The 30-second version
Illusory superiority (also called the better-than-average effect, the above-average effect, or overplacement) is the tendency for most people in a group to rate themselves as better than their peers on positive traits and skills — a pattern that cannot be true in aggregate, since only about half of any group can genuinely be above its own average. The best-known demonstration is Ola Svenson's 1981 study of drivers: 88% of his American sample rated themselves safer than the median driver, and 93% rated themselves more skillful; a 2018 preregistered replication with over 1,200 participants found the same pattern at similar or larger scale. The effect shows up across intelligence, ethics, health habits, and job performance, and it is strongest in subjective, hard-to-score domains and weakest where feedback is immediate and objective, such as a timed race or a graded exam. The practical reading isn't "people are arrogant" — it's that self-assessment without a specific, shared standard to compare against is structurally unreliable, for reasons that apply to careful, honest people as much as anyone else.
Mental Model Compass
Where Illusory Superiority sits next to three other models already in this collection about misjudging ourselves. Click a dot to see why it sits where it does.
Click or focus a dot to read why it sits where it does.
Build the idea, one step at a time
Two different stories feed this model, and it helps to keep them separate. The psychological research traces to studies like Ola Svenson's 1981 paper "Are We All Less Risky and More Skillful Than Our Fellow Drivers?," which asked American and Swedish drivers to rate their own safety and skill against other people on the road. The popular name "Lake Wobegon effect" has a separate, later origin: it was coined around 1988 after West Virginia physician John Jacob Cannell found that every single U.S. state reported its elementary students scoring above the national average on major standardized tests — a statistically impossible result, since a national average cannot be exceeded by everyone being averaged into it. Commentators named this after Garrison Keillor's fictional Minnesota town in his 1985 book "Lake Wobegon Days," where, in his telling, "all the women are strong, all the men are good-looking, and all the children are above average." The two strands describe the same underlying statistical impossibility and are now used almost interchangeably, even though one comes from psychology research and the other from an education-testing scandal.
Svenson's original 1981 study asked 81 American and 80 Swedish college students to compare their own driving safety and skill to the other people in the room. Among the Americans, 88% placed themselves above the median on safety and 93% above the median on skill; among the Swedes, the numbers were lower but still far past the mathematical ceiling of 50% — 77% on safety and 69% on skill. A large, preregistered replication published decades later, with over 1,200 participants and several different rating scales, found the same lopsided pattern: as many as 93% of respondents still placed themselves above the median on at least one scale. The replication's authors note real differences in method from the original study — an online sample instead of an in-person one, and a different comparison group — so the exact percentages shift somewhat by scale and sample, but the core finding, a large majority placing themselves above a line only half of them can actually be above, has held up.
The effect isn't constant everywhere — it's strongest in vague, subjective, hard-to-score domains, like "ethics" or "being a good listener," and weakest where feedback is fast and objective, like a timed 100-meter race or a graded multiple-choice exam. Three mechanisms are commonly proposed for why it happens even to careful, honest people: you have privileged access to your own good intentions and circumstances, but mostly see other people's visible behavior, not their intentions; you unconsciously give more weight to whichever specific criteria you happen to be good at, within a broad trait like "skill" or "leadership"; and the comparison target, "the average person," is vague enough that each person is free to define it in a way that flatters them. None of this requires dishonesty or arrogance — it requires only an ambiguous standard and an evaluator with a stake in the answer.
Research on debiasing this effect has found that merely informing people the bias exists, or even paying them for accuracy, does little to shrink it — the vague comparison standard is the real problem, not a lack of motivation to be accurate. What has been shown to help is replacing the vague standard with a concrete one: comparing yourself to specific, named individuals instead of an abstract "average person," and defining the criteria for a judgment in observable, shared terms before anyone rates themselves against it. It's also worth knowing the failure runs in both directions — overcorrecting into chronic self-doubt or false modesty is its own cost, since it can lead people to pass up opportunities, negotiations, or promotions they were genuinely qualified for.
What it looks like in different fields
Self-rating surveys at performance-review time routinely show most employees placing themselves in the top third or higher, which is exactly the pattern that fuels resentment when only a fraction of them receive the top rating, raise, or promotion. The fix organizations have found most effective isn't a lecture about bias — it's publishing specific, observable criteria for each rating level in advance, so a self-rating can be checked against a shared standard rather than a vague sense of "I work hard."
Candidates asked to self-rate their own skills on a resume or in an interview ("rate your proficiency in this tool from 1 to 10") are subject to the same ceiling-defying pattern as Svenson's drivers. Structured interviews that score candidates against the same specific, predefined rubric for every applicant sidestep the problem better than open-ended self-assessment questions do, because the comparison standard is fixed instead of left for each candidate to define favorably.
Teammates often disagree about who contributed the most to a shared project, not necessarily because anyone is lying, but because each person is quietly using a different personal yardstick — one weighting hours logged, another weighting the hardest technical problem solved — and each can rate themselves "best" by their own measure. Naming the specific, agreed metric before the work starts, rather than after, heads off the disagreement.
The original domain: most drivers think they're safer and more skillful than the person next to them in traffic, most people rate their own sense of humor or fairness as above average, and most couples in surveys report doing more than their fair share of the household chores than their partner credits them with — all mathematically impossible in aggregate, and all evidence of the same ambiguous-standard problem rather than anyone being unusually dishonest.
Worked application: running a calibrated review cycle
- Before anyone self-rates, publish the specific, observable criteria for each rating level. "Above expectations" should name the behaviors or results that earn it, not be left for each person to interpret in their own favor.
- Ask people to compare themselves to named peers or a concrete example of work, not to "the average person on the team." A vague comparison target is exactly what lets a flattering self-definition slip in unnoticed.
- Collect multiple perspectives on the same fixed criteria — manager, peers, and self — rather than relying on self-rating alone. The gap between self-rating and others' ratings, measured against the same standard, is itself useful information.
- Share real calibration examples: concrete work samples at each rating level. Concrete anchors reduce overplacement far more reliably than reminding people the bias exists.
- Separate the conversation about standards from the conversation about any one person's score. Agreeing on the yardstick before scores are revealed avoids the standard itself being renegotiated defensively after the fact.
- Watch for the opposite failure too. Some people underrate themselves against the same vague standard; a calibrated process should catch both overplacement and undue self-doubt, not just the one that's easier to notice.
Common mistakes
Assuming awareness alone fixes it
Telling people the bias exists, or even paying them to be accurate, has been shown to do little on its own. The vague comparison standard is the actual mechanism, and only replacing it with something concrete reliably helps.
Confusing it with the Dunning-Kruger Effect
Illusory superiority is a broad, population-wide pattern found at every skill level. The Dunning-Kruger Effect is a narrower, specific claim: that overestimation is concentrated among the least skilled, who lack the skill to notice their own gap.
Overcorrecting into chronic self-doubt
Treating every confident self-assessment as proof of bias can push people toward false modesty, which has its own cost: passing up roles, raises, or chances they were genuinely qualified for.
Using it to dismiss a specific, well-evidenced claim
Knowing the bias exists in general doesn't tell you that any one person's confident self-rating is wrong. Someone with a strong, documented track record can, in fact, be genuinely above average.
When not to use it
- As a blanket explanation for every workplace disagreement over ratings or credit — people also genuinely differ in which criteria they believe matter most, which isn't the same thing as bias.
- As grounds to dismiss someone's self-assessment without offering a concrete standard of your own to compare it against — that just substitutes one unmoored opinion for another.
- As the basis for compensation or promotion decisions by itself — "most people probably overrate themselves" is not a substitute for clear, agreed criteria applied consistently.
- To diagnose a specific individual from a single data point — it describes a pattern across groups, not proof that any one person's particular confidence is misplaced.
Try It Yourself: build your own prompt
Answer three quick questions and get a ready-to-use prompt for applying Illusory Superiority to your own situation.
Your prompt:
Save as Skill
A portable summary of Illusory Superiority you can hand to an AI assistant, a teammate, or your future self, so the model gets applied the same careful way every time.
Preview