Age-Related Differences in Credit Assignment: A pilot study of model-based and model-free learning
-
Colleen E. Dollst
Institute of Psychology
Universität Hamburg
Hamburg, Germany
colleen.dollst@uni-hamburg.de -
Paul Schreiner
Institute of Psychology
Universität Hamburg
Hamburg, Germany
paul.schreiner-2@uni-hamburg.de -
Rani Moran
Department of Psychology
Queen Mary University of London
London, UK
r.moran@qmul.ac.uk -
Nicolas W. Schuck
Institute of Psychology
Universität Hamburg
Hamburg, Germany
nicolas.schuck@uni-hamburg.de -
Xiangjuan Ren
Institute of Psychology
Universität Hamburg
Hamburg, Germany
xiangjuan.ren@uni-hamburg.de -
Alexa Ruel
Institute of Psychology
Universität Hamburg
Hamburg, Germany
alexa.ruel@uni-hamburg.de
Abstract
Credit Assignment (CA) – the ability to assign value to the reward-generating aspects of an environment or action – is essential for adaptive decision-making but becomes challenging in complex, multi-step environments. Previous research has found that younger adults can flexibly switch between Model-Free (MF) and Model-Based (MB) strategies to solve the credit assignment problem. In contrast, older adults are less likely to exhibit such flexibility, potentially relying more on MF learning due to task representation difficulties. This pilot study explored age-related differences in CA by adapting a dual-bandit task designed to assess MF and MB contributions and their interactions with behavior. An initial sample of six younger (19-25 years) and six older adults (67-73 years) completed a sequential decision-making task involving binary choices between bandits (or pairs) and inferring a hidden option based on observed outcomes. We could assess the contributions of the MF and MB systems by designing two scenarios based on how the chosen bandit and its counterpart were repeated in successive trials. Preliminary results revealed evident MF contributions in both younger and older adults. However, our conclusion of MB contributions was obscured due to the entanglement of the MF learning. Additionally, we found older adults exhibited slower response times and significant learning across blocks, suggesting greater cognitive effort than younger adults. This study provides initial evidence for potential age-related shifts in CA mechanisms, with significant implications for future research. The need for further work integrating neuroimaging and computational modeling to disentangle MF and MB contributions underlying CA in aging is underscored by our findings.
Acknowledgements
This work was supported by a grant from Hamburg University's Ideas and Venture Fund (IVF) for Early Career Researchers, awarded to Dr. Xiangjuan Ren and Dr. Alexa Ruel.