Broken Reciprocity as Fatigue Accelerant in Human-AI Interaction

Human-AI interaction is genuinely social and effortful, but AI provides no emotional reciprocity — so the cognitive depletion from each relational switch accumulates without attenuation. Banks (2026, “Ghosting the Machine”) argues that human-agent interaction is genuinely social, not parasocial, which means the depletion from emotional labor is real, not performative.

In human interaction, social reciprocity partially replenishes cognitive resources between role switches. A thank-you, visible understanding, a smile — these recovery mechanisms are built into the interaction. This is why you can switch relational modes all day with colleagues and still function: each switch costs, but the reciprocity loop partially offsets the depletion between switches.

AI returns well-formed output and functional feedback, but not emotional reciprocity. The loop is structurally broken. You’re paying the performance cost at every switch with no recovery mechanism between switches. The switching costs from relational-mode-oscillation accumulate without attenuation.