Behaviorism

In the AI Story

The historical trajectory of behaviorism runs through three major phases. The Watsonian phase (1913–1930) established the methodological commitment to publicly observable behavior and environmental explanation. The neo-behaviorist phase (1930–1960) — represented by Hull, Tolman, Guthrie, and others — attempted to build formal theoretical systems that retained behaviorist methodology while incorporating intervening variables and hypothetical constructs. The Skinnerian phase (1938–present) rejected intervening variables in favor of strict contingency analysis and extended the framework from laboratory research to applied and cultural domains.

The cognitive revolution beginning in the late 1950s — represented by Miller, Chomsky, Newell, Simon, and others — mounted a sustained critique of behaviorism that academic psychology largely accepted. The critique held that behaviorist principles were inadequate to explain complex symbolic behavior, particularly language, and that psychology required theoretical constructs about internal representation that behaviorism excluded. The displacement was nearly total in academic psychology departments by the 1970s.

The tradition continued in applied domains throughout the cognitive ascendancy — applied behavior analysis, behavior therapy, organizational behavior management, educational technology — and produced empirically validated interventions that the cognitive tradition did not match in applied reliability. The return to theoretical prominence through reinforcement learning in AI has been unexpected and substantive: the computational systems that produce human-like language were built on principles Skinner's framework specified, not on the symbolic principles of classical cognitive science.

The Skinner volume treats this history diagnostically. The AI moment is not merely a vindication of behaviorism against the cognitive revolution; it is an occasion for applying the framework to a phenomenon the framework is unexpectedly suited to analyze, and for recognizing that the behavioral consequences of AI-assisted work are predictable from principles the academic psychology of the cognitive era largely set aside.

Origin

John B. Watson, "Psychology as the Behaviorist Views It," Psychological Review 20: 158–177 (1913). The founding manifesto of behaviorism, which developed through multiple subsequent phases including Skinner's radical behaviorism.

Key Ideas

Behavior is the subject matter. Psychology, on the behaviorist view, is a science of behavior, not of mental states inferred from behavior.

Environmental contingencies are the explanatory mechanism. Behavior is accounted for by its relationship to antecedent stimuli and consequent events.

Radical behaviorism includes private events. Thoughts and feelings are themselves behavior, not hidden causes of behavior.

The tradition returned through AI. Reinforcement learning implements operant principles in computational systems, producing renewed theoretical relevance.

Debates & Critiques

The enduring debate between behaviorism and cognitive science has not been resolved, only relocated. The cognitive position that complex symbolic behavior requires internal representation remains widely held; the behaviorist position that contingency analysis is sufficient and that internal representations are themselves behavior also remains defensible. The AI moment complicates both positions: large language models produce sophisticated symbolic behavior through reinforcement principles without anything resembling classical symbolic representation, suggesting that the binary that structured the twentieth-century debate may have been inadequate to both positions.