Explaining Reinforcement Learning to Mere Mortals: An Empirical Study
This addresses the problem of making RL more accessible to non-experts, though it is incremental as it builds on existing explanation methods.
The study investigated how different explanations affect non-experts' understanding of reinforcement learning agents in a game, finding that combining saliency maps and reward-decomposition bars significantly improved mental model scores compared to a control group.
We present a user study to investigate the impact of explanations on non-experts' understanding of reinforcement learning (RL) agents. We investigate both a common RL visualization, saliency maps (the focus of attention), and a more recent explanation type, reward-decomposition bars (predictions of future types of rewards). We designed a 124 participant, four-treatment experiment to compare participants' mental models of an RL agent in a simple Real-Time Strategy (RTS) game. Our results show that the combination of both saliency and reward bars were needed to achieve a statistically significant improvement in mental model score over the control. In addition, our qualitative analysis of the data reveals a number of effects for further study.