Comparative and animal cognition literature describes many models of associative learning and a multitude of experimental protocols for exploring learning phenomena. These methodologies can serve as inspiration for reinforcement learning (RL) algorithms and tasks (Shanahan et al., 2020). However, there is a considerable gap between animal cognition and RL research, both conceptually and in the assumptions made about the learning process. Associative learning models assume the presence of a “stimulus” guiding a behavioural response, which in the field of RL usually translates loosely into a state-action pair. Our research attempts to investigate and bridge this gap by implementing the A-Learning model (Ghirlanda et al., 2020) into an embodied AI system, using the purpose-built Animal-AI environment (Beyret et al., 2019). Here we present early findings of our research.

This content is only available as a PDF.
This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. For a full description of the license, please visit