Yoking-Based Identification of Learning Behavior in Artificial and Biological Agents

We want to understand how animals can learn to solve complex tasks. To achieve this, it makes sense to first hypothesize learning models and then compare these models to real biological learning data. But how to perform such a comparison is still unclear. We propose that yoking is an important component to such an analysis. In yoking, two agents are made to experience the same inputs, rewards or perform the same actions – possibly in combination. We use yoking as an analytical tool to identify the algorithm that drives learning in a target agent. We evaluate this approach in a synthetic task, where we know the ground truth learning algorithm. Then we apply it to biological data from a physical puzzle task, to identify the learning algorithm behind physical problem solving in Goffin’s cockatoos. Our results show that yoking works, and can be used to identify the target algorithm more reliably, with less variance and assumptions, than a more unconstrained approach to identify learning algorithms.