Cognitive biases
Hindsight bias: why everything seems obvious after the fact
Quiz night at a friend’s place. The question is: “How many keys does a standard piano have?” Dina thinks and writes on her slip of paper: 60. The right answer appears on the screen: 88. A minute later Dina says: “I knew it was around ninety!” And she isn’t pretending. It truly seems that way to her.
This is hindsight bias, also called the “I knew it all along” effect. When the outcome is already known, it seems it could have been predicted. And the memory of our own earlier guess quietly shifts toward what we know now.
Below, we look at what research has shown, why memory is so easy to rewrite, and how to learn from the past more honestly. In the middle of the article there is a small memory game: now it is your turn to answer.
Contents
In short: how hindsight bias works
What hindsight bias is
The classic experiments were run by the American psychologist Baruch Fischhoff. In a 1975 paper, people read a description of an event that could have had several outcomes. Some were told how it ended, others were not. Then everyone was asked to rate how likely each outcome had been, as if they had known nothing in advance. Those who were told the outcome considered it more likely and barely noticed this influence. So they overestimated what they would have known without the hint. Fischhoff concluded: if we do not notice such influence, it is hard to judge the past honestly and learn from it.
Fischhoff ran a second experiment together with Ruth Beyth (the paper came out in 1975). In 1972, U.S. President Richard Nixon was getting ready to travel to China and the USSR. Students were asked in advance to rate how likely various outcomes of those trips were. After the trips were over, the participants were unexpectedly asked to recall their ratings. If an event had actually happened, people more often “remembered” that they had considered it more likely than they really had. The title of the paper speaks for itself: “I knew it would happen.”
Since then the effect has been tested many times. In a review of 95 studies it was found consistently, though on average it was small to moderate. It showed up more strongly in guessing questions about facts, like “how many meters tall is this mountain,” than in questions about real events. A large replication in 2021 confirmed Fischhoff’s classic result. But in a check in 2025, which has not yet been peer-reviewed in a journal, the effect turned out to be very weak. So it is more honest to say: the effect exists, but its size depends on exactly how it is measured.
Try it yourself: guess and recall
Now you are in Dina’s place. The game takes a couple of minutes. Here is what will happen.
Step 1. We ask five questions about the world, for example about a skyscraper and about Mars. You do not have to know for sure. Move the slider and set your guess.
Step 2. We show the right answers one at a time. Your answers will not be on this screen.
Step 3. There will be one more small task. You will find out which one in the game.
Why are the questions so odd? We need facts that do not change and that few people know exactly. Then you have to estimate, and there is something to recall later. If we asked how many days are in a week, there would be nothing to recall.
What usually happens. In studies with such questions, people who learned the right answer recalled their guess as closer to it than it really was. This was shown, for example, by Ulrich Hoffrage and colleagues. An important detail: the task was to recall your earlier answer, not to guess the right one.
If your memory did not shift, that is a good result, not a failure of the experiment. Only a couple of minutes passed between the answer and the recall, so your answers are still fresh. Five questions are few: one shift toward the right answer could be chance. But unlike many other experiments, here you can see the effect in yourself. Look at where your memory shifted more often: toward the right answer or away from it. If it is about equal, it looks like ordinary forgetfulness. If it is mostly toward the right answer, it looks like hindsight bias.
Why memory shifts toward the answer
Picture a document where edits are made with no change history. The text got more correct, but how it looked yesterday can no longer be seen. This is an image, not the exact way memory works, but it conveys the main thing: new knowledge does not lie next to the old; it partly replaces it.
Knowledge gets updated. Hoffrage and colleagues proposed this explanation. When you learn the right answer, your knowledge on the topic gets updated. If you later need to recall your old guess and you did not memorize it exactly, we reconstruct it from what we know now. And it comes out closer to the right answer. The authors see this as a side effect of a useful process: it is good that we update our knowledge. The only bad part is that we do not notice how this changes our memories. When participants were helped to recall what they had known before the hint, the effect weakened.
The story falls into one line. The psychologists Neal Roese and Kathleen Vohs, in a 2012 review, write that, knowing the outcome, we pick from memory what agrees with it and build a clear story. Such a story is easy to retell, and this ease quietly turns into the feeling that “it was predictable.”
We want the world to make sense. According to the same authors, the effect has a motivation too: we like to think that the world is orderly and predictable. And nobody wants to be blamed for failing to see it coming.
Where it gets in the way
Learning from experience. After a failure it seems the cause was obvious: “I should have realized right away that it would turn out this way.” Roese and Vohs write that hindsight bias narrows our view of the past to a single cause and feeds overconfidence in our own judgments. If “everything was clear,” there is nothing to examine. Yet the most useful lessons often hide in what was unclear back then.
Judging other people’s decisions. When the outcome is bad, the decision looks worse in retrospect. In one study, ordinary people read the same story about a psychotherapist working with a patient who might become dangerous. Only the ending differed. Those who were told that the patient committed an act of violence more often considered that the therapist had acted carelessly and that the violence could have been foreseen. The therapist’s decision was the same in all versions. Something similar happens at work: a manager reviews a failed project knowing how it ended and easily says “anyone could have guessed.”
Relationships and blaming yourself. “I told you so” after a quarrel or a bad purchase sounds convincing, though nobody said anything in advance. There is a quiet side to the same effect. Here we rely not on research but on everyday observation: after a breakup or a loss, it often seems that all the signs were visible from the very beginning, and a person blames themselves for “not noticing.” But back then, without knowing the outcome, those signs looked different and were lost among others. It is fair to judge yourself by what you knew then.
News, sports, forecasts. After an election, a match, or a sudden jump in prices, many people appear who “predicted it all.” There is only one way to check: look at what they said before the event. Hindsight bias goes well with confirmation bias: knowing the outcome, we recall the arguments in its favor and forget the rest.
How to learn from experience more honestly
Hindsight bias is one of the biases on the cognitive bias map, in the section on memory. If you are curious about how we judge our own knowledge, read about the Dunning-Kruger effect.
Frequently asked questions
Is this the same as “I told you so”?
Almost. “I told you so” is one side of the effect, the one tied to memory: the earlier opinion is remembered as closer to what happened. But hindsight bias is broader. It also includes the feeling that the outcome was inevitable and the certainty that it could have been predicted.
If you know about the effect, does it stop working?
Unfortunately, not really. Fischhoff’s participants did not notice how knowing the outcome affected their ratings. In the review of 95 studies, attempts to weaken the effect on average gave no noticeable result. In individual experiments specific techniques helped, for example thinking through two other outcomes, but there is no magic cure. The most reliable thing is to write the forecast down in advance: then you compare the outcome with a text, not with a memory.
Is there an everyday phrase for this?
Yes: “hindsight is 20/20,” meaning that looking back, everything seems clear. People also say “I knew it all along.”
