Effects of Agent Transparency on Multi-Robot Management Effectiveness

Abstract

Abstract : The objective of the study was to investigate the effects of agent transparency on operator performance in the context of joint human-agent decision making in multi-robot management. The agent display configurations were based on the 3 levels of the situation awareness-based agent transparency model (basic-information, reasoning, and projections/uncertainty). Results showed that participants calibrated their trust in the agent more effectively (proper reliance and correct rejections) and reported higher levels of trust when they were provided with the agent's reasoning and uncertainty information. No speed-accuracy trade-offs were observed. Nor did the participants report higher levels of workload when agent transparency increased. Working memory capacity was found to be a significant predictor of participants' trust in the agent. Individual differences in spatial ability accounted for variations in ocular indices of workload across display configurations.