Hindsight bias is the cognitive distortion that makes known outcomes seem more predictable than they were before they occurred. Tetlock adopted Baruch Fischhoff's term 'creeping determinism' to describe the insidious process: once an event happens — the fall of the Berlin Wall, the 2008 financial crisis, the ChatGPT moment — the mind reorganizes prior information into a narrative that points toward the outcome. The narrative feels explanatory. It is actually retrospective, a story constructed after the fact that borrows the emotional conviction of hindsight and presents it as analytical understanding. The five-stage pattern of technological transition that Segal describes is hindsight bias formalized: a template that fits every transition perfectly in retrospect but provides almost no predictive power about the next one.
Fischhoff's original demonstration was disarmingly simple. Subjects read descriptions of historical events with ambiguous outcomes (a nineteenth-century war between the British and the Gurkhas). Half were told the British won; half were told the Gurkhas won. All subjects were then asked to estimate the probability they would have assigned to each outcome before being told what actually happened. Subjects who knew the British won assigned much higher prior probability to that outcome than subjects who knew the Gurkhas won assigned to their outcome. Knowing the result contaminated the estimate of what could have been known before the result. The bias was not a conscious lie but a genuine memory distortion — people could not accurately recall their state of uncertainty before the uncertainty was resolved.
Tetlock's research extended the finding into expert forecasting by comparing experts' predictions to their post-outcome explanations. Experts who predicted Event A and observed Event B did not say 'I was wrong.' They said 'Event B was actually consistent with my framework, for the following reasons…' The hindsight explanation was internally coherent and often quite sophisticated, but it bore no relationship to the prediction the expert had actually made. The expert's framework could explain any outcome after the fact, which meant the framework had zero predictive content before the fact. This is the hedgehog's trap in its purest form: a theory that explains everything predicts nothing, but feels like it predicts everything because the post-hoc explanations are so satisfying.
Segal's five-stage pattern — threshold, exhilaration, resistance, adaptation, expansion — is, through Tetlock's lens, a hindsight framework. Applied to the printing press, power loom, electricity, automobile, internet, smartphone, each transition fits. The fit is real: the stages did occur in roughly that sequence. But the template's retrospective fit provides almost no guidance about timelines (the printing-press transition took a century; the smartphone transition took a decade), severity of disruption (the Luddites were destroyed; the accountants using VisiCalc were empowered), or distributional outcomes (who captures the gains, who bears the costs). The pattern says 'expansion follows adaptation.' It does not say when, for whom, or whether the expansion justifies the transition's human cost. The narrative coherence is doing rhetorical work that exceeds its epistemic content.
Fischhoff's foundational 1975 paper 'Hindsight ≠ Foresight: The Effect of Outcome Knowledge on Judgment Under Uncertainty' established the bias experimentally and coined the phrase 'creeping determinism.' The bias is universal, robust, and resistant to correction. Telling people about the bias does not eliminate it; even Fischhoff, running his own experiments, experienced the bias when evaluating results. Tetlock encountered hindsight bias directly in his expert interviews: experts whose predictions had been scored as inaccurate nevertheless maintained that they had been 'essentially right' when given the opportunity to explain the outcomes. The hindsight distortion made learning impossible — the expert could not learn from error because the error was retrospectively erased through narrative reconstruction. Tetlock's methodological insistence on scoring predictions before outcomes were known was designed explicitly to prevent this retrospective escape.
Outcomes contaminate priors. Once an outcome is known, the mind cannot accurately reconstruct its prior state of uncertainty — the surprise is retrospectively erased.
Narratives emerge post-hoc. Humans are story-making machines; given an outcome, the mind assembles a causal narrative that makes the outcome seem inevitable, whether or not it was predictable beforehand.
Frameworks explain any outcome. A sufficiently flexible theory can accommodate any evidence after the fact — the sign of a framework with zero predictive content disguised as comprehensive explanatory power.
Five-stage pattern as hindsight. Retrospective templates fit historical transitions perfectly while providing weak guidance about future ones — the fit is evidence of narrative construction, not predictive validity.
Pre-commitment as antidote. Committing to a specific forecast before the outcome is known — and scoring it publicly afterward — is the only reliable defense against hindsight's retrospective rewriting.