Human Trust Between Real and Virtual Agents for a Real-World Assistive Task

The objective is investigating how humans perceive and compare their trust between similar real (physical) and virtual intelligent agents (a social robot and a virtual human) in real-world assistive collaborative tasks performed between humans and the agents. An intelligent anthropomorphic humanoid robot and a human-like intelligent virtual human were developed. Each agent was made capable of separately performing a specific real-world task such as finding an object (a small doll) in a home setting for an old or disabled human. An experiment was designed and conducted to investigate the factors that influenced human trust in the agents and compare human trust between the agents for the same real-world task. We learned from the results that the trust factors were almost the same for both social agents though humans trusted the robot more than the virtual human for the same real-world task. It was assumed that the difference in embodiment between the agents created the difference in human trust between them (the agents). We proposed to use the results obtained from the study for developing trustworthy collaborative intelligent social agents performing real-world complex tasks for humans.