@gershbrain.bsky.social and I have a new paper in PLOS Comp Bio!

We study how two cognitive constraints—action consideration set size & policy complexity—interact in context-dependent decision making, and how humans exploit their synergy to reduce behavioral suboptimality.

osf.io/preprints/ps...

If you work on artificial or natural intelligence and are finishing your PhD, consider applying for a Kempner research fellowship at Harvard:
kempnerinstitute.harvard.edu/kempner-inst...

People learn from rewards differently when outcomes are shared with others. Learning slows when receiving smaller reward shares, and social stereotypes about partners further impair learning with cognitive demands.
@hamhuang.bsky.social @annajenkins.bsky.social
www.nature.com/articles/s44...

Great mentor! Can’t go wrong!

Imagine this scenario: would you rather me giving $2 to you and $1000 to a politician or not giving you or the politician any money at all? Many people would choose the latter option, even though the first option gives them more monetary benefits.

People don’t always choose the option with more money because what people find rewarding is not simply the amount of money, but also social equity. When deciding which option is more rewarding, people consider both the amount of money and the inequity.

However, making decisions is different from learning from rewards. People always know which option feels more rewarding to them when making decisions, but learning from reward can happen very implicitly. People learn to choose the rewarding options without knowing it.

How can social inequity matter if people may not even realize that they learned from reward? We found that surprisingly, it matters! A lot! This suggests that social inequity profoundly impacts reward representation at all levels.

In 3 experiments (N=280), subjects tried to learn which of 3 keys gave more reward ($0, $1, or $2) in different contexts. The reward was split with another person. We manipulated the inequity level from 0% to 100%. The keypress only determined the total reward.

We found a huge inequity effect. When subjects received less than 50% of the reward, they learned more slowly and less effectively. Moreover, they were more sensitive to inequity levels under 50%. For example, 80% vs 90% made less difference to learning quality than 10% vs 20%.

We discovered that reinforcement learning is also socially contextual! We are currently working on an fMRI study to tackle follow-up questions about how the social brain ties to the learning brain, led by Mu-chen and in collaboration with Joe Kable. Stay tuned!

🚨New paper!🚨My new work with @annajenkins.bsky.social is published
@commspsychol.nature.com
! Social inequity shapes people’s ability to learn from rewards during reinforcement learning! www.nature.com/articles/s44...

People don’t always choose the option with more money because what people find rewarding is not simply the amount of money, but also social equity. When deciding which option is more rewarding, people consider both the amount of money and the inequity.

Imagine this scenario: would you rather me giving $2 to you and $1000 to a politician or not giving you or the politician any money at all? Many people would choose the latter option, even though the first option gives them more monetary benefits.

Here is the full poster ;)

The CoLab is headed to #CogSci2025!! 🥳 Here's where to find us!

"We introduce Collaborative Rational Speech Act (CRSA), an information-theoretic (IT) extension of RSA that models multi-turn dialog by optimizing a gain function adapted from rate-distortion theory."

arxiv.org/abs/2507.14063

Excited to see the work during my undergrad times finally come out! Learned so much from working with Sam and Anne on this project.

We are so excited to welcome our summer RA's to the lab! Looking forward to a productive summer💫 @bellafascendini.bsky.social @hamhuang.bsky.social @creppy-r.bsky.social