STAI CDT PhD student, Lennart Wachowiak (2021 cohort), presented a paper and poster on “A Survey of Evaluation Methods and Metrics for Explanations in Human-Robot Interaction (HRI)” at the Explainable Robotics Workshop at IEEE International Conference on Robotics and Automation (ICRA) 2023.
The paper and poster focus on the crucial role of explanations in making human-robot interactions safe and trustworthy and is a comprehensive overview of methods used to evaluate these explanations.
As Lennart explains, “The ability to explain its behavior is essential for a collaborative robot to foster a reliable and trustworthy human–robot interaction. However, when researchers develop a new way of generating a robot explanation, they need some means to test whether the explanation is any good. Since I will encounter that situation myself, I wrote a survey of how you can evaluate a new robot explanation. I looked at a variety of methods ranging from participatory design, in which you listen to user feedback before you even start developing your explanation method, to more quantitative metrics that can be automatically computed”.
In addition to presenting the paper and poster, Lennart also got involved with other ICRA activities. Lennart says, “Later in the week, I volunteered at one of the ICRA competitions, where university teams had to program a robot that finds people in an apartment and then brings them a requested item. As a volunteer, I was tasked with judging the teams’ performances and being an actor in the apartment, asking the robot to bring me different objects”.
Lennart enjoyed the opportunity to meet fellow researchers with similar interests and research areas. He says, “Regarding the conference itself, the explainable robotics workshop was the highlight, as I got to know people looking into similar things as me. One of the workshop organizers also asked me to be a guest on the Talking Robotics podcast, which was a fun experience”.
The paper is co-authored with Oya Celiktutan, Andrew Coles and Gerard Canal, and you can find out more about the paper here: A Survey of Evaluation Methods and Metrics for Explanations in Human–Robot Interaction (HRI) — King’s College London (kcl.ac.uk)