Evaluating and benchmarking the interactions between a humanoid robot and a virtual human for a real-world social task

We developed two social agents (a virtual human and a humanoid robot) with various similar functionalities, interaction modalities, intelligence, autonomy etc. and integrated them through a common communication platform based on a novel control algorithm to assist each other in a real-world social task (searching for a hidden object).We also studied human's interactions with those social agents and with some other allied agents for that task to benchmark the interactions. We developed the standards of the performances as well as the performance measurement methods for the agents for the task. We also adopted several hypotheses regarding the attributes and performances of the agents for their interactions for the task. We evaluated the attributes and performances of the robot and the virtual human in their interactions for the task, analyzed them and compared them with the standards. The results showed that both the robot and the virtual human performed satisfactorily in their social interactions though the performances varied slightly.We also found a trade-off between the attributes and the performances of the agents. The results will help develop intelligent social agents of different realities to assist humans in various real- world social tasks, or to get the real-world social tasks done in cooperation between artificial social agents of different realities.