iiinova — where your digital self keeps evolving

Tobias Gerstenberg @tobigerstenberg.bsky.social · 12/10/2026

We share some tips on how to develop better intuition for how observational and experimental data should be combined to make good decisions. Counterfactual thinking is at the heart of it! You can try out your own intuitions by playing around with data: cicl-stanford.github.io/good_decisio...

110 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 12/10/2026

While both novices and experts make incorrect inferences about the counterfactual types from the data (e.g. believing people benefit from acting when they don't), their decisions systematically follow these (incorrect) inferences.

110 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 12/10/2026

People's decisions are well-explained by a simple decision model that uses a linear additive combination of the experimental and observational data. But these decisions don't correlate well with the predictions of the normative model.

110 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 12/10/2026

People leave a lot of money on the table! Getting feedback doesn't improve performance, but engaging with the data more closely helps. Experts at Stanford (postdocs & students in statistics, economics, or data science) performed better than novices but still far from optimally.

110 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 12/10/2026

Balke & Pearl (1994) developed a normative model of how to do this. How do people's decisions compare? Not well! People are better than chance but far from optimal. They have difficulty inferring the counterfactual types from the data, relying on a simple linear additive heuristic instead.

110 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 12/10/2026

One should use the data to infer whether someone will benefit from the treatment, be harmed by it, or be unaffected. These are counterfactual notions! Benefit = better because of action. By placing a value on each counterfactual type, one can compute the expected benefit of acting vs. not acting.

110 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 12/10/2026

I'm very excited to share a new preprint with you all (5 years in the making). "Making good decisions with counterfactuals" with Ricky Ma, Scott Mueller, and Thomas Icard. We investigate how people combine experimental and observational data to make decisions. 📃 osf.io/preprints/ps...

16115 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 23/09/2026

I'm so grateful for the continued support from my lab, and for everyone who helped along the way, now as an associate professor with tenure at Stanford 🧑‍🏫🥂🌲

222204 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 24/07/2026

Wonderful keynote by @hyogweon.bsky.social at #cogsci2026 on "Distinctively human intelligence: Thinking, learning, and communicating about the self". Hyo argues that better understanding how people think about the self is a key ingredient for uncovering what makes us intelligent.

0352 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 20/07/2026

The Causality in Cognition Lab is pumped for #CogSci2026 🇧🇷

05814 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 16/07/2026

Now out in Noûs: A communication-first account of explanation, with Jacqueline Harding (jacquelineharding.github.io) and Thomas Icard (stanford.edu/~icard/). Paper: onlinelibrary.wiley.com/doi/10.1111/... Preprint: arxiv.org/abs/2505.03732

05116 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 01/07/2026

Very proud Doktorvater moment from a few weeks back. Pictured with Dr Sarah Wu and Dr David Rose (and with a large hat thanks to University College London's regalia). Sarah will postdoc with @tomerullman.bsky.social. David with @shaunnichols.bsky.social and @davidpierce.xyz. I will miss you!! ♥️

2330 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 17/06/2026

New paper in Cognitive Psychology on how people combine visual 👀, auditory 👂, and haptic information 🖐️ to make causal judgments. Led by Elyse Chase (www.elysechase.com), with Kevin Smith, and Sean Follmer. 📜 authors.elsevier.com/a/1nHnI2Hxok...

0264 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 02/06/2026

It was such an honor and so much fun to participate in Phil Johnson-Laird's 90th birthday conference. Thank you @ruthbyrne.bsky.social and Sunny Khemlani for organizing the event, and the folks at NYU for hosting it. Here is Phil asking whether people are smarter than logic. Phil certainly is!

0100 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 01/06/2026

Congratulations to Dr. Dr. David Rose! 🥳 David (davdrose.github.io) first got a PhD in Philosophy, and now added another one in Psychology.

0280 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 15/05/2026

Congratulations Dr. Sarah Wu 🎓🎉 (sarahawu.github.io)

0310 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 12/04/2026

From Susan Gelman's talk (who met John thrice 📖☎️👨‍🏫): 1) Don't be afraid to ask questions. 2) "Research is hard (and fun!)" 3) Appearances can be deceiving. From Alison Gopnik's comments: John showed us that it's possible to be a good man and a good scientist.

060 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 12/04/2026

Some notes from the wonderful celebration of John Flavell's life and work at #CDS2026. "Age early and get it over with." (John on his looks not having changed in many years.) "Use the simplest word to get the idea across." "If I could only step into a child's mind even if just for a moment."

1150 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 09/04/2026

Excited (and cold) for my first ever #CDS2026! I'll be presenting work by the 'Dropping Things' dream team in the workshop on "Counterfactual thinking: What is it good for?" Paper sneak peek: cicl.stanford.edu/papers/rose2...

0321 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 03/04/2026

The Causality in Cognition Lab -- a supportive, bluesky-colored team -- is looking for a predoc to join us! Here are infos about the lab (cicl.stanford.edu) and the position (careersearch.stanford.edu/jobs/iriss-p...). The application deadline is May 1st. Please share, thank you 🙏

26536 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 10/03/2026

Congratulations @judithfan.bsky.social on winning the Lila R. Gleitman Prize for early-career contributions to Cognitive Science 🥳 Amazing!! cognitivesciencesociety.org/gleitman-pri...

47310 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 06/03/2026

Thanks Arthur Le Pargneux (arthurlepargneux.wixsite.com/arthurleparg...) for your talk "Contractualist moral cognition: From fair divisions to the emergence of rules via implicit agreements". A novel take on morality backed by clever experiments + elegant models 👍 📃 psycnet.apa.org/fulltext/202...

0122 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 27/02/2026

Thank you Yoonseo Zoh (zohyos7.github.io) for sharing your work with us on "Intuitive Theories in Moral Cognition". Intuitive theories structure how people represent dilemmas, how they generalize to new contexts, and how they switch between representations based on resource-rational constraints.

1141 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 20/02/2026

Thanks @maxtaylordavies.bsky.social for sharing your work with us on "Using the information bottleneck to study social cognition". Max develops resource-rational models that elegantly unify existing theories for various phenomena such as stereotyping and ToM development. 📃 osf.io/preprints/ps...

0151 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 15/02/2026

I'm very sad to have learned today that Joe Halpern passed away. Joe was a giant who knew no scientific boundaries. He loved science with a contagious, child-like enthusiasm. He was wonderful and I'm so grateful that I got to learn from him. Thank you Joe 🙏 www.bangsfuneralhome.com/obituaries/j...

15112 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 13/02/2026

Thanks @katenuss.bsky.social for sharing your work with us! Kate studies how people learn about the world through external exploration (acting) and internal exploration (imagining). Modeling & experiments reveal that people learn by simulating counterfactuals! 📃 elifesciences.org/articles/84260

1142 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 07/02/2026

Thank you @ltreiman.bsky.social for sharing your work with us on how people act differently when they know that their behavior is used to train AI. In the ultimatum game, people are more likely to reject disadvantageous offers when an AI is watching and learning. 📃 www.pnas.org/doi/10.1073/...

0113 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 22/01/2026

Thanks @miriam-hauptman.bsky.social for sharing your work with us! How people learn about the visual world from language is mediated through causal models. Both sighted and blind people infer how many colors an object has based on how color ➡️ function. 📃 m-hauptman.github.io/files/Hauptm...

091 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 15/01/2026

Thank you Aniket Vashishtha (aniketvashishtha.github.io) for sharing your work on counterfactual reasoning in LLMs with us! Most current benchmarks don't assess genuine counterfactual reasoning. Aniket's work does, showing that LLMs struggle and what to do about that. 📃 arxiv.org/abs/2510.015...

092 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 13/11/2025

I'm very excited about this work led by @nbalamur.bsky.social Inspired by the classic "Spot the ball ⚽" task, we develop a benchmark for visual social inference. We find that human participants perform much better than vision-language models. Try it here: v0-new-project-9b5vt6k9ugb.vercel.app

0133 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 05/11/2025

It was great to hear from Brian Leahy (brianleahy.net) in the devo lunch at Stanford today!! He presented a beautiful set of studies that suggest that many 4-year-old children have a minimal concept of possibility: they simulate only once and treat the outcome as a fact. 🎱⬅️➡️🤔💭💡

2333 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 24/10/2025

We develop a causal abstraction that infers a causal story of how the data was generated, paying more attention to factors that mattered for the prediction task. This model captures participants' generalization judgments better than a feature-based model, despite having many fewer parameters.

100 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 24/10/2025

This time, we also asked participants to predict what would happen in novel situations. For example, we showed them two familiar cubes on a novel ramp. These generalization trials also featured ramps that were facing the opposite direction from what they had seen before.

100 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 24/10/2025

In Exp 3, participants either viewed forward-facing ramps, or backward-facing ones. The cubes always ended up on the right side. Again, after having learned to predict which cubes cross the finish line, we surprised and asked them where exactly the cubes would be. Ps made similar errors as in Exp 2.

100 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 24/10/2025

Exp 2: Participants predict whether a cube on ramp will cross a finish line. Either cube color or ramp color is diagnostic. A surprise question about exactly where the cube will end up reveals systematic errors: they knew on which side of the line the cube would end up, but not the exact location.

100 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 24/10/2025

Exp 1: Participants learn whether color or shape matter for turning on a machine. In a surprise test, we ask them what they saw last. They frequently misremember (e.g., choosing a differently colored object when only the shape mattered). Only happens with enough evidence to learn the rule!

100 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 24/10/2025

When people are asked to predict what happens next, do they learn simple feature-outcome mappings, or do they learn causal models that capture the underlying generative process? If they do learn causal models, how can we tell? We ran 3 experiments (N=1080) using two paradigms to find out.

100 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 24/10/2025

🚨 New preprint 🚨 How do people's mental models shape memory, prediction, and generalization? We find that people spontaneously construct goal-dependent causal abstractions that compress experience to privilege relevant information. 📃 osf.io/preprints/ps... 🔗 github.com/cicl-stanfor...

25410 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 17/10/2025

The Causality in Cognition Lab at Stanford University is recruiting PhD students this cycle! We are a supportive team who happened to wear bluesky appropriate colors for the lab photo (this wasn't planned). 💙 Lab info: cicl.stanford.edu Application details: psychology.stanford.edu/admissions/p...

0427 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 13/10/2025

This project was expertly led by David Rose (davdrose.github.io) in collaboration with @siyingzhg.bsky.social, Sophie Bridgers, Hyo Gweon, and myself. 📄 doi.org/10.31234/osf... 🔗https://github.com/cicl-stanford/counterfactual_development

1100 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 13/10/2025

So what did we find? We tested 480 children and 91 adults online. Participants saw 4 (Exp 1) or 6 different scenarios. We find that children perform above chance when they're around 5 years of age. And we find a marked shift in performance around 7 years of age (where most children seem to get it).

130 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 13/10/2025

Three experiments rule out simpler explanations: 1️⃣ Different objects; children might answer based on preference. 2️⃣ Same objects; children might anticipate what would happen (hypothetical thinking). 3️⃣ Same objects, outcome revealed later; children need genuine counterfactual thinking.

120 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 13/10/2025

The "dropping things" task removes language and tests genuine counterfactual thinking. Granny drops two objects: an 🥚 and a 🏀. Two friends catch them. Granny would like to thank them but only has one sticker. Who should she give it to? Not catching the 🥚 would have been worse, so "Suzy"!

120 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 13/10/2025

Estimates of when counterfactual thinking develops range from 2-12 years. Two potential reasons: language & reasoning. 💬 A question like: "Where would Peter have been if there hadn’t been a fire?” is difficult to understand! 🤔 Counterfactual and hypothetical thinking are different!

130 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 13/10/2025

🚨New Preprint: We develop a novel task that probes counterfactual thinking without using counterfactual language, and that teases apart genuine counterfactual thinking from related forms of thinking. Using this task, we find that the ability for counterfactual thinking emerges around 5 years of age.

27614 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 10/10/2025

Thanks Evan Orticio (orticio.com) for sharing your fascinating work with us on how children and adults form beliefs without direct evidence. In one super cool study, he shows how children become more diligent fact checkers in less reliable environments. 📃 orticio.com/assets/Ortic...

081 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 02/10/2025

I had a wonderful time visiting UC Irvine to give a talk in the cognitive science colloquium. Thank you @annaleshinskaya.bsky.social for being a fantastic host, and to all the other faculty, students, and postdocs I got to meet during my visit 🙏

3181 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 16/09/2025

Simulating world models supports strong multimodal inferences. Prior work modeled multimodal inference as optimal averaging. But in the "Sound + Ball Occluded" condition each modality alone is useless (only hearing sounds, or only seeing obstacles). Combining both sources reveals what happened!

010 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 16/09/2025

The Sequential Sampler accurately captures people's judgments and eye-movements across the three inference conditions.

100 0

Tobias Gerstenberg @tobigerstenberg.bsky.social · 16/09/2025

The model also predicts eye-movements. It assumes that people look at visual features of the scene, but also at dynamic features that are the consequence of mentally simulating how the ball would fall and collide with the obstacles and walls if it was dropped into the different holes.

100 0