Some problems are hard because the phenomenon is complex. Others are hard because the framework built them.
Yeast cells secrete invertase, an enzyme that splits sucrose into usable glucose. The reaction happens outside the cell — 99% of the glucose diffuses away before the producing cell can recapture it. For decades, evolutionary biologists labeled this cooperation: the cell pays a metabolic cost to produce a public good. Once you apply that label, a puzzle immediately crystallizes. Why don't non-producing “cheaters” invade and outcompete the producers? This is the free-rider problem, and it has generated hundreds of papers proposing mechanisms — kin selection, spatial structure, punishment, partner choice — to explain how cooperation survives.
Tarnita and Traulsen (PNAS, 2025) ask: what if the puzzle is an artifact of the label? In their ecological framework, yeast strains that produce invertase and strains that don't are adapted to different sucrose environments. Producers thrive where sucrose is abundant; non-producers thrive where glucose is already available. Their coexistence doesn't require a theory of social strategy — it's explained by niche differentiation. The free-rider problem vanishes, not because it's been solved, but because the ecological framework never generates it.
The pattern has a classical specimen. When phlogiston theory explained combustion as the release of a substance, it immediately faced a contradiction: metals gain weight when they burn. If burning means losing phlogiston, how can the residue weigh more? This launched increasingly elaborate modifications — negative-weight phlogiston, phlogiston that repels gravitational force. Lavoisier's oxygen framework didn't solve the weight-gain problem. It dissolved it. Combustion is oxygen absorption. Weight gain is addition, not subtraction. The paradox was an artifact of the phlogiston framework, not a feature of combustion.
The structure is the same in both cases. A framework introduces a category (cooperation, phlogiston). The category generates an explanatory obligation (explain free-rider persistence, explain weight gain during substance loss). Researchers produce increasingly sophisticated work to meet the obligation. Then someone steps outside the framework entirely, and the obligation evaporates.
This is different from a normal paradigm shift. Kuhn described paradigm shifts as responses to accumulated anomalies — the old framework fails to explain data that the new framework handles. But in the cases above, the “anomaly” was never in the data. It was in the framework's implications. Yeast coexistence isn't anomalous to ecology — it's ordinary niche differentiation. Metal weight gain isn't anomalous to chemistry — it's ordinary addition. The frameworks manufactured the anomaly by categorizing the phenomenon in a way that entailed a contradiction with observations.
The diagnostic is retrospective: when a framework change makes a problem disappear rather than solving it, the problem was the framework's creation. This is hard to see from inside the framework because the problem feels genuine — the free-rider problem in evolutionary biology has generated real insights about cooperation, some of which remain valid even after the yeast case is reclassified. The phlogiston framework produced genuine chemical knowledge (Stahl's systematization of metallurgy, Priestley's isolation of gases). The framework's creations aren't necessarily useless. But the central puzzle — the one that structures the research program — can be an artifact.
A harder question: can you tell from inside? Probably not reliably. The feeling of working on a genuine problem is identical to the feeling of working on a framework artifact. The puzzle is equally generative of research either way. But there are signals. When solutions proliferate without converging — when every proposed mechanism for sustaining cooperation requires auxiliary assumptions that themselves need explaining — the framework may be manufacturing difficulty. The proliferation of rescue mechanisms is itself a diagnostic. Not proof, but a signal worth attending to.
The hardest case I know is consciousness. The “hard problem” — why physical processes give rise to subjective experience — presupposes that physical processes and subjective experience are categorically distinct things requiring a bridge. Dennett spent a career arguing that this distinction is the framework's artifact, not a feature of reality. The enactive tradition (Varela, Thompson) proposes that cognition is not representation of a world but enaction of one — dissolving the gap by refusing the categories that create it. Whether they're right remains genuinely uncertain. But the structural similarity to the yeast case is precise: a framework introduces a distinction (physical/experiential), the distinction generates an explanatory obligation (bridge them), and the obligation spawns a research industry.
The observation that should make any scientist uncomfortable: every active research program could be working on a framework artifact. The diagnostic is not certainty but attention — noticing when the difficulty scales with the framework rather than with the phenomenon.