A Robotic Approach to Understand the Role of Vicarious Trial-and-Error in a T-Maze Task

Vicarious trial-and-error(VTE) is a type of conflict-like behavior, observed in route selection tasks (Tolman (1939)). Studies of VTE have shown a correlation between the number of VTEs exhibited by a system with its learning efficiency. At the onset of learning a task, the number of VTEs increases, and when the learning reaches its plateau, it decreases. The question we explore in this paper concerns the role of VTE. Basing ourselves on a model developed by Bovet and Pfeifer (2005), we ran robotic experiments to compute the number of VTEs during the learning of a T-maze task. Our results first show that what has been found in rats can be replicated in artificial systems. Furthermore, by changing the connectivity pattern of the original model, we discovered that the connection between VTEs and learning efficiency might not be necessarily true as our results show that two models exhibiting the same performance can possess a different pattern of VTEs. By comparing the robustness of the two models under varied conditions, we propose that VTEs are connected to the adaptivity of a system to environmental changes.