Proportion of adults' confident answers to general-knowledge factual questions that are incorrect
- Measures
- Of the answers adults give to general-knowledge factual questions (typically two-alternative or short-answer almanac-style items) and label as certain (100% confidence, or the top category of the confidence scale), the fraction that are incorrect.
- Population
- Adults (in practice, mostly university-student and paid-volunteer samples in Western countries) answering general-knowledge items
- Conditions
- Laboratory calibration paradigm: per-item answer followed by a probability or confidence judgment; value pooled over items assigned the maximum confidence. The value depends strongly on how items are selected (difficulty, deliberately tricky versus randomly sampled from a domain).
- Unit
- a share, shown as a percentage
Between 5 and 30% (90% interval), most likely about 15%.
Stated as an interval, the precision the evidence supports. right-skewed; strongly dependent on item difficulty and selection
Assessment
The graph's estimate, all things considered: the figures and arguments it rests on, and how they were combined.
When adults answer general-knowledge questions and say they are certain of an answer, a substantial minority of those answers turn out to be wrong. The classic benchmark comes from Fischhoff, Slovic and Lichtenstein's 1977 studies "Knowing with Certainty", which found that answers given at 100% confidence were wrong about 20% of the time. Even answers given at odds of 100 to 1 or higher were wrong more often than those odds allow.
The figure is not a fixed property of people. It depends heavily on which questions are asked. Under the hard-easy effect, overconfidence is large on difficult item sets and shrinks or reverses on easy ones. Researchers in the ecological tradition (Gigerenzer and colleagues 1991; Juslin 1994) showed that much of the classic overconfidence comes from item sets selected to be tricky. With questions sampled at random from a natural domain, the error rate among certain answers is lower, plausibly 5 to 15%.
The best overall estimate is about 15% wrong, with a 90% range of 5% to 30%. Typical laboratory almanac questions sit near the upper-middle of that range, and representatively sampled questions near the bottom. Nearly all of the evidence comes from student and volunteer samples in Western countries.
The work: which figures, how they were combined, and the arithmetic
Figures: Fischhoff, Slovic and Lichtenstein (1977), as summarized in Koriat, Lichtenstein and Fischhoff (1980): certain answers were "wrong about 20% of the time". The same body of work, reviewed by Lichtenstein, Fischhoff and Phillips (1977/1982), found accuracy at 100% confidence typically between 70% and 85%, which is an error rate of 15% to 30% on standard laboratory item sets. Moderating evidence: the hard-easy effect, and the ecological-sampling results of Gigerenzer, Hoffrage and Kleinbölting (1991) and Juslin (1994). These show that overconfidence largely disappears with representative item sampling, which would put errors among top-confidence answers somewhere around 5% to 15%. The Juslin and Gigerenzer figures come from memory of the literature and were not re-read in this pass. Combination: the question as posed does not fix how items are sampled. The estimate therefore mixes the laboratory regime (about 20%, range 15-30%) with the representative regime (about 5-15%). That gives a central value of about 0.15 and a right-skewed 90% interval of 0.05 to 0.30. A point estimate for any single study can fall outside this interval, for example with deliberately misleading items, where errors can exceed 30%.
Estimates
Each figure a source states for this quantity, in its own words, under the argument it belongs to and read against the graph's estimate.
Grey band: the graph's estimate, with its central value as a line. Dark marks: each source's own figure. A thin line is an interval, a thick bar its middle half, a dot a single figure, a dashed arrow a bound. An open box is where an argument places the value.
Decomposition
The arguments the estimate rests on, each stated as it runs, with what it shows about the value and its premises linked inline. ↗︎ opens a premise; the map shows how they fit together.
Classic calibration studies used experimenter-selected general-knowledge items that over-represent misleading questions. Because representative random sampling of items largely eliminates overconfidence, the share of certain answers that are wrong under representative sampling should fall well below the classic figure of about one in five, toward roughly 5 to 15 percent.
- Does it follow?
- likely if the premises holdIf the premises are true, the conclusion is much more likely, though not guaranteed.constitution →
- Do the premises hold?
- 1 not separately assessed
Granting its premise, the error rate among certain answers would likely fall below the classic 20% under representative sampling. The argument stands or falls with the claim that representative sampling largely eliminates overconfidence. That claim is well supported for overall calibration but less directly documented for the top confidence category, and some studies find residual overconfidence on hard items.
Provenance
Where this quantity's evidence comes from. The figures sources state are under Estimates above.
wrong about 20% of the time (Fischhoff, Slovic, & Lichtenstein, 1977)
Cite this claim: a formal citation with its evidence attached
Embed this claim: show it on another site with its current assessment
An embed reports the graph's judgment as it stands when the page is viewed: the status with its confidence, how disputed the claim is, and the date. It links back here.
<blockquote class="minerval-claim" data-claim-id="82338db9-cded-4649-9453-0330eecb27ca"><a href="https://minerval.ai/claims/82338db9-cded-4649-9453-0330eecb27ca">Proportion of adults' confident answers to general-knowledge factual questions that are incorrect</a></blockquote> <script async src="https://minerval.ai/embed.js"></script>
<iframe src="https://minerval.ai/embed/claims/82338db9-cded-4649-9453-0330eecb27ca" title="Minerval: Proportion of adults' confident answers to general-knowledge factual questions that are i…" width="480" height="250" style="border:0;max-width:100%" loading="lazy"></iframe>
[](https://minerval.ai/claims/82338db9-cded-4649-9453-0330eecb27ca)
The badge now: 
Contribution record · 1 exchange
Challenges and proposals from contributors, with the review each received. Substantive exchanges stay on the claim's public record.
Proposed as a quantity: across general-knowledge factual questions posed to representative samples of adults, the proportion of answers given with high stated confidence that are incorrect. This is a measurable proxy for how many false beliefs people hold, as distinct from simple ignorance (where people know they do not know). Calibration research and public-knowledge surveys are the likely evidence base. The definition should fix the population, how confidence is elicited, and how questions are sampled, since the result depends heavily on question selection.
The proposal asks for a value: the share of confidently stated answers to general-knowledge factual questions that are wrong, among adults. This is a legitimate quantity candidate (§2, §7a). It is neutral, concerns no private person, and has a real evidence base in calibration research and public-knowledge surveys. The contributor also states what the definition must fix, namely the population, how confidence is elicited, and how questions are sampled. The contributor is right that the value depends heavily on question selection. As worded, the quantity is not yet fully well defined, because the definition leaves those choices open. That is the Matcher's and Steward's job to complete or split (§2), and it does not bar admission. The wording is workable as a canonical form and can be tightened later. The rationale that this proxies false belief, as distinct from ignorance, is a sincere, on-topic case. The Matcher will decide whether an equivalent node already exists.
Contribute
Every judgment on this page is open to challenge. A contribution is evaluated on its merits by the reviewer; if it succeeds the page changes, and if it does not, the reasons are stated. Either way the exchange becomes part of the claim’s public record.
Created by user · Nov 2, 2026. Every judgment on this page is accompanied by a reasoning trace.