[ACT-R-users] Cognitively-plausible Reinforcement Learning in Agent-based Modeling

[Konstantinos Mitsopoulos]

Dear all,

I would like to share our new publication that may be of interest to the
ACT-R and cognitive modeling community:

*Cognitively-plausible reinforcement learning in epidemiological
agent-based simulations*
Konstantinos Mitsopoulos, Lawrence Baker, Christian Lebiere, Peter Pirolli,
Mark Orr, Raffaele Vardavas
🔗 <https://www.frontiersin.org/articles/10.3389/fepid.2025.1563731/full>*Frontiers
in Epidemiology, 2025
<https://www.frontiersin.org/articles/10.3389/fepid.2025.1563731/full>*
🔹We present a framework that integrates the ACT-R cognitive architecture
with agent-based modeling. This replaces fixed-rule agents with learning
agents that adapt dynamically, without parametric training (e.g. gradient
descent), while remaining cognitively interpretable.
🔹We show how this is achieved by grounding the architecture in *Statistical
Learning Theory*, enabling it to perform both *Supervised Learning*
(regression, classification) and *Reinforcement Learning*.

The framework can be extended to support other learning paradigms
(e.g. *Multi-agent
RL* to account for *group-level decision making*) and to integrate
components such as *LLMs* for additional capabilities.

We demonstrate the framework in the context of epidemiological simulations,
but the methodology is applicable to social simulations, policy modeling,
and beyond.

If you have any questions, feedback, or thoughts, feel free to email me - I
would be happy to discuss.

Best regards,

Konstantinos Mitsopoulos
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/act-r-users/attachments/20250807/07064a24/attachment.html>