Improving robot transparency: Real-time visualisation of robot AI substantially improves understanding in naive observers

摘要

Deciphering the behaviour of intelligent others is a fundamental characteristic of our own intelligence. As we interact with complex intelligent artefacts, humans inevitably construct mental models to understand and predict their behaviour. If these models are incorrect or inadequate, we run the risk of self deception or even harm. Here we demonstrate that providing even a simple, abstracted real-time visualisation of a robot's AI can radically improve the transparency of machine cognition. Findings from both an online experiment using a video recording of a robot, and from direct observation of a robot show substantial improvements in observers' understanding of the robot's behaviour. Unexpectedly, this improved understanding was correlated in one condition with an increased perception that the robot was `thinking', but in no conditions was the robot's assessed intelligence impacted. In addition to our results, we describe our approach, tools used, implications, and potential future research directions.