Variable reward
Also known as: Intermittent reinforcement, Variable-ratio reinforcement
Rewards delivered unpredictably are far more compelling and habit-forming than predictable ones.
What it means
A variable reward is a reinforcer delivered on an unpredictable schedule, so that the learner cannot know which action will pay off or when, and this uncertainty makes the resulting behavior unusually frequent, persistent, and resistant to extinction. In Skinner's analysis of reinforcement schedules, variable-ratio schedules—reward after an unpredictable number of responses—produce the highest and steadiest rates of responding, far exceeding what fixed, predictable rewards elicit. The power lies in anticipation: because any given attempt might be the one that delivers, the gap before reward is filled with a dopamine-driven seeking response, and the variability itself becomes engaging. This is the engine of slot machines and the documented basis of much 'persuasive design' in apps—pull-to-refresh feeds, unpredictable likes and notifications, and loot boxes—where intermittent, uncertain payoffs hook attention and form compulsive habits. The same mechanism that makes gambling addictive makes infinite-scroll feeds hard to put down. It matters because it is one of the most exploited principles in technology and a key reason behaviors become compulsive rather than merely frequent.
Examples
A slot machine pays out unpredictably, and a phone delivers likes and messages at random intervals—both keep people pulling the lever or refreshing far longer than any fixed reward would.
A charity shop rarely has anything good, but the occasional cashmere coat for three pounds keeps regulars coming back weekly. The unpredictability of the find, not the average haul, drives the visits.
A dog given a treat for every sit stops sitting the moment the treats run out. One rewarded unpredictably keeps sitting for weeks, because this attempt might be the one that pays.
First described in B. F. Skinner (schedules of reinforcement); applied by Eyal and others.