Reinforcement learning: Computational modeling of learning and decision-making

Zach Cohen, Akshay Kumar Jagadish, Eghbal A. Hosseini, Maria K Eckstein
Proceedings of the Analytical Connectionism Schools 2023--2024, PMLR 320:87-101, 2026.

Abstract

Reinforcement learning (RL), as a computational modeling framework, is a formal approach to understanding and building agents, natural or artificial, that learn to make decisions based on rewards they receive from the environment. In this Lecture Notes, we begin by exploring how RL has historically been used in psychology and neuroscience to investigate reward-driven learning, before introducing it more formally from a machine learning perspective. We then demonstrate its utility in building cognitive models that explain the processes and mechanisms underlying human learning and decision-making at both the behavioral and neural levels. Finally, we discuss recent work that brings together the theory-driven approach taken by RL and the data-driven approach taken by artificial neural networks to build more predictive, yet interpretable, models of human behavior. Together, this work highlights the value of RL as a computational modeling framework for cognitive neuroscience.

Cite this Paper

BibTeX
@InProceedings{pmlr-v320-cohen26a, title = {Reinforcement learning: Computational modeling of learning and decision-making}, author = {Cohen, Zach and Jagadish, Akshay Kumar and Hosseini, Eghbal A. and Eckstein, Maria K}, booktitle = {Proceedings of the Analytical Connectionism Schools 2023--2024}, pages = {87--101}, year = {2026}, editor = {Sarao Mannelli, Stefano and Mignacco, Francesca and Chou, Chi-Ning and Chung, SueYeon and Saxe, Andrew}, volume = {320}, series = {Proceedings of Machine Learning Research}, month = {01 Jan--31 Dec}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v320/main/assets/cohen26a/cohen26a.pdf}, url = {https://proceedings.mlr.press/v320/cohen26a.html}, abstract = {Reinforcement learning (RL), as a computational modeling framework, is a formal approach to understanding and building agents, natural or artificial, that learn to make decisions based on rewards they receive from the environment. In this Lecture Notes, we begin by exploring how RL has historically been used in psychology and neuroscience to investigate reward-driven learning, before introducing it more formally from a machine learning perspective. We then demonstrate its utility in building cognitive models that explain the processes and mechanisms underlying human learning and decision-making at both the behavioral and neural levels. Finally, we discuss recent work that brings together the theory-driven approach taken by RL and the data-driven approach taken by artificial neural networks to build more predictive, yet interpretable, models of human behavior. Together, this work highlights the value of RL as a computational modeling framework for cognitive neuroscience.} }
Endnote
%0 Conference Paper %T Reinforcement learning: Computational modeling of learning and decision-making %A Zach Cohen %A Akshay Kumar Jagadish %A Eghbal A. Hosseini %A Maria K Eckstein %B Proceedings of the Analytical Connectionism Schools 2023--2024 %C Proceedings of Machine Learning Research %D 2026 %E Stefano Sarao Mannelli %E Francesca Mignacco %E Chi-Ning Chou %E SueYeon Chung %E Andrew Saxe %F pmlr-v320-cohen26a %I PMLR %P 87--101 %U https://proceedings.mlr.press/v320/cohen26a.html %V 320 %X Reinforcement learning (RL), as a computational modeling framework, is a formal approach to understanding and building agents, natural or artificial, that learn to make decisions based on rewards they receive from the environment. In this Lecture Notes, we begin by exploring how RL has historically been used in psychology and neuroscience to investigate reward-driven learning, before introducing it more formally from a machine learning perspective. We then demonstrate its utility in building cognitive models that explain the processes and mechanisms underlying human learning and decision-making at both the behavioral and neural levels. Finally, we discuss recent work that brings together the theory-driven approach taken by RL and the data-driven approach taken by artificial neural networks to build more predictive, yet interpretable, models of human behavior. Together, this work highlights the value of RL as a computational modeling framework for cognitive neuroscience.
APA
Cohen, Z., Jagadish, A.K., Hosseini, E.A. & Eckstein, M.K.. (2026). Reinforcement learning: Computational modeling of learning and decision-making. Proceedings of the Analytical Connectionism Schools 2023--2024, in Proceedings of Machine Learning Research 320:87-101 Available from https://proceedings.mlr.press/v320/cohen26a.html.

Related Material