Baruch Fischhoff’s Experiment, 1975
In 1975, a researcher named Baruch Fischhoff gave participants a passage describing the historical situation before Nixon’s 1972 trip to China and asked them to estimate the probability of various outcomes — diplomatic relations established, a Summit meeting, Nixon declaring the trip a failure, and so on. One group made predictions before reading about what actually happened. Another group was told the outcomes and asked to remember what probability they would have assigned before knowing.
The results established a finding that has been replicated hundreds of times since: people systematically overestimate how predictable an outcome was once they know how it turned out. The group told the outcomes believed they would have predicted them with much higher confidence than the group that actually predicted them beforehand. More striking, subjects who were explicitly instructed to ignore the outcome information when reconstructing their prior judgments were largely unable to do so. The knowledge had become structurally integrated into how they thought about the situation.
Fischhoff called this creeping determinism — the sense that what happened had to happen, that it was visible in the situation all along. The more informal term that has stuck in the literature is hindsight bias, or the “I knew it all along” effect. It is, by the standards of cognitive psychology, one of the most robust and least correctable biases documented.
What Hindsight Bias Actually Does
Hindsight bias is not a simple memory distortion in the sense of misremembering specific facts. It operates at the level of narrative reconstruction. When an event occurs, the brain immediately begins integrating it into its existing model of the world — linking it to prior knowledge, prior causes, prior signals. This integration is largely automatic and happens within seconds of learning an outcome.
The problem is that once this integration has occurred, it becomes very difficult to mentally undo it. The causal chain from prior conditions to outcome now feels visible and obvious. Events that were ambiguous before the outcome are retroactively assigned the role of harbingers. The outcome doesn’t just feel predictable — it feels like it was predictable from the evidence available at the time. But the evidence available at the time didn’t unambiguously point anywhere.
Three distinct effects have been identified within what gets broadly called hindsight bias. Memory distortion refers to the misremembering of prior predictions in the direction of the actual outcome. Inevitability judgments refer to the sense that the outcome was bound to happen. Foreseeability judgments refer to the belief that the outcome should have been anticipated. All three operate together in most real-world contexts, and all three are relevant to how people evaluate divination.
The Reading That Was Right All Along
The typical encounter with hindsight bias in divination goes like this. A person receives a reading — a tarot spread, a BaZi analysis, a Nine Star Ki annual reading — in January. The reading describes a period of upheaval in relationships, a professional transition, a need to let something old dissolve before something new can form. The language is rich and reasonably specific within the conventions of the system, but it doesn’t specify what the upheaval will be, which relationship, which transition.
By December, something has happened. A job changed. A relationship shifted. A period of uncertainty resolved. The person looks back at the January reading and the language now maps precisely onto what occurred. The upheaval was the job. The dissolution was the relationship. The reading was accurate.
What makes this feel so compelling is that the match is often genuinely striking. The language of the reading really does describe what happened. The problem is that the language of the reading was broad enough to describe a wide range of possible developments — and that the brain, integrating the December outcome into the January record, has now made the connection feel inevitable. The equally plausible readings of the same January language that didn’t come true have been quietly deemphasized. They were possible interpretations before December. They’re not the interpretation now.
This is not deliberate cherry-picking. It’s hindsight bias operating automatically on narrative material.
Why Divination Language Is Especially Vulnerable
Hindsight bias operates more powerfully when prior information is ambiguous. This is well-established in the experimental literature: the vaguer the prior information, the more the brain can reinterpret it to align with the actual outcome, and the more inevitable the outcome consequently feels.
Divination language is almost always constructed at a level of abstraction that allows for multiple interpretations. This is not accidental — it’s structurally necessary for a system that aims to generate useful reflections for a diverse range of people with a diverse range of circumstances. A tarot card meaning “transformation” or “endings and beginnings” can map onto a move, a job change, a death, the end of a relationship, a creative breakthrough, a religious conversion. A BaZi luck pillar analysis describing “metal pressure on wood” can map onto dozens of different experiential textures.
The abstractness that makes these readings feel applicable at the time of reading is the same property that makes them maximally susceptible to hindsight reinterpretation. After the fact, the abstract language molds itself to the specific event so naturally that the fit feels like precision rather than post-hoc alignment.
This creates a specific evaluative trap. The more ambiguous the reading, the more likely it is to feel accurate after any significant event. But “feels accurate after the fact” is exactly the wrong metric for evaluating whether a system generates useful information before the fact. The two are not equivalent, and hindsight bias ensures that people routinely treat them as if they were.
The Interaction with Memory Distortion
Hindsight bias doesn’t only affect how readings are interpreted after events. It also distorts the memory of what the reading actually said.
Several studies have demonstrated that people systematically misremember prior predictions to align with actual outcomes. In the context of divination, this means that a reading remembered as “predicting” a relationship ending may have originally said something more like “a period of emotional turbulence and reevaluation.” The memory of the reading has been retroactively sharpened in the direction of what happened.
This memory distortion compounds the problem considerably. Not only does the brain make actual outcomes feel inevitable from the original reading — it also modifies the record of what the original reading contained, making the match appear more specific than it was. The person isn’t lying. Their memory is simply doing what memory does under the influence of outcome knowledge.
This has practical implications for any attempt to systematically track reading accuracy. Written records of readings — made at the time, before outcomes are known — show substantially less accuracy when evaluated prospectively than memories of the same readings show when evaluated retrospectively. The prospective evaluation and the retrospective evaluation are measuring different things. Only the prospective one is measuring predictive accuracy. The retrospective one is mostly measuring hindsight bias.
What This Means for the “Accurate Reading” Experience
None of this means that the experience of receiving an accurate reading is worthless or that the reading contained no signal. The phenomenology of “this reading captured something true about my situation” is real and worth taking seriously. The question is what it’s evidence of.
The felt accuracy of a retrospective reading assessment is evidence of several things: that the reading’s language was rich enough to support meaningful interpretation, that the event it was matched to was genuinely significant, and that the brain did its normal work of integrating new experience into existing narrative. It is not straightforward evidence that the divination system generated accurate predictions, because the process of evaluation is contaminated by hindsight bias throughout.
This leaves the honest position in an uncomfortable place. You can’t use your retrospective sense of accuracy to evaluate the system. You can’t fully trust prospective evaluation either, given the other cognitive biases operating on it. The only evaluation methodology that actually isolates predictive accuracy — preregistered prospective tracking with written records, evaluated blind to outcomes — is one almost no divination user employs.
This doesn’t make divination useless. But it means the question “is this system accurate?” cannot be answered by accumulating personal experiences of accurate readings, however numerous and compelling those experiences feel. The experiences are real. Their evidential weight for system accuracy is limited in specific, well-characterized ways.
The Harder Question
Hindsight bias research raises a question that goes beyond divination: if outcome knowledge automatically restructures how we perceive prior evidence, what does it mean to evaluate the accuracy of any prediction-adjacent system?
This problem is not unique to astrology or tarot. It applies to medical prognosis, financial forecasting, geopolitical analysis, and clinical judgment. In all of these domains, Fischhoff and subsequent researchers found the same pattern: professionals and non-professionals alike systematically overestimate how predictable the outcomes were. The doctor remembers the case as showing clearer warning signs than the chart actually recorded. The analyst remembers the geopolitical signals as more obvious than the original briefing contained.
The solution in rigorous fields is structural: preregistration, blinding, prospective tracking, base rate calibration. These methods exist precisely because the brain can’t be trusted to evaluate prediction accuracy without them.
For divination, these methods are almost never employed. The systems are evaluated through accumulated retrospective impressions — which is to say, through exactly the mechanism that hindsight bias exploits most effectively.
What this suggests is not that divination systems are definitely inaccurate. It suggests that we don’t actually know, and that the evidence most people cite for their accuracy — the compelling experience of retrospective fit — is the wrong kind of evidence for the question being asked.
The reading felt right. That’s worth something. It’s just not worth as much as it feels.