Lennart Wachowiak and Peter Tisnikar of the 2021 cohort of the UKRI Centre for Doctoral Training in Safe and Trusted Artificial Intelligence presented their work on the paper ‘Analysing Eye Gaze Patterns during Confusion and Errors in Human–Agent Collaborations’ at the 31st IEEE International Conference on Robot & Human Interactive Communication (RO-MAN 2022) and the 39th IEEE International Conference on Robotics and Automation (ICRA 2022) respectively.
Lennart, Peter and their co-authors investigated people’s gaze patterns during human-agent collaborations. As Lennart explains, “With this paper, we explored whether eye gaze signals can tell us if something goes wrong during a human–agent collaboration, for instance, the robot making mistakes or the human not knowing what they should do next. Being able to interpret social cues such as gaze is vital for social robots that want to proactively provide explanations and assistance at the right time. The scenario we looked at is a simulated cooking task in which a human has to prepare dishes with an AI agent.”
They found that people’s gaze patterns did indicate if something goes wrong in the collaboration. Their research is a key step towards developing robots that can understand and adapt to their collaborators and towards building people’s positive perception and trust in robots.
Lennart presented the paper at the RO-MAN 2022 conference in Naples, and said, “I not only enjoyed lots of pizza and pasta but also met tons of other like-minded researchers working on the interplay between robots, humans, and society.”
Peter Tisnikar presented a snapshot of their work at the ICRA’22 workshop ‘Prediction and Anticipation Reasoning in Human Robot Interaction’ in Philadelphia.
The paper is co-authored with Gerard Canal, Andrew Coles, Matteo Leonetti and Oya Celiktutan.