Human-agent trust relationships in a real-time collaborative game

Abstract

Collaborative virtual agents are often deployed to help users make decisions in real-time. For this collaboration to work, users must adequately trust the agents that they are interacting with. In my research, I use a game where human-agent interactions are recorded via a logging system and survey instruments in order to explore this trust relationship. I then study the impact that different agents have on reliance, performance, cognitive load and trust. I seek to understand which aspects of an agent influence the development of trust the most. With this research, I hope to pave the way for trust-aware agents, capable of adapting their behaviours to users in real-time.