CONCEPT
Triple Contingency (Skinner)
The combined reinforcement architecture of AI engagement — positive reinforcement for continuing, negative reinforcement for resuming, punishment for stopping — that the Skinner volume identifies as the mechanism behind the specific difficulty of disengagement.
The Skinner volume's triple contingency is not to be confused with the three-term contingency — it is a different analytical construct at a higher level, identifying the three concurrent reinforcement relationships that jointly maintain AI engagement: (1) positive reinforcement of continuing (each prompt produces a useful response that strengthens prompting); (2) negative reinforcement of resuming (the aversive state of incomplete work is removed when the user returns to the interaction); and (3) punishment of stopping (the continuous reinforcement the system provides is withdrawn when the user disengages, and the withdrawal is experienced as loss). The three contingencies operate simultaneously and produce maintenance effects that exceed what any single schedule would generate alone. Escaping the trap requires overcoming all three contingencies simultaneously — which is why the behavior is so difficult to terminate and why single-pronged interventions consistently fail.
In The You On AI Field Guide
The first contingency — positive reinforcement of continuing — is the familiar schedule analysis: