Theory of mind across biological and artificial embodiment: theory, experiments and computational models

A recurrent, yet still incredibly interesting, debate surrounds the development of human Theory of Mind (ToM), as well as the identification of the mechanisms and factors implicated in this cognitive ability. In addition to psychology, this debate has attracted several research fields, including robotics for building social robots. Ultimately, while humans represent the best model of ToM for implementing a machine ToM for social robots, unresolved questions regarding human ToM development and ability need addressing. In this thesis, I used a mixed approach involving different disciplines and methodologies to contribute to human and machine ToM. Specifically, through research in the infant (Part 1 of this thesis), as well as the limb difference and general populations (Part 2), and computational modelling (Part 3), I aimed at investigating (i) ToM emergence, (ii) the mechanisms underlying ToM ability and development, and (iii) factors implicated in this cognition. Merging findings from experimental and computational modelling research, I contribute to these topics in three important ways. First, I provide support for early ToM emergence and indicate its advantage towards improved prediction of others’ behaviours. Second, I suggest the coexistence of the association, simulation, and teleological for mentalising mechanisms and their collaboration for achieving ToM in different scenarios. Third, I identify sensorimotor-driven embodiment and perspective taking as factors implicated in ToM ability and development, while suggesting that self-other similarity may not be a requirement for ToM. Whether multisensory integration and mental rotation impact ToM remains undetermined. Overall, my multidisciplinary findings provide insights into human and machine ToM. Specifically, experimental findings increase our general knowledge of human ToM, computational findings inform the implementation of social robotic architectures, whilst knowledge from both disciplines is exploited to advance one another and lead to a more global understanding of ToM. Future studies are warranted to validate and extend these results.