Conceptualizing Agency: A Framework for Human-AI Interaction

This paper proposes a framework for understanding intentional agency in the context of increasingly autonomous AI systems. As AI tools perform complex tasks independently, they challenge our traditional notions of agency-particularly the ability to act intentionally, grounded in what Anscombe calls "practical knowledge." We explore intentional agency through three lenses: (1) Anscombe's concept of intention, where we examine how AI tools complicate human intentionality and our ability to give reasons for actions; (2) shared intention, which frames agency in joint human activities; and (3) relational intention, a novel synthesis drawing on Anscombe and Fiske's Relational Models Theory which captures how relationships shift between different relational models. Our analysis reveals that agency is not merely an individual capacity but is fundamentally shaped by relationships with others who can interpret and respond to our reasons. AI systems, however, disrupt this dynamic: while they assist in tasks, they lack the intersubjectivity required to engage in mutual reason-giving, potentially diminishing our practical knowledge and intentional agency. We conclude that AI should be designed to support human-human relationships rather than supplant them, with humans retaining relational authority in human-AI interactions to preserve agency.