How do individuals prefer to interact with robots in crowded environments? What algorithms are suitable for roboticists to program robots for human interaction?
A group of mechanical engineers and computer scientists from the University of California San Diego delved into these questions in a recent study presented at the ICRA 2024 conference in Japan.
The study, led by Aamodh Suresh and Angelique Taylor, aimed to investigate how humans perceive risk when interacting with robots in everyday scenarios. Suresh, a postdoctoral researcher for the U.S. Army Research Lab and Taylor, now a faculty member at Cornell Tech, developed a framework to gauge human risk aversion in robot interactions by drawing on models from behavioral economics.
To gather data, the researchers conducted an online experiment during the pandemic. Participants, mainly STEM students, engaged in a simulated game where they acted as Instacart shoppers selecting a path in a grocery store to reach the milk aisle. The paths varied in risk levels related to potential exposure to COVID-19, with some paths passing by individuals with the virus. Participants were incentivized to choose the fastest route, despite the associated risks.
The study revealed that individuals often underestimated their willingness to take risks when there was a reward involved. This observation led the researchers to adopt prospect theory, a concept from behavioral economics, to program robots for human interactions. Prospect theory, pioneered by Nobel laureate Daniel Kahneman, suggests that individuals tend to weigh losses more heavily than gains when making decisions.
Furthermore, the study explored how individuals preferred robots to communicate intentions, with responses ranging from speech to gestures and touchscreens.
Future plans include conducting an in-person study with a more diverse participant pool to further investigate human risk aversion in robot interactions.
