Operant Conditioning
Learning through consequences: reinforcement increases behavior and punishment decreases it, as studied by B. F. Skinner.
Operant conditioning is learning in which behavior is modified by its consequences. The framework was developed by Edward Thorndike (law of effect, 1898) and B. F. Skinner (1930s–1950s). While classical conditioning pairs stimuli with reflexes, operant conditioning deals with voluntary behavior: an organism emits a behavior, the environment delivers a consequence, and the probability of the behavior changes accordingly. Skinner studied this in operant chambers (Skinner boxes), where animals pressed levers or pecked keys for food.
The core distinction is between reinforcement and punishment, each with two forms. Positive reinforcement adds a pleasant stimulus after a behavior (food for a lever press), increasing the behavior; negative reinforcement removes an aversive stimulus (stopping a shock), also increasing it — negative reinforcement is not punishment. Positive punishment adds an aversive stimulus (a scold), decreasing the behavior; negative punishment removes something pleasant (a fine, loss of privileges), also decreasing it. The two-by-two table — add/remove × pleasant/aversive — organizes all cases, and reinforcement is generally more effective than punishment at shaping behavior.
Schedules of reinforcement determine how consequences are delivered and powerfully affect response patterns. Continuous reinforcement (every response rewarded) produces fast learning but rapid extinction; partial schedules are more resistant to extinction. Fixed ratio (reward every N responses) yields high steady rates; variable ratio (reward after unpredictable counts — slot machines) yields high, persistent rates; fixed interval produces a scalloped pattern with a burst near the expected time; variable interval produces slow, steady responding. These schedules are a classic result of the experimental analysis of behavior.
Operant principles appear throughout human life: wages, grades, praise, fines, and points systems are all reinforcement arrangements. Applied behavior analysis uses them systematically in education and autism therapy, and token economies manage behavior in institutions. The same logic powers reinforcement learning in artificial intelligence — an agent's actions are shaped by rewards and punishments — making Skinner's ideas, stripped of their behaviorist philosophy, the conceptual ancestor of modern AI training.
Tags
behavior learning psychology reinforcement