Alan R. Wagner presented at the latest virtual IEAI Speaker Series on October 14, 2021.
The topic of the presentation was “To Trust a Robot: Examining Human-Robot Trust during Emergency Evacuations”. Dr. Wagner examined the opportunities and challenges at the intersection of human-robots interaction in safety critical contexts. Particularly interesting was the analysis of the dynamics that lead individuals to delegate trust to robots in emergency situations and the related ethical issues at the intersection of trustworthiness and autonomy.
Dr. Wagner showed several simulations to explain the role played by robots in such scenarios, as well as, the opportunities and challenges. The goal of the research is to develop emergency evacuation robots, that can reduce the number of emergency personnel needed, reduce the response time and immediately react to face an emergency to ultimately increase the number of people that can be saved. For instance, robots can be very helpful to guide people to different exits in emergency situations to avoid chaos that could endanger the individuals even more
A fundamental aspect of this research is the analysis of trustworthiness in such human-robot interaction. For instance, in case of poor performance of the robot, it becomes challenging to repair the broken trust of the individuals in order to conduct a more successful operation. The way the robot should operate to repair such trust is not trivial and the results of the simulations showed that the timing is fundamental. If the robots would apologize for the poor performance right after the mistake, there would not be a considerable difference in the way people respond. On the other hand, if such apology would be done later, the trust would be completely repaired.
To test the outcome of the deployment of such robots in a potential real-world scenario as well as analyzing how subjects would react to the robot´s guidance, tests were conducted in physical simulations. The results suggested that in moments of fear all people have the tendency to follow the robot. However, such high trust can be considered worrysome because total trust can be dangerous in the case the robots would perform poorly, which is a similar tendency that has been detected in studies concerning the interaction with people and autonomous vehicles. In conclusion, Dr. Wagner explained the technical challenges of building emergency evacuation robots and the necessity to conduct more tests to accurately measure how people react to such machine´s deployment and how to mitigate the unintended consequences, such as automation bias, which leads to overreliance to such machines.