The mind doesn't like loose facts. Give it a handful of events and it will stitch them into a tidy story, whether or not that story is true. Spotting the stitching is the whole skill.
The narrative fallacy is the habit of taking a string of facts and compressing them into a single, coherent, cause-and-effect story — then mistaking the story's neatness for evidence that it's correct. The clearer and more satisfying the explanation feels, the less likely anyone stops to check it against what was actually knowable at the time.
Named and popularized by Nassim Nicholas Taleb in The Black Swan (2007), building on decades of hindsight-bias research, including Baruch Fischhoff's 1975 studies showing people reliably overestimate how predictable past events were.
A good story is sticky. Once you have one, you stop generating alternative explanations, stop looking for disconfirming facts, and start treating a plausible account as a verified cause.
Market commentary, biographies of successful founders, after-action reports, performance reviews, history textbooks, and any time someone says "and that's when I realized…" about a decision that was really just lucky timing.
Ask: could I have predicted this outcome in advance, using only what was known then — or am I just explaining it well after the fact?
Narrative Fallacy lives in the story-driven, backward-looking corner: it's specifically about the explanations we build after we already know how things turned out. Nearby models share the neighborhood for different reasons — some are about the same backward-looking distortion, some are about forward-looking blind spots, and some are about data problems rather than story problems.
Close but different: Hindsight Bias is the feeling that an outcome was predictable; Narrative Fallacy is the story you build to make it feel that way. Survivorship Bias is about which examples you even see; Narrative Fallacy is about how you explain the ones you do.
The real test: if you can't state the explanation as a prediction you'd have made before the outcome, you're probably looking at a narrative built backward, not a cause found forward.
Show people a handful of unrelated facts and, within seconds, they'll connect them into a plot. This isn't a flaw that shows up occasionally — it's how comprehension normally works. The problem is that the same machinery runs identically whether the facts actually belong together or not.
A narrative replaces a pile of loose, effortful-to-hold facts with one compact, low-effort story. That compression is useful for memory and communication. It is not evidence that the story is the right one.
Three forces push toward narrative over nuance: coherence feels like truth (a story that hangs together is judged as more likely to be accurate, even though coherence and accuracy are different properties); raw uncertainty is uncomfortable to hold, so a single confident cause relieves tension; and stories are what get remembered, repeated, and rewarded, so they get selected for even when they're wrong.
A stock index drops 1.4% in a single afternoon. By evening, four different outlets have each published a confident, different explanation. What does the narrative fallacy predict is happening?
A few tells show up again and again: the explanation arrives after the outcome is already known; it names a single hero or villain for something that had many contributing causes; it uses certainty words like "clearly" or "inevitably"; and it survives by cherry-picking the details that fit while quietly dropping the ones that don't.
Biographies of successful founders are a reliable source of this pattern: the same decisions that look "visionary" in hindsight would have looked like one gamble among many equally plausible gambles at the time they were made.
Before accepting an explanation, ask whether it could have been stated as a prediction beforehand, using only the information available then. If the explanation only works once you already know the ending, it's narrative, not cause. Writing predictions down before outcomes resolve — a practice sometimes called a prediction journal — makes this check almost automatic, because you can't quietly rewrite a prediction you already committed to paper.
A product team ships a feature that becomes unexpectedly popular. In the retrospective, the team writes that they "always knew" the feature would work. What's the single best next question to ask them?
The goal isn't to become someone who refuses to explain anything — stories are how humans coordinate and learn. The goal is to hold explanations as working hypotheses rather than settled facts, especially the satisfying ones, and to reserve real confidence for causes that were stated and tested before the result was known.
This simulator mimics how financial headlines get written. Generate a random day. Notice that the confident tone of the headline never seems to depend on how big — or how explainable — the actual move really was.
Even a barely-there move tends to get a confidently worded headline. The story-generating instinct doesn't check whether there's enough signal to justify the certainty — that check has to come from you.
Fill in the fields below and this page will assemble a ready-to-use prompt you can hand to an AI assistant, or simply ask yourself.
Preview the model as a portable skill below, copy it, download it as a Markdown file, or generate a version tailored to a specific problem you're facing right now.