Effects of action instruction source on the sense of agency: humans vs. artificial agents

The sense of agency (SoA)-the experience of being in control of one's own actions and outcomes-is a fundamental aspect of daily life. Prior research shows that SoA can be disturbed when actions are externally instructed rather than voluntarily initiated; however, the role of the instructing agent specifically in shaping this effect remains underexplored. As artificial agents (e.g., online chatbots, virtual avatars) become increasingly embedded in everyday interactions, their potential to influence human action raises important questions about the experience of agency. Across three studies, we investigate how action instructions delivered by a human or an artificial agent (in the form of an on-screen embodied chatbot) influence both implicit (temporal binding) and explicit (self-reported control) measures of agency. The results of Study 1 (implicit) showed that binding was strongest in the free choice condition wherein participants' actions were of their own volition compared to actions conducted under external instruction. Notably, binding was significantly reduced when actions were directed by a chatbot compared to the free choice condition. Similarly, the results of Study 2 (explicit) showed that self-reported control ratings were the highest in the free choice condition and decreased significantly when comparing the free choice condition with both instruction conditions. After conducting a third follow-up study that integrates both implicit and explicit methods, we were able to replicate the findings of Study 1 and 2. These results highlight a distinction in the experience of agency when responding to human- versus technology-driven instructions.