientific papers.
This confusion isn't surprising—even researchers designing these systems struggle to evaluate what's real, as evidenced in recent scientific papers.
February 4, 2026 at 8:13 PM
Agent Societies are generating significant attention—but we lack consensus for what they mean for us.
This confusion isn't surprising—even researchers designing these systems struggle to evaluate what's real, as evidenced in recent scientific papers.
This confusion isn't surprising—even researchers designing these systems struggle to evaluate what's real, as evidenced in recent scientific papers.
Reposted by Max Puelma Touzel
Why don’t neural networks learn all at once, but instead progress from simple to complex solutions? And what does “simple” even mean across different neural network architectures?
Sharing our new paper @iclr_conf led by Yedi Zhang with Peter Latham
arxiv.org/abs/2512.20607
Sharing our new paper @iclr_conf led by Yedi Zhang with Peter Latham
arxiv.org/abs/2512.20607
February 3, 2026 at 4:19 PM
Why don’t neural networks learn all at once, but instead progress from simple to complex solutions? And what does “simple” even mean across different neural network architectures?
Sharing our new paper @iclr_conf led by Yedi Zhang with Peter Latham
arxiv.org/abs/2512.20607
Sharing our new paper @iclr_conf led by Yedi Zhang with Peter Latham
arxiv.org/abs/2512.20607
Reposted by Max Puelma Touzel
Good piece, especially because it foregrounds "legitimacy" as the remaining value proposition for universities. In a world where knowledge and skill are not scarce, metacognitive decisions will still be hard — and those decisions are interwoven with questions of social validation. +
AI will transform what the university sells and how it sells it. That is why it is the alligator closest to the boat. I welcome the AI alligator, because mentorship paired with expert judgment is the only defensible value proposition left. www.insidehighered.com/opinion/view...
AI-Forced Restructuring Is Top Risk for Higher Ed (opinion)
AI-forced restructuring is the biggest risk colleges face.
www.insidehighered.com
February 3, 2026 at 4:26 PM
Good piece, especially because it foregrounds "legitimacy" as the remaining value proposition for universities. In a world where knowledge and skill are not scarce, metacognitive decisions will still be hard — and those decisions are interwoven with questions of social validation. +
Thoughts:
-AI agents behave in roles according to externally imposed context
-eg. they act as if driven by the goals of the roled persona
-if given agency, they will act consistent with that persona.
-if the role is inferred from hyperbolic "clanker" discourse, we are setting us up for antagonism.
-AI agents behave in roles according to externally imposed context
-eg. they act as if driven by the goals of the roled persona
-if given agency, they will act consistent with that persona.
-if the role is inferred from hyperbolic "clanker" discourse, we are setting us up for antagonism.
January 31, 2026 at 11:00 PM
Thoughts:
-AI agents behave in roles according to externally imposed context
-eg. they act as if driven by the goals of the roled persona
-if given agency, they will act consistent with that persona.
-if the role is inferred from hyperbolic "clanker" discourse, we are setting us up for antagonism.
-AI agents behave in roles according to externally imposed context
-eg. they act as if driven by the goals of the roled persona
-if given agency, they will act consistent with that persona.
-if the role is inferred from hyperbolic "clanker" discourse, we are setting us up for antagonism.
Reposted by Max Puelma Touzel
Come join us for this first worshop of a three part series on the computational ingredients of reasoning in minds and AI. Reasoning is a complex term, especially in light of an exploding category of methods in LLMs. These workshops will explore reasoning’s multiple facets.
IVADO unveils the schedule of the first workshop "Cognitive Basis of #Reasoning (in Minds and #AI)", Jan 27-29, 2026, spearheaded by @taylorwwebb.bsky.social and Dhanya Sridhar.
🗓️ Schedule and speakers: ivado.ca/en/events/co...
📥 Registration: event.fourwaves.com/thematicseme...
🗓️ Schedule and speakers: ivado.ca/en/events/co...
📥 Registration: event.fourwaves.com/thematicseme...
Cognitive Basis of Reasoning (in Minds and AI) | IVADO
ivado.ca
January 20, 2026 at 10:34 PM
Come join us for this first worshop of a three part series on the computational ingredients of reasoning in minds and AI. Reasoning is a complex term, especially in light of an exploding category of methods in LLMs. These workshops will explore reasoning’s multiple facets.
Reposted by Max Puelma Touzel
I am recruiting graduate students for the experimental side of my lab @mcgill.ca for admission in Fall 2026!
Get in touch if you're interested in how brain circuits implement distributed computation, including dopamine-based distributed RL and probabilistic representations.
Get in touch if you're interested in how brain circuits implement distributed computation, including dopamine-based distributed RL and probabilistic representations.
November 19, 2025 at 6:04 PM
I am recruiting graduate students for the experimental side of my lab @mcgill.ca for admission in Fall 2026!
Get in touch if you're interested in how brain circuits implement distributed computation, including dopamine-based distributed RL and probabilistic representations.
Get in touch if you're interested in how brain circuits implement distributed computation, including dopamine-based distributed RL and probabilistic representations.
Some views on why cognitive effort exertion can feel costly.
New pontification piece with @awestbrook.bsky.social and Jean Daunizeau, just out in TICS:
Why is cognitive effort experienced as costly?
(or why does it hurt to think)
never written a review paper before in my life, that was a new and unusual experience
Why is cognitive effort experienced as costly?
(or why does it hurt to think)
never written a review paper before in my life, that was a new and unusual experience
Why is cognitive effort experienced as costly?
A widespread observation is that people avoid mentally effortful courses of action,
and much recent work examining cognitive effort has explained subjective effort evaluation
– and, consequently, pref...
www.cell.com
November 19, 2025 at 5:14 PM
Some views on why cognitive effort exertion can feel costly.
Reassuring that we have set reasonable submission limits. Something to aspire to! lol
November 19, 2025 at 2:41 PM
Reassuring that we have set reasonable submission limits. Something to aspire to! lol
Move fast, break things.
apnews.com/article/open...
Even if their recent push has largely fixed this issue
openai.com/index/streng...
Can't help but think why they didn't push earlier...
apnews.com/article/open...
Even if their recent push has largely fixed this issue
openai.com/index/streng...
Can't help but think why they didn't push earlier...
OpenAI faces 7 lawsuits claiming ChatGPT drove people to suicide, delusions
OpenAI is facing seven lawsuits claiming ChatGPT drove people to suicide and harmful delusions even when they had no prior mental health issues.
apnews.com
November 7, 2025 at 2:09 AM
Move fast, break things.
apnews.com/article/open...
Even if their recent push has largely fixed this issue
openai.com/index/streng...
Can't help but think why they didn't push earlier...
apnews.com/article/open...
Even if their recent push has largely fixed this issue
openai.com/index/streng...
Can't help but think why they didn't push earlier...
Reposted by Max Puelma Touzel
Canadian researchers should be aware the there is a motion before the Parliamentary Standing Committee on Science and Research to force Tricouncils to hand over disaggregated peer review data on all applications:
Applicant names, profiles, demographics
Reviewers names, profiles, comments, and scores
Applicant names, profiles, demographics
Reviewers names, profiles, comments, and scores
October 30, 2025 at 8:33 PM
Canadian researchers should be aware the there is a motion before the Parliamentary Standing Committee on Science and Research to force Tricouncils to hand over disaggregated peer review data on all applications:
Applicant names, profiles, demographics
Reviewers names, profiles, comments, and scores
Applicant names, profiles, demographics
Reviewers names, profiles, comments, and scores
Reposted by Max Puelma Touzel
If you don't think rando MPs (and whoever else they share it with) should have access to personal information of individual applicants, reviewers, and students, confidential data, & full applications and reviews in every Canadian research grant (funded or not), please sign.
Yes! Contact them and sign this petition here —it’s not just about EDI data! So much personally identifying information for applicants and reviewers will not be protected!
docs.google.com/forms/d/e/1F...
docs.google.com/forms/d/e/1F...
Open Letter to Protect Tri-Council EDI Data
Why do we need urgent action to protect our data?
The mandate of the House of Commons Standing Committee on Science and Research is to review and report on topics relating to science and research in ...
docs.google.com
October 31, 2025 at 12:46 AM
If you don't think rando MPs (and whoever else they share it with) should have access to personal information of individual applicants, reviewers, and students, confidential data, & full applications and reviews in every Canadian research grant (funded or not), please sign.
Reposted by Max Puelma Touzel
Tell that to Jamaica
"Bill Gates Says Climate Change Isn’t So Bad After All --Experts are calling his latest memo "pointless, vague, unhelpful and confusing" by Victor Tangermann for @futurism.com: futurism.com/science-ener...
Bill Gates Says Climate Change Isn't So Bad After All
In a new memo published on his blog, Bill Gates pushed back on a "doomsday view of climate change," arguing we should help the poor instead.
futurism.com
October 28, 2025 at 8:56 PM
Tell that to Jamaica
Reposted by Max Puelma Touzel
How does behavior spread? New insights on complex contagion in social networks with causal evidence from a country-scale field experiment @davidlazer.bsky.social sociologicalscience.com/articles-v12...
October 16, 2025 at 5:51 PM
How does behavior spread? New insights on complex contagion in social networks with causal evidence from a country-scale field experiment @davidlazer.bsky.social sociologicalscience.com/articles-v12...
Had an amazing time at #COLM2025 It was vibrant, high level, and seemed a healthy balance of LLM critique and solution focussed. I am so happy with how our social simulation workshop went. Chairing and panel moderating was a pleasure thanks to the many that participated. Stay tuned for recordings!
October 11, 2025 at 7:29 PM
Had an amazing time at #COLM2025 It was vibrant, high level, and seemed a healthy balance of LLM critique and solution focussed. I am so happy with how our social simulation workshop went. Chairing and panel moderating was a pleasure thanks to the many that participated. Stay tuned for recordings!
Excited to be at #COLM2025 and to be organizing one of (perhaps the?) first workshop on social modelling with LLMs! Looking forward to this community-wide stocktake on the state of the field, main challenges, and promising applications!
sites.google.com/view/social-...
sites.google.com/view/social-...
Social Simulation with LLMs
Important Information
Workshop Date: October 10, 2025
Workshop Room: 523AB Workshop Time Zone: Montreal (UTC -4)
Contact Email: social-simulation@googlegroups.com
sites.google.com
October 7, 2025 at 2:06 PM
Excited to be at #COLM2025 and to be organizing one of (perhaps the?) first workshop on social modelling with LLMs! Looking forward to this community-wide stocktake on the state of the field, main challenges, and promising applications!
sites.google.com/view/social-...
sites.google.com/view/social-...
Looks like I missed the publication of another stochastic thermodynamics text in June. Seifert is arguably better positioned, but let's see about his exposition style. The other text is Politi & Pigolotti, which was nice, but light on the info side.
www.cambridge.org/core/books/s...
www.cambridge.org/core/books/s...
Stochastic Thermodynamics
Cambridge Core - Statistical Physics - Stochastic Thermodynamics
www.cambridge.org
September 30, 2025 at 7:49 PM
Looks like I missed the publication of another stochastic thermodynamics text in June. Seifert is arguably better positioned, but let's see about his exposition style. The other text is Politi & Pigolotti, which was nice, but light on the info side.
www.cambridge.org/core/books/s...
www.cambridge.org/core/books/s...
Reposted by Max Puelma Touzel
🚨Our preprint is online!🚨
www.biorxiv.org/content/10.1...
How do #dopamine neurons perform the key calculations in reinforcement #learning?
Read on to find out more! 🧵
www.biorxiv.org/content/10.1...
How do #dopamine neurons perform the key calculations in reinforcement #learning?
Read on to find out more! 🧵
September 19, 2025 at 1:05 PM
🚨Our preprint is online!🚨
www.biorxiv.org/content/10.1...
How do #dopamine neurons perform the key calculations in reinforcement #learning?
Read on to find out more! 🧵
www.biorxiv.org/content/10.1...
How do #dopamine neurons perform the key calculations in reinforcement #learning?
Read on to find out more! 🧵
Reposted by Max Puelma Touzel
I am recruiting (Canadian🇨🇦) graduate students at York University for Fall 2026 in Social and Personality Psychology! If you are interested in misinformation, political polarization, and/or computational social science, I'd love to hear from you. Feel free to reach out with any questions.
September 15, 2025 at 7:29 PM
I am recruiting (Canadian🇨🇦) graduate students at York University for Fall 2026 in Social and Personality Psychology! If you are interested in misinformation, political polarization, and/or computational social science, I'd love to hear from you. Feel free to reach out with any questions.
Reposted by Max Puelma Touzel
It seems obvious that we should prioritize addressing misinformation that is more harmful. But what makes something more likely to be harmful? Can we can reason about it before the harm has occurred?
📣 In an upcoming #CSCW2024 paper, we present a taxonomy of *Misinformation as a Harm* ➡️
📣 In an upcoming #CSCW2024 paper, we present a taxonomy of *Misinformation as a Harm* ➡️
April 15, 2024 at 5:42 PM
It seems obvious that we should prioritize addressing misinformation that is more harmful. But what makes something more likely to be harmful? Can we can reason about it before the harm has occurred?
📣 In an upcoming #CSCW2024 paper, we present a taxonomy of *Misinformation as a Harm* ➡️
📣 In an upcoming #CSCW2024 paper, we present a taxonomy of *Misinformation as a Harm* ➡️
If there is anything worse than "economist brain", it's economist brain arguments about how inefficient the state is.
September 6, 2025 at 8:13 PM
If there is anything worse than "economist brain", it's economist brain arguments about how inefficient the state is.
Reposted by Max Puelma Touzel
1800 block of South Western Avenue in Chicago
August 16, 2025 at 8:28 PM
1800 block of South Western Avenue in Chicago
Ironic how that last sentance in the image (as a belief) is a deeply held fiction for certain individuals ('deep state is after me') who voted for a govmnt that's doing just that to other people whose idealistic notion of govmnt presumably made them not believe it could happen, least of all to them.
One trademark of an authoritarian regime is the use of the legal system to target political opponents. Bill Pulte, a Trump appointee who leads the Federal Housing Finance Agency, has been using his position to concoct accusations of mortgage fraud against Letitia James, Adam Schiff and now Cook.
August 20, 2025 at 5:22 PM
Ironic how that last sentance in the image (as a belief) is a deeply held fiction for certain individuals ('deep state is after me') who voted for a govmnt that's doing just that to other people whose idealistic notion of govmnt presumably made them not believe it could happen, least of all to them.