A psychology and game theory approach to human–robot cooperation

Social robots have great practical potentials to be applied to, for example, education, autism therapy, and commercial settings. However, currently, few commercially available social robots meet our expectations of ‘social agents’ due to their limited social skills and the abilities to maintain smooth and sophisticated rea-life social interactions. Psychological and human-centred perspectives are therefore crucial to be incorporated in for better understanding and development of social robots that can be deployed as assistants and companions to enhance human life quality. In this thesis, I present a research approach that draws together psychological literature, Open Science initiatives, and game theory paradigms, aiming to systemically and structurally investigate the cooperative and social aspects of human–robot interactions. In Chapter 1, the three components of this research approach are illustrated, with the main focus on their relevance and value in more rigorously researching human–robot interactions. Chapter 2 to 4 describe the three empirical studies that I adopted this research approach to examine the roles of contextual factors, personal factors, and robotic factors in human–robot interactions. Specifically, findings in Chapter 2 revealed that people’s cooperative decisions in prisoner’s dilemma games played with the embodied Cozmo robot were not influenced by the incentive structures of the games, contrary to the evidence from interpersonal prisoner’s dilemma games, but their decisions demonstrated a reciprocal (tit-for-tat) pattern in response to the robot opponent. In Chapter 3, we verified that this Cozmo robotic platform can displays highly recognisable emotional expressions to people, and people’s affective empathic might be counterintuitively associated with the emotion contagion effects of Cozmo’s emotional displays. Chapter 4 presents a study that examined the effects of Cozmo’s negative emotional displays on shaping people’s cooperative tendencies in prisoner’s dilemma games. We did not find evidence supporting an interaction between the effects of the robots’ emotions and people’s cooperative predispositions, which was inconsistent with our predictions informed by psychological emotion theories. However, exploratory analyses suggested that people who correctly recognised the Cozmo robots’ sad and angry expressions were less cooperative to the robots in games. Throughout the two studies on prisoner’s dilemma games played with the embodied Cozmo robots, we revealed consistent cooperative tendencies by people that cooperative willingness was the highest at the start of games and gradually decreased as more game rounds were played. In Chapter 5, I summarised the current findings and identified some limitations of these studies. Also, I outlined the future directions in relation to these topics, including further investigations into the generalisability of different robotic platforms and incorporating neurocognitive and qualitative methods for in-depth understanding of mechanisms supporting people’s cooperative willingness towards social robots. Social interactions with robots are highly dynamic and complex, which have brought about some unique challenges to robotic designers and researchers in the relevant fields. The thesis provides a point of departure for understanding cooperative willingness towards small-size social robots at a behavioural level. The research approach and empirical findings presented in the thesis could help enhance reproducibility in human–robot interaction research and more importantly, have practical implications of real-life human–robot cooperation