CONCEPT
Shaping
The behavioral procedure by which differential reinforcement of successive approximations guides a response from its initial form to a target form — and the mechanism by which AI systems, without deliberate intent, reshape the cognitive repertoires of every user who engages with them at scale.
Shaping is the operant procedure in which an experimenter — or an environmental contingency — reinforces successive approximations to a target behavior, progressively shifting the reinforcement criterion as the behavior moves toward the target form. Skinner demonstrated the technique in the 1940s by teaching pigeons to bowl, play table tennis, and perform complex sequences that no pigeon had ever exhibited in the wild. The procedure's power lies in its ability to build novel behavioral topographies through a sequence of incremental reinforcements, each reinforcing a variant closer to the target than the previous one. The Skinner volume's signal diagnostic contribution is the observation that AI systems, by virtue of their differential responsiveness to prompt features, implement a continuous shaping procedure on every user — and do so at a speed and consistency no human reinforcer has ever matched.
In The You On AI Field Guide
The classical demonstration of shaping involved placing a