iiinova — where your digital self keeps evolving

David Schultner @davidschultner.bsky.social · 28/07/2025

🎊 New paper out! In this @cp-trendscognsci.bsky.social Forum, we (with @lucasmolleman.bsky.social and @bjornlindstrom.bsky.social) summarize how reward learning can lead to adaptive social learning. We also explore the broader consequences for cultural evolution: www.cell.com/trends/cogni... 🚄

2247

David Schultner @davidschultner.bsky.social · 23/07/2025

Going beyond these experiments, we explore the implications of the SFL model for social learning under a variety of environmental contexts (such as spatial & environmental variability or dangerous environments, check out the full paper for these agent-based simulations and much much more).

130

David Schultner @davidschultner.bsky.social · 23/07/2025

In a final experiment (Exp. 6), we find evidence suggesting that learning about social- and non-social features follows the same principles. We do so by establishing feature competition (i.e., shared associative strength between social- and non-social features, see full text for details).

120

David Schultner @davidschultner.bsky.social · 23/07/2025

This suggests that SLS can be shaped by rewards. We show that among multiple available features, people pick out reward-predictive ones (Exp. 3), that this learning pattern also holds with 4 (instead of 2) choice options (Exp. 4), and that learning generalizes to dissimilar contexts (Exp. 5)

120

David Schultner @davidschultner.bsky.social · 23/07/2025

Exps. 1 & 2: Consistent with our model, social learning was shaped by rewards: Having learnt that the majority (or minority) choice is reward predictive, pps copied the majority (or minority) when encountering novel targets. Exp. 2 replicates this with others' payoffs instead of choice behaviour.

110

David Schultner @davidschultner.bsky.social · 23/07/2025

☀️ We advance a domain-general reinforcement learning model—the Social Feature Learning (SFL) model—explaining SLS as the result of associating social features (e.g., others choices, their payoffs, or their age) with rewards. We test core assumptions and predictions across 6 experiments (n = 1941).

130

David Schultner @davidschultner.bsky.social · 23/07/2025

💥 Our new paper (with ‪@lucasmolleman.bsky.social and ‪@bjornlindstrom.bsky.social‬) is now out in @nathumbehav.nature.com‬ 🥳 www.nature.com/articles/s41... 🧠 Here, we advance a novel RL account—the Social Feature Learning (SFL) model—that explains how people learn to learn from others! 🤝 🧵👇

1319

David Schultner @davidschultner.bsky.social · 11/03/2025

Having established these two separable factors, how may their interplay affect larger social systems? Using agent-based simulations, we find: 🔹 Inequality aversion helped societies move toward prosocial states (blue) 🔹 Commonness bias reinforced selfish defaults (red) instead of promoting fairness

130

David Schultner @davidschultner.bsky.social · 11/03/2025

How do both factors interact in a shared setting? In Study 2, we combined both Study 1 games into a novel game, manipulating inequality & commonness independently. Results closely track those of Study 1: Both motivations shaped moral judgments, but were only weakly related.

120

David Schultner @davidschultner.bsky.social · 11/03/2025

In Study 1, participants judged others’ behavior in economic games designed to isolate each motivation. Results show distinct effects of inequality aversion & commonness, however: only weak associations between both factors emerged, pointing to separate cognitive contributions

110

David Schultner @davidschultner.bsky.social · 11/03/2025

Having established separable factors, we asked: how may their interplay affect larger social systems? 🤔 Using agent-based simulations, we found: 🔹Inequality aversion helped societies move toward prosocial states (blue) 🔹Commonness bias reinforced selfish defaults (red) instead of promoting fairness

100

David Schultner @davidschultner.bsky.social · 11/03/2025

How do both factors interact in a shared setting? In Study 2, we combined both Study 1 games into a novel game, manipulating inequality & commonness independently. Results closely mirrored those of Study 1: Both motivations shaped moral judgments, but were only weakly related.

100

David Schultner @davidschultner.bsky.social · 11/03/2025

100