Toward a Psychology of Deep Reinforcement Learning Agents Using a Cognitive Architecture.

We argue that cognitive models can provide a common ground between human users and deep reinforcement learning (Deep RL) algorithms for purposes of explainable artificial intelligence (AI). Casting both the human and learner as cognitive models provides common mechanisms to compare and understand their underlying decision-making processes. This common grounding allows us to identify divergences and explain the learner's behavior in human understandable terms.

Saved information is remembered less well than deleted information, if the saving process is perceived as reliable.

Prior research has shown that people are more likely to remember information that is deleted from a computer than information that is saved on a computer, presumably because saving serves as a form of cognitive offloading. Given recent concerns about the robustness and replicability of this "Google Effect," we conducted two experiments seeking to replicate and extend the phenomenon by identifying a potential boundary condition for when it is observed. In Experiment 1, we replicated the Google Effect, but only when participants experienced a practice phase demonstrating the reliability of the saving process.