The influence of human agency beliefs on the perception of gaze-signalled communicative intent

Communication with artificial agents (e.g., animated characters in virtual reality, social robots) is becoming increasingly common. Understanding how to design these agents' behaviours to best support social interactions is therefore essential. One key communicative behaviour that supports human-human interactions is eye gaze, which can signal an agent's focus of attention, goals, and intentions to interact. Previous research has demonstrated that beliefs about whether an agent is human- or artificially controlled influence the perceived significance of gaze at subjective, behavioural, and neurophysiological levels. The present study specifically examined how these beliefs shape the evaluation of gaze dynamics as signals of an agent's communicative intent---an essential process for recognising and responding to communication opportunities. Participants (n = 160) completed a semi-interactive online task with a virtual agent that displayed dynamic gaze sequences. In each trial, we manipulated whether eye contact was established and whether the agent looked at the same object twice. Participants decided whether the agent was communicating a request for assistance to access an object or privately inspecting it. Across two task blocks, we also manipulated whether participants believed the agent was human- or AI-controlled. We found that believing the agent to be human-controlled increased perceived communicativeness when gaze cues were ambiguous, but this effect was attenuated in the second block of the experiment. This pattern may reflect a learned adjustment specific to the experimental setup that involved relatively consistent agent behaviour. Nevertheless, subjective ratings indicated a preference for human- over AI-modelled data. Our findings provide insights into the complexities of gaze-based communication and highlight the role of user expectations in shaping interactions with artificial agents.