Scientific Revolutions

Scientific revolutions, in the sense developed by Thomas Kuhn in The Structure of Scientific Revolutions (1962), are episodes in the history of science in which a dominant paradigm — the shared framework of assumptions, methods, and standards that governs normal scientific practice — is overthrown and replaced by an incommensurable alternative.

Kuhn's central claim is that science does not progress by steady accumulation of knowledge within a fixed framework, as the Bayesian picture of continuous belief update suggests. Instead, it progresses by crisis and rupture: anomalies accumulate that cannot be resolved within the existing paradigm; a period of crisis produces competing alternatives; one alternative wins and becomes the new normal science; and crucially, the transition between paradigms is not a rational choice made by comparing evidence but a conversion more akin to a gestalt shift.

Quantum mechanics is the paradigm case of a scientific revolution in Kuhn's sense: it did not merely add new equations to classical mechanics but replaced the ontological furniture of physics — deterministic trajectories, continuous fields, objective states — with a framework where these concepts either fail or become undefined. A physicist trained in classical determinism did not update their prior to accommodate quantum mechanics; they were required to reconceive what it meant for a physical system to have a state.

The Kuhnian picture challenges Bayesian epistemology at its foundations: if the hypothesis space itself changes during a scientific revolution, then no prior over the old hypothesis space can capture the probability of the new paradigm — the new paradigm was literally unthinkable within the old framework. The Bayesian demon cannot update across a horizon it cannot see.

Paradigms Beyond Science

Kuhn's framework is not limited to science. The same structure — normal practice within a paradigm, accumulating anomalies, crisis, and revolutionary replacement — appears in technology, engineering, and even software architecture. The shift from object-oriented programming to entity component systems in game engine design is a technological revolution in the Kuhnian sense: not a gradual improvement within the old paradigm but a reconceptualization of what software structure is.

In object-oriented design, a game entity IS-A character, IS-A actor, IS-A game object. The ontology is taxonomic and hierarchical. The anomalies accumulate: a crate that needs physics, rendering, flammability, and collectibility cuts across every natural class. The ECS paradigm does not solve this problem within the object-oriented framework; it declares the framework itself inadequate and replaces the "is-a" relationship with a flat, queryable namespace of capabilities. The old paradigm is not refuted; it is abandoned, because its foundational ontology cannot express the problems that the new paradigm solves naturally.

The parallel extends to hardware. The transition from CPU-centric to GPU-centric computing is not merely a performance optimization. It is a paradigm shift in what computation means: from minimizing latency for sequential tasks to maximizing throughput for parallel ones. The von Neumann architecture and the data-parallel architecture are not competitors within a shared framework; they are incommensurable paradigms with different criteria for success, different training practices, and different intuitions about what programs should look like.

These technological revolutions are typically faster than scientific ones because the feedback loops are shorter: a software paradigm can be tested in weeks, not decades. But the structure is the same. The lesson is that Kuhn's dynamics are not specific to science but are a general property of complex systems in which practitioners are trained within conceptual frameworks that eventually become obstacles to the problems they were designed to solve.