Setayesh Radkani

A window into my research (and life)…

It's been more than a year, but the EWoK (Elements of World Knowledge) paper is finally out in TACL! tl;dr: language models learn basic social concepts way easier than physical and spatial concepts. direct.mit.edu/tacl/article/d…

My thoughts are with British Jews who have had to watch anti-Semites march through our streets for years, our government legitimise and reward anti-Semitic Islamist terror by recognising Palestine, and now this hideous atrocity. Our country has consistently let Jews down.

Are you an aspiring scientist coming from a non-traditional academic background or facing unusual adversity during graduate school application? Sign up for the student-run BCS Application Assistance Program by November 20 for free advice: lnkd.in/eJ2hjYr8

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

🚨Out in PNAS🚨 with @MITCoCoSci & @rebecca_saxe 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 🧵

In a paper publishing in @PNASNews this week, McGovern researchers @SetayeshRadkani + @rebecca_saxe show the same punishment can either build respect for authority or deepen distrust—depending on what people already believe. mcgovern.mit.edu/2025/08/04/lea…

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

Finding 4⃣: 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

Is the Language of Thought == Language? A Thread 🧵 New Preprint (link: tinyurl.com/LangLOT) with @alexanderdfung, @ParisJaggers, Jason Chen, Josh Rule, @BennYael, @MITCoCoSci, @spiantado, Rosemary Varley, @ev_fedorenko 1/8

🎬 Runway Aleph: hype or helpful? • In-context video editing/generation with simple prompts (add/remove/transform). • Big step beyond Gen-3 Alpha: from text-to-video → in-context edits. • Demos show day→night, rain, reflection removal, green screen, scene extension & new angles. • Waitlist only so far—expect variability, artifacts on non-curated footage, and high compute costs (which means expensive to use). • Likely better identity/motion preservation vs older models, but real-world consistency is TBD. Verdict: Sora-level wow in demos; promising but early. Worth watching, especially for post-production. Are you joining the waitlist? 👇 — Louis-François — PhD-dropout & CTO/co-founder @ Towards AI. Follow for tomorrow’s no-BS roundup. #Runway #Aleph #GenAI #VideoAI #VFX #VideoEditing #TechNews #HypeOrNot #AI

How do people reason while still staying coherent – as if they have an internal ‘world model’ for situations they’ve never encountered? A new paper on open-world cognition (preview at the world models workshop at #ICML2025!)

@BlackHC Very cool example! We recently made a similar argument on how rational belief updating could lead to sustained (or worsened) polarization by considering structured interactions among beliefs x.com/setayeshradkan…

Ever wondered why presenting more facts can sometimes *worsen* disagreements, even among rational people? 🤔 It turns out, Bayesian reasoning has some surprising answers - no cognitive biases needed! Let's explore this fascinating paradox quickly ☺️

🚨New paper🚨How do we decide who has the 'moral standing' to call out wrongdoing? New research, led by former YES Lab member @ghezae_isaias and in collaboration with @FanYangUChicago explores folk perception of moral standing to blame. Now in Open Mind direct.mit.edu/opmi/article/d…