iiinova — where your digital self keeps evolving

Reposted by Setayesh Radkani

Anna (Anya) Ivanova @neuranna.bsky.social · 20/10/2025

Finally out in TACL: 🌎EWoK (Elements of World Knowledge)🌎: A cognition-inspired framework for evaluating basic world knowledge in language models tl;dr: LLMs learn basic social concepts way easier than physical&spatial concepts Paper: direct.mit.edu/tacl/article... Website: ewok-core.github.io

16910

Reposted by Setayesh Radkani

Simon Columbus @simoncolumbus.bsky.social · 26/08/2025

⚡ Autumn 2025 Cooperation Colloquia ⚡ We are excited to announce the next run of Cooperation Colloquia. With @talbotmandrews.bsky.social, @setayeshradkani.bsky.social, @kris-smith.bsky.social, @alexmesoudi.com, & more. Every second Friday, 15:00 CE(S)T Sign up here: list.ku.dk/postorius/li...

Speaker list for the Autumn 2025 Cooperation Colloquia:
2025-09-12 | Doruk İriş (Sogang University)
2025-09-26 | Talbot M. Andrews (Cornell University)
2025-10-10 | Setayesh Radkani (MIT)
2025-10-24 | Christian Ruff (University of Zürich)
2025-11-07 | Bianca Beersma (Vrije Universiteit Amsterdam)
2025-11-21 | Kristopher M. Smith (Washington State University)
2025-12-05 | Alex Mesoudi (University of Exeter)

088

Setayesh Radkani @setayeshradkani.bsky.social · 08/08/2025

Also sharing a beautiful illustration of these ideas by my lovely and talented 👩‍🎨 friend, Adhara Martellini!

061

Setayesh Radkani @setayeshradkani.bsky.social · 08/08/2025

Bottom line: The same punishment can teach different lessons to different people depending on their prior beliefs, even when everyone is reasoning rationally. So, even well-meant punishment can widen divides or fuel polarization. 10/10

030

Setayesh Radkani @setayeshradkani.bsky.social · 08/08/2025

Finding 4️⃣: In a separate study, we found that repeated punishments can fail to close societal divides, and may even polarize initially shared beliefs. Our model predicts when punishment works in reducing polarization and when it backfires! 9/N

130

Setayesh Radkani @setayeshradkani.bsky.social · 08/08/2025

Finding 3️⃣: Our computational model simultaneously captures people’s belief updates about the act and the authority, even in novel prior conditions that the model has never seen before 👀; even better than control models that are fit to predict each belief separately! 8/N

120

Setayesh Radkani @setayeshradkani.bsky.social · 08/08/2025

Finding 2️⃣: People’s prior beliefs shape their reasoning about punishment. The same punishment can lead to contrasting inferences, depending on the value and uncertainty of prior beliefs about both the act and the authority. 7/N

120

Setayesh Radkani @setayeshradkani.bsky.social · 08/08/2025

Finding 1️⃣: From observing punishment (or no punishment!), people simultaneously update their beliefs about the wrongness of the target act, and about the authority’s motivations and values. 6/N

120

Setayesh Radkani @setayeshradkani.bsky.social · 08/08/2025

Across 3 studies, we used imaginary villages to experimentally control people’s pre-existing beliefs about the target act and the authority. We then measured how observing punishment with different severities moves their beliefs. 5/N

130

Setayesh Radkani @setayeshradkani.bsky.social · 08/08/2025

We used an inverse planning framework: people assume authorities plan punishment to achieve their desires based on their beliefs. By inverting this model, people infer the hidden beliefs and desires that most likely produced the observed punishment. 4/N

130

Setayesh Radkani @setayeshradkani.bsky.social · 08/08/2025

We solved this puzzle by building a computational model that characterizes “how” people interpret punishment. Key insight: People use their pre-existing beliefs and opinions to simultaneously evaluate both the norm to be learned and the authority who’s punishing. But how?! 3/N

130

Setayesh Radkani @setayeshradkani.bsky.social · 08/08/2025

This is a puzzle in life and in punishment literature! Even when nobody disagrees about the facts—everybody knows what action happened, who punished it, and what they did to punish it—different observers of the same punishment could come to drastically different conclusions. 2/N

130

Setayesh Radkani @setayeshradkani.bsky.social · 08/08/2025

🚨Out in PNAS🚨 with @joshtenenbaum.bsky.social & @rebeccasaxe.bsky.social Punishment, even when intended to teach norms and change minds for the good, may backfire. Our computational cognitive model explains why! Paper: tinyurl.com/yc7fs4x7 News: tinyurl.com/3h3446wu 🧵

tinyurl.com

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...

36728

Reposted by Setayesh Radkani

Larisa Heiphetz Solomon @drlarisa.bsky.social · 06/04/2025

1. This Friday I hosted a workshop on morality with some fabulous humans who took take time out of their lives to talk about ideas with each other. Here is a thread with brief summaries of what folks presented in case you too would like to hear about their science: #PsychSciSky #SocialPsyc #DevPsyc

3305

Reposted by Setayesh Radkani

MIT Kids Brains @kidsbrains.bsky.social · 03/03/2025

The Saxe Lab @ MIT is hiring! We seek one lab manager to start in summer 2025. Research in our lab focuses on social cognition (learn more on saxelab.mit.edu). Please apply at: tinyurl.com/saxe2025 (Job ID 31993). Review of applications starts on March 24, 2025. Sharing appreciated. Thank you!

saxelab.mit.edu

Saxelab Social Cognitive Neuroscience Laboratory at MIT |

25444

Reposted by Setayesh Radkani

Benoît de Courson @bdecourson.bsky.social · 07/02/2025

Does poverty lead to risk taking or risk avoidance? Turns out, to both. Our new paper (with D. Nettle & W. Frankenhuis) in @royalsocietypublishing.org explains why, and conducts preregistered tests of our ‘desperation threshold’ model. royalsocietypublishing.org/doi/10.1098/... A 🧵

12615

Setayesh Radkani @setayeshradkani.bsky.social · 11/11/2024

Thanks for this! Could I please be added as well?

100

Setayesh Radkani @setayeshradkani.bsky.social · 15/10/2024

Huge thanks to @dgrand.bsky.social , @falklab.bsky.social and @anthlittle.bsky.social for their feedback on this work, and to our funding sources! 13/13

000

Setayesh Radkani @setayeshradkani.bsky.social · 15/10/2024

We need these authorities to succeed in being seen as independent and truthful, because in this space of uncertainty, those are the voices that can move people toward an accurate outcome 12/N

000

Setayesh Radkani @setayeshradkani.bsky.social · 15/10/2024

Given the current 👋 state of things, this work offers insight into the challenges faced by independent election observers, public health professionals, and others seeking to cultivate credibility as debunkers 11/N

000

Setayesh Radkani @setayeshradkani.bsky.social · 15/10/2024

Finding 3: Differing beliefs about authorities can spread polarization When beliefs about the authority diverge in the original domain (election fraud), their debunking can polarize the two groups in a new domain (e.g., public health) even when the groups initially share the same perspective 10/N

000

Setayesh Radkani @setayeshradkani.bsky.social · 15/10/2024

Finding 2: Credibility isn’t just about unbiasedness Consistent with other work in political science, authorities seen as biased can still be effective debunkers – if they are seen as biased *in favor* of the perspective they are debunking, although this depends on the uncertainty of beliefs 9/N

010

Setayesh Radkani @setayeshradkani.bsky.social · 15/10/2024

if people are certain, debunking – even by a reasonably unbiased, committed authority – fails: 1) The groups' beliefs about the perspective remain polarized 2) The groups' initially shared beliefs about the authority additionally diverge! 8/N

010

Setayesh Radkani @setayeshradkani.bsky.social · 15/10/2024

Finding 1: Debunking can work…sometimes If initial beliefs in the perspective (“the election was unfair”) are held with some uncertainty, then groups can eventually converge on shared beliefs, if they also believe the authority is unbiased and committed to the truth. But … 7/N

020

Setayesh Radkani @setayeshradkani.bsky.social · 15/10/2024

Using simulations, we capture how three related beliefs evolve: (1) commitment to the initial perspective; (2) views on the bias of the authority; (3) views on the authority’s commitment to the truth 6/N

010

Setayesh Radkani @setayeshradkani.bsky.social · 15/10/2024

We use this inverse-planning framework [pubmed.ncbi.nlm.nih.gov/19729154] to develop a computational model of how observers’ beliefs evolve as each group witnesses 5 debunking acts that counter a single perspective on a topic (e.g., “the election was unfair”) 5/N

010

Setayesh Radkani @setayeshradkani.bsky.social · 15/10/2024

We make minimal assumptions that both groups: - Are made up of equally rational Bayesians - Differ in their belief on a topic (ground truth may not be known) - Hold shared beliefs about the authority’s motives with some uncertainty (see the paper for a discussion of why these assumptions) 4/N

010

Setayesh Radkani @setayeshradkani.bsky.social · 15/10/2024

We focus instead on how observers interpret debunking acts (and update their beliefs) based on their intuitive theory of the authority’s motives: degree of bias, commitment to the truth 3/N

010

Setayesh Radkani @setayeshradkani.bsky.social · 15/10/2024

Lots of great work on misinformation in politics, health, etc. focuses on the content and delivery method of corrections, and on the role of biases, identities, etc. in sustaining polarized beliefs 2/N

010

Setayesh Radkani @setayeshradkani.bsky.social · 07/02/2024

How do people interpret punitive events (differently) and what do they learn from it? Excited to present our work (with @rebeccasaxe.bsky.social) at #SPSP2024 and moral psychology preconference (in collaboration with Josh Tenenbaum) @spspnews.bsky.social

2115