Human Assessment of Robot Capability

Abstract

In this thesis, we have explored the relationship between the perceived cognitive intelligence of robots and the trustworthiness they inspire in their human collaborators. To achieve this, we ran two classic trust games whose experimental conditions were the perceived cognitive intelligence of the robot. With the first study (Price Game), we measured participants' willingness to change their minds and follow the robot's suggestion on the matter of object pricing. In the second study (Investment Game), we measured how much participants were inclined to invest financially in the robot. In both experiments, the experimental conditions depended on the participant watching a video that depicted the robot as either passing the Sally-Anne test (i.e. cognitively intelligent) or failing the Sally-Anne test (i.e. not cognitively intelligent). Building upon the findings and limitations of the first two studies, a third experiment was conducted to further investigate the application of thematic analysis in capturing the mental modelling process of a robotic agent during a predictive game. As a result, we managed to extract a first draft of a mental modelling of a robotic agent (``Theory of Robotic Mind''). The results of this study are promising, supporting the use of a qualitative protocol in constructing a framework for monitoring the belief system of humans during their interactions with a robotic companion.

Date of Award	1 Aug 2025
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Caroline Jay (Supervisor) & Bijan Parsia (Supervisor)