Joint Human-AI decision making and sense of agency
Abstract
In this study, we address shared Human-AI decision making through the lens of agency: what is the feeling of control experienced by an operator when making a decision that is shared with an AI-based artificial agent? Using an optimization trajectory task, we explored how the level of autonomy and the explainability of the decision support system impact the human operator's feelings of agency, responsibility and confidence. Results indicated that the level of automation modulate the sense of agency, but also the confidence and feeling of responsibility. More interestingly, our results also indicated that explanations could increase the number of mind change, but also the feelings of agency, confidence and responsibility during such mind change. These results are discussed in term of mechanisms underlying the development of a sense of agency and aims to provide tools and models to explore the notion of 'meaningful control' during our interactions with AI-enabled systems.