Research

Relative value learning

Decision makers sometimes have to use feedback from the environment to learn the values of different choice options. When options are repeatedly encountered in separate contexts, people may learn the option values in a context-dependent fashion. Instead of encoding absolute outcomes ("Option X won $25 and Option Y won $20"), people behave as if they encode relative outcomes (“Option X wins more than Option Y”). Learning relative outcomes in one context and then using them to make decisions in a new context may lead to "irrational" preferences for options with lower payoffs.

We are currently using eye-tracking and computational modeling to study the relationship between visual fixation patterns and learned option values in sequential decision tasks.

See our latest preprint: A Reinforcement Learning and Sequential Sampling Model Constrained by Gaze Data

Relevant papers and presentations

Hayes, W. M., & Wedell, D. H. (2023). Testing models of context-dependent outcome encoding in reinforcement learning. Cognition, 230, 105280. (link)

Hayes, W. M., & Wedell, D. H. (2023). Reinforcement learning in and out of context: The effects of attentional focus. Journal of Experimental Psychology: Learning, Memory, and Cognition, 49(8), 1193-1217. (link)

Hayes, W. M., & Wedell, D. H. (2023). Effects of blocked versus interleaved training on relative value learning. Psychonomic Bulletin & Review, 30, 1895-1907. (link)

Hayes, W. M., Yax, N., & Palminteri, S. (2025). Relative value encoding in large language models: A multi-task, multi-model investigation. Open Mind, 9, 709-725. (link)

Touchard, M. J., & Hayes, W. M. (2025). "Can visual fixations explain context-dependent reinforcement learning?" Accepted to CogSci 2025.

Hayes, W. M., & Touchard, M. J. (submitted). A reinforcement learning and sequential sampling model constrained by gaze data. (link)

Context effects in decision making

In decisions between multiple alternatives with multiple attributes, a person's relative preference for one alternative over another often depends on other, "irrelevant" alternatives. These context effects are notoriously fragile and can disappear or even reverse depending on the way that information is presented. The fragile nature of context effects may be partly due to attentional processes. Different choice presentation formats facilitate certain types of comparisons while making other types of comparisons more difficult, thereby decreasing the likelihood of attending to the latter. For example, incommensurable attributes, such as CPU speed (GHz) and RAM size (GB) for laptops, should be more difficult to compare than attributes on a common scale, such as quality ratings. The varying levels of attention allocated to different attribute comparisons may then influence the occurrence or nonoccurrence of context effects.

We are currently planning experiments that will leverage process tracing methods (e.g., eye-tracking) to directly test this hypothesis.

Relevant papers and presentations

Hayes, W. M., Holmes, W. R., & Trueblood, J. S. (2025). Attribute commensurability and context effects in preferential choice. Psychonomic Bulletin & Review, 32, 737-748. (link)

Trueblood, J. S., Liu, Y., Murrow, M., Hayes, W. M., & Holmes, W. R. (revision submitted). Attentional dynamics explain the elusive nature of context effects.

Page updated

Report abuse