Adjacent Complementarity

In The You On AI Field Guide

The mechanism explains why the Berkeley study found workers unable to stop even when continuation was visibly harming them. Each building session was not just enjoyable — it was recalibrating the baseline. The pre-AI rhythm of meetings, documentation, and slow collaborative iteration began to feel intolerably slow, not because it had become objectively worse but because the complementarity had shifted the reference point against which the slower rhythm was evaluated.

Adjacent complementarity also explains the characteristic instability of productive compulsion — the binge-crash pattern the Berkeley researchers documented. Systems governed by adjacent complementarity with a high discount rate and a delayed cost function tend toward unstable equilibria. Small perturbations produce large swings. Periods of extreme intensity are followed by periods of collapse, because the same feedback loop that drives escalation drives the crash when the cost threshold is finally breached.

The mechanism has a feature that makes productive addiction structurally harder to address than substance addiction: the adjacent complementarity in productive work shifts the agent's preference structure itself. Revealed preference — the economic principle that individuals reveal their true preferences through choices — becomes a treacherous guide when the activity in question alters the very preferences being revealed.

Origin

Becker and Murphy developed the concept to formalize how consumption goods can generate self-reinforcing demand patterns that classical consumer theory could not explain. The concept extended earlier work on habit formation while introducing the forward-looking optimization that distinguished rational addiction from mere habit. The formalization allowed empirical testing — and the tests confirmed that consumers of addictive goods respond to expected future prices in ways that require the complementarity assumption.

Key Ideas

Self-reinforcement through recalibration. Current consumption does not just produce satisfaction — it shifts the reference point against which future consumption is evaluated.

The optimization trap. The agent is not failing to calculate. The agent is calculating correctly on a path the calculation itself has reshaped.

Instability. Complementarity-driven systems produce unstable equilibria with characteristic binge-crash dynamics.

Preference-shifting. Revealed preference becomes an unreliable guide when the consumption alters the preference structure it purports to reveal.