I also recommend Gervais et al.'s thoughtful reply: pnas.org/doi/10.1073/...

(As they highlight, their broader arguments about cultural evolution and religion are not the target of my critique, and their cross-cultural findings remain intriguing despite the interpretive difficulties)

So this commentary is my small contribution to the skeptic's side of that broader debate.

One motivation for writing this commentary: There is a large appetite in psychology for methods for covertly measuring people's implicit attitudes and preferences. But these efforts have mostly failed to live to the hype: www.nature.com/articles/s44...
wires.onlinelibrary.wiley.com/doi/abs/10.1...

The problem is that this strategy would also allow us to 'infer' that people have intuitive preferences for nuclear explosions, or sending people to concentration camps:

People tend to attribute more intentionality to agents that did something bad. It is tempting to use this effect as a way to covertly measure people's implicit attitudes, as Will Gervais and colleagues do in a recent paper on attitudes toward atheism:
www.pnas.org/doi/10.1073/...

Imagine if all the money academics pour into OA fees at for-profit publishing corporations went instead to academic societies, which exist to support science; or to university presses, which exist to support academia

Link to the paper: escholarship.org/uc/item/86c4...

For more in this line of research see also our model of resource-limited Theory of Mind: bsky.app/profile/tade...

Overall, our work makes sense of why people tend to rely on social stereotypes more than they should if they were idealized Bayesian observers: they tend to preferentially allocate their limited cognitive resources to encoding group membership information.

We also find that the optimal policy exhibits the same 'outgroup homogeneity bias' as people: it tends to represent outgroup members are more similar to each other than they are. Again, this is especially the case when cognitive resources are very limited.

This explains why social stereotypes often act as 'energy-saving' devices for prediction, as social psychologists have found: people are for example more likely to use stereotypes when they are under cognitive load.

Our key result: under limited resources, the optimal policy preferentially encodes information about group membership (blue) and tends to discard individuating information (teal).

This tendency only reverses if group membership has very low task relevance.

We study agents who have to predict the behavior of other agents.

Agents have limited cognitive resources: they can only extract so much information from the environment, and have to prioritize which information to encode (group membership or individuating info).

When predicting someone's behavior, people typically rely on both:
-individuating information (e.g. what that person did in the past),
-group membership,
We want to explain why people integrate these two types of information as they do, working from first principles.