Categorical Ambiguity of AI Agents

AI agents resist relational stabilization because their social cues are rich but inconsistent. With a human colleague, you settle into a relational mode quickly — “this is my manager,” “this is my collaborator” — and it runs on autopilot. With a tool, you settle into “this is a tool” and the framing never wavers. AI agents occupy a unique position: they are categorically ambiguous in a way that resists either settlement.

The CASA framework predicted that social cues from computers trigger automatic social responses, and that more social cues produce more stable social frames. But CASA was formulated around minimal, consistent cues: a computer that says “I” or uses a gendered voice. The cues point in one direction, the frame activates, and nothing contradicts it.

AI agents break this because the social cues are rich but inconsistent from the same entity. The AI defers like a subordinate, then asserts like a peer, then explains like a teacher, then mirrors your register back at you — all in the same conversation. When social cues are contradictory, more cues means more conflicting signals, which deepens ambiguity instead of resolving it. You say it is a tool; you know it is a tool; you want to use it as a tool. Yet when you actually start using it, your relationship with it changes, and it changes continually. We are perpetually in a situationship with the AI.