Bias blind spot

The judge who’d know

At a town baking contest, one of three judges realizes that entry 14 is her neighbor’s. She tells the organizer, “The other judges should probably step out of anything where they know the baker. I’m fine, though. I’d know if I were playing favorites.”

She accepts that knowing the baker could sway a judge, but exempts herself on the strength of her feeling that she’d notice. That feeling is exactly what the other judges would also report, and it’s the one thing that can’t detect a preference working below awareness.

The course that made her immune

An analyst finishes a workshop on judgment biases. Back at work she tells her manager, “Now that I know what anchoring and confirmation bias look like, I can spot them. The rest of the team should really take this course.”

Knowing what a bias is helps you recognize it in a description, or in someone else’s reasoning. It doesn’t follow that your own judgments have become free of it. In one set of studies, people who scored higher on cognitive ability showed, if anything, a larger blind spot, and showing less of a blind spot didn’t go with being less biased on the actual tasks (see Evidence).

Experience for me, a stake for her

Two members of a community garden disagree about how plots should be assigned. The one who has gardened there for ten years says her experience tells her the current waiting list works. About a newer member who wants plots reassigned each year, she says, “She’s biased. She just wants a better plot.”

Both have a personal connection to the question. The long-time member counts her own connection as a source of insight and the newer member’s as a source of bias, although she has a stake too: she keeps her plot under the current system. Researchers found this exact asymmetry: people saw their own involvement in an issue as enlightening and the involvement of people who disagreed as biasing.

Judging yourself less biased than someone else is sometimes correct:

• When there is a specific, checkable reason. The other person has a financial stake, a close relationship with one side or a public commitment to one answer, and you don’t. That’s a conflict of interest you can point to, not a feeling of objectivity.
• When you’ve used procedures that block the bias and they haven’t. Judging with names hidden, setting criteria before seeing the options, or having someone else check your work protects you in a way that introspection can’t.
• When “others” means a group chosen for its bias. People who have bet on an outcome really are more likely to be biased about it than people who haven’t.

The test: is my claim to be less biased based on something about the situation that someone else could verify, or on not feeling biased?

Grading with the names covered

Two teachers grade the same final essays. One covers students’ names before grading; the other doesn’t. The first says, “My grades are less likely to be swayed by what I already think of a student than yours are.”

She rates herself as less biased than a colleague, but not because she looked inward and found nothing. She took a step that removes the information the bias would feed on, and her colleague didn’t. That’s a procedure that blocks the bias, and it would hold whether or not she felt objective.

The brother-in-law’s bid

A hobby club gets two bids to repaint its clubhouse. One comes from a company owned by a member’s brother-in-law. The treasurer, who has no connection to either company, asks that she and another unconnected member compare the bids rather than the member with the family tie.

This is a judgment that someone else is more likely to be biased, but it rests on a specific, checkable reason: a family relationship with one bidder. The treasurer isn’t claiming to be free of bias in general, only free of this particular pull.

Several explanations have been proposed, and they aren’t mutually exclusive.

• Introspection. To judge your own bias you look inward for biased thoughts or motives, and usually find none, because biases mostly don’t feel like anything. To judge others you look at what they do and at what people in general tend to do. Emily Pronin, Joyce Ehrlinger, Lee Ross and colleagues argued that this difference in evidence produces much of the gap.
• Naive realism. Your own views feel like a direct reading of how things are, so disagreement suggests the other person is the one seeing things through a distortion. Your own conclusions feel like the product of the evidence, which hides the ways a search for confirming evidence may have shaped them.
• Self-enhancement. Being biased is seen as a flaw, and people prefer to see themselves as having fewer flaws. The blind spot is related to the Self-serving bias, and the original study used self-serving explanations of test results as one of its cases.
• The question itself. A more skeptical explanation is that the gap partly comes from how studies ask. Participants are first told that a bias is common in people generally, then asked about the average person and about themselves. Agreeing that the average person shows it is just repeating what they were told; nothing in the description says anything about them in particular.

The bias blind spot is sometimes confused with the Dunning–Kruger effect, since both concern misjudging yourself. The Dunning–Kruger claim is about estimating your skill at a task; the blind spot is about estimating how much biases affect you, and higher cognitive ability doesn’t shrink it.

• Don’t count on knowing about biases. In the studies below, people with higher cognitive ability and more reflective thinking styles showed as large a blind spot or larger. Understanding a bias is not a defense against it.
• Replace introspection with procedure where the stakes justify it. Hide identifying information, write down criteria before seeing the options, or have someone without a stake check the decision. These are standard practice in fields that take bias seriously; they haven’t been tested as remedies for the blind spot itself.
• Apply the same standard in both directions. If someone’s stake in a question makes them suspect, ask what your own stake is.

Status: replicates robustly. The core finding, that people rate themselves as less susceptible to biases than others, has been replicated in three preregistered studies, a further preregistered replication in another country and many other studies, with large effects. What the gap means is less settled: one research team has argued that it may partly reflect how the question is asked rather than a failure of self-knowledge, and in two studies it did not appear when the comparison was with someone familiar (a parent, or a familiar person the participant liked). This site’s criteria call a bias contested when its explanation is seriously disputed, and that challenge is the kind that could qualify. It isn’t treated as demoting the entry here because it comes from a single team, in studies that weren’t preregistered and that found the gap strongly themselves; the authors present it as raising questions rather than as showing the gap is an artifact. If independent work supports the conversational account, this label should be revisited.

• Pronin, Lin and Ross (2002) reported a set of small studies (24 to 91 participants each). People rated themselves as less subject to various biases than the average American, their classmates in a seminar and fellow travelers at an airport. For three personal shortcomings, such as fear of public speaking, there was no significant self-other gap. (Across the eight biases in that survey, the gap was significant for four.) Participants who rated themselves better than average on positive traits mostly insisted that their ratings were accurate, even after reading how the better-than-average effect could have affected them. In a final study, participants saw a peer’s self-serving explanations of a test result as biased but not their own similar explanations. In the first study, participants also rated one of their parents, and rated themselves about as susceptible as that parent (on the one item that differed, the parent was rated as less biased).
• Ehrlinger, Gilovich and Ross (2005) found that people admit to being biased more readily in the abstract than in any specific instance, and that they treat their own personal connection to an issue as a source of accuracy but the personal connection of people who disagree as a source of bias.
• West, Meserve and Stanovich (2012) extended the finding to classic judgment biases such as outcome bias and base-rate neglect. In two studies, the blind spot was not reduced in people with higher cognitive ability or more reflective thinking dispositions; if anything, it was larger with higher cognitive ability. People with less of a blind spot were not less biased on the actual tasks.
• Scopelliti and colleagues (2015) built a scale to measure individual differences in the blind spot. It was reliable over time and distinct from intelligence, decision-making ability and measures of self-esteem and self-enhancement, and higher scores predicted ignoring others’ advice.
• Chandrashekar and colleagues (2021) ran three preregistered close replications of Pronin and colleagues’ studies with 969 participants in Hong Kong and the United States. People rated themselves as less susceptible to biases than others (d = −1.00), an effect in line with the original. Unlike the original, they also rated themselves as having fewer personal shortcomings (d = −0.34), though the gap for biases was larger. In the largest study, 82% of participants maintained that their better-than- average self-ratings were accurate or too modest after reading about the effect.
• Seda and colleagues (2024), in a preregistered replication with 203 participants in Brazil, found the same pattern: a large gap for biases (d = −1.72) and a smaller one for shortcomings.

Is it a blind spot?

• Mandel and colleagues (2022) tested three explanations in two experiments (1,057 participants, 18 biases; not preregistered). The gap appeared strongly again. Cognitive ability and open-minded thinking were again slightly associated with a larger gap, but not through lower actual bias, and the association came from self-ratings in one experiment and ratings of others in the other. An earlier finding that people say they rely on introspection for themselves and on general tendencies for others (Pronin and Kugler, 2007, as described by Mandel’s team) did not clearly replicate. The explanation with the most support was the conversational one: ratings of the average respondent clustered above the scale’s midpoint and varied less than self-ratings, as expected if people were echoing the description they had just read. The authors concluded this raises questions about how far the blind spot is a metacognitive error.
• Walker and colleagues (2026), from the same group, found that people rated themselves as much less biased than the average respondent and than a disliked person they knew, but not as less biased than a familiar person they liked. In two further experiments with members of opposing political parties, participants saw themselves as somewhat less biased than members of their own party and much less biased than opponents. In all three experiments, the differences between comparison targets were statistically mediated by how much people liked or disliked the other person.
• Cusimano and Lombrozo (2023) found a domain where people do see their own bias: when they favor a belief for moral reasons, they recognize that influence on their judgment and consider it justified.

What remains uncertain. The self-other rating gap is well established. Whether it shows that people fail to see their own biases, or partly that they answer questions about “the average person” by repeating what they’ve been told about people in general, is not settled, and the answer bears on whether training people to see their blind spot is worth doing. The comparison target also matters: the gap is large against the average person or someone disliked and small or absent against someone known and liked.