Neural Networks Uncover Human Decision-Making Patterns through Symbolic Regression

Strategic Insights -

Researchers have made a breakthrough in understanding human decision-making by combining artificial neural networks (ANNs) with symbolic regression. This approach allows for the extraction of an expressive and interpretable model that specifies how individuals evaluate decision-relevant information during choice. The model was able to account for behavior in the researchers' own data and in previous work, outperforming existing accounts of information sampling such as the Upper Confidence Bound heuristic. This modeling approach has broad potential for uncovering novel patterns in behavior and cognitive processes, while also specifying them in human-interpretable formats.

Key Takeaways:

  • Researchers used artificial neural networks (ANNs) combined with symbolic regression to develop a model that explains how humans evaluate decision-relevant information during choice.
  • The model accounted for behavior in the researchers' own data and in previous work, outperforming existing accounts of information sampling such as the Upper Confidence Bound heuristic.
  • The model specified roles for midbrain dopaminergic nuclei, anterior cingulate cortex, and anterior insula in mediating the influence of value of information on behavior.
  • The research used ultra-high field neuroimaging to examine activity across a suite of subcortical neuromodulatory nuclei and two cortical regions.
  • The approach has broad potential for uncovering novel patterns in behavior and cognitive processes.
  • The model provides a human-interpretable format for understanding decision-making processes.
  • The research focused on midbrain dopaminergic nuclei, anterior cingulate cortex, and anterior insula in mediating the influence of value of information on behavior.

Statistics:

  • The model outperformed existing accounts of information sampling such as the Upper Confidence Bound heuristic.
  • The research used ultra-high field neuroimaging to examine activity across a suite of 5 subcortical neuromodulatory nuclei and 2 cortical regions.
  • The study analyzed data collected from human participants during choice tasks.
  • The value of information derived by the model was used to examine activity across a suite of subcortical neuromodulatory nuclei and two cortical regions.

Sources:

  • biorxiv.org/content/10.1101/2025.06.24.661282v1