Reinforcement learning as a foundation for psychodynamics
Abstract
In this article we consider the use of reinforcement learning (RL) as a formal framework for understanding traditional psychodynamic phenomena. We show how RL can be used to provide new insights as well as normative explanations for key psychodynamic constructs, grounding them in principles of reward maximization and adaptation. We illustrate this by reviewing two recent studies in which simulations of adaptive agents offer new explanations for psychodynamic processes, discuss how RL could account for both similarities as well as individual differences in human psychological profiles, and outline possible future directions for this approach. We conclude that, by providing a computational foundation for psychodynamic concepts, RL may contribute to the development of a more rigorous science of psychological dynamics and their disruption in clinical conditions.
Related
- RNASeek: A Cross-Phyla Generative Foundation Model for Multipurpose RNA Modeling and Reinforcement Learning-Based Design
- Reliability-based arbitration reveals heterogeneous model-based and model-free reinforcement learning strategies linked to compulsivity and impulsivity
- Increased monetary incentives enhance goal-directed reinforcement learning but fail to suppress automatic learning processes
- Reference-point dependent reinforcement learning in humans and rats
- Hippocampal-cingulate dynamics in the human brain link reinforcement-learning and memory
BibTeX
@misc{dulberg2026reinforcement,
title = {{Reinforcement learning as a foundation for psychodynamics}},
author = {Zack Dulberg and Jonathan D. Cohen},
howpublished = {PsyArXiv},
year = {2026},
month = sep,
doi = {10.31234/osf.io/vcnxt_v3},
url = {https://osf.io/preprints/psyarxiv/vcnxt_v3/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Reinforcement learning as a foundation for psychodynamics
AU - Zack Dulberg
AU - Jonathan D. Cohen
PY - 2026
DA - 2026/09/04
PB - PsyArXiv
DO - 10.31234/osf.io/vcnxt_v3
UR - https://osf.io/preprints/psyarxiv/vcnxt_v3/
AB - In this article we consider the use of reinforcement learning (RL) as a formal framework for understanding traditional psychodynamic phenomena. We show how RL can be used to provide new insights as well as normative explanations for key psychodynamic constructs, grounding them in principles of reward maximization and adaptation. We illustrate this by reviewing two recent studies in which simulations of adaptive agents offer new explanations for psychodynamic processes, discuss how RL could account for both similarities as well as individual differences in human psychological profiles, and outline possible future directions for this approach. We conclude that, by providing a computational foundation for psychodynamic concepts, RL may contribute to the development of a more rigorous science of psychological dynamics and their disruption in clinical conditions.
ER -