Capturing the complexity of human strategic decision-making with machine learning
Abstract
Strategic decision-making is a crucial component of human interaction. Here we conduct a large-scale study of strategic decision-making in the context of initial play in two-player matrix games, analysing over 90,000 human decisions across more than 2,400 procedurally generated games that span a much wider space than previous datasets. We show that a deep neural network trained on this dataset predicts human choices with greater accuracy than leading theories of strategic behaviour, revealing systematic variation unexplained by existing models. By modifying this network, we develop an interpretable behavioural model that uncovers key insights: individuals' abilities to respond optimally and reason about others' actions are highly context dependent, influenced by the complexity of the game matrices. Our findings illustrate the potential of machine learning as a tool for generating new theoretical insights into complex human behaviours.
Study specs
Deep neural networks trained on data from procedurally generated matrix games with over 2,400 variations; models were modified for interpretability.
- Authors
- JQ Zhu,JC Peterson,B Enke,TL Griffiths
- Discipline
- Artificial Intelligence
- Sample Size
- N=90,000
- Study Type
- Experimental Study
- Year
- 2025
- Human Data Platform
- Prolific
- Source
- View Source DOI Google Scholar
Measured Outcomes
Human choices and reasoning in initial play of two-player matrix games, focusing on strategic decision-making and response to game complexity.
Peer Review & Critical Discussion
Potential Selection Bias in 2023 Cohort
The participant pool shows a concerning overrepresentation of users from high-income demographics. Looking at Table 3, we can see that 78% of respondents had annual incomes above $75k, which significantly limits the generalizability of these findings to broader populations.
Non-naive Participants Issue
I've noticed a methodological concern regarding participant naivety. Given that Prolific users often complete multiple studies, there's a real risk that participants had prior exposure to similar experimental paradigms, which could confound the results.
RLHF Applicability to This Study Design
The implications for RLHF training pipelines are understated. If we accept the authors' conclusions about preference stability, this has direct consequences for how we should structure reward model training. The temporal decay effect described in Section 4.2 is particularly relevant.
Verify your expertise to join discussion
Create an account and verify your credentials to participate in peer discussions.