How susceptible is your decision-making to cognitive bias? A ~15-minute assessment using behavioral tasks — not personality quiz questions — to measure your susceptibility to seven well-studied cognitive biases. Grounded in two decades of research from MIT, Princeton, Harvard, Toronto, and Carnegie Mellon. Everything runs in your browser. Nothing is transmitted or stored.
Cognitive bias measurement is an active and contested area of psychology. Reasonable researchers reading the same evidence reach different conclusions. Some (Stanovich, University of Toronto) argue rational thinking is a measurable general ability. Others (Teovanović, Belgrade) find cognitive biases don't correlate strongly enough to support a single "rationality" factor.
Your per-bias scores (anchoring, sunk cost, etc.) are reasonably well-established — those individual effects replicate across hundreds of studies. But your archetype and composite factor scores reflect one interpretation of your results. A different researcher using different analytical choices might characterize you differently.
Treat your results as a starting point for self-reflection, not a fixed label. Cognitive biases are universal human cognitive architecture — every person has all seven to some degree. High scores are an invitation to try the research-backed interventions we'll suggest, not a character judgment.
For researchers and curious users: read the full methodology — the documented framework, the variables measured, the scoring algorithm, the limitations, and the references.
You'll work through seven short tasks — not a personality quiz, but actual behavioral measurements that probe how your mind handles specific decision challenges. The tasks are drawn from the peer-reviewed heuristics-and-biases literature (Kahneman, Stanovich, Bruine de Bruin, others). Some will feel like estimation games. Some will feel like small logic puzzles. Some are scenarios where you rate how reasonable a decision was.
1. Don't look things up. Some tasks ask you to estimate facts you're unlikely to know precisely. The point isn't to be correct — it's to see how your mind handles uncertainty. Looking things up defeats the measurement.
2. Don't overthink. Give the answer that feels right at first pass, then commit. The tasks are designed to work with quick, genuine responses. Excessive deliberation can mask the effects we're measuring.
3. You can't fail. There are no wrong answers in the punitive sense — only responses that do or don't show susceptibility to a given bias. Every human has all seven biases to some degree. This measures the shape of yours.
For each question, give your best estimate. You're unlikely to know the exact answer — that's by design. Don't look anything up; don't spend more than 10 seconds per item. If you have no idea, guess a number that feels plausibly in the right neighborhood.
Each scenario describes a real event before its resolution. Without looking anything up, estimate the probability you'd have assigned to the outcome described, given only what was known at the time. Move the slider to your best guess. We'll return to this task near the end.
For each question, choose the answer you think is correct, then rate how confident you are (50% = just guessing, 100% = absolutely certain). Don't look anything up. This task measures calibration — the gap between how confident you feel and how often you're actually right.
You'll see the same six estimation questions from Task 1. This time, each comes with a reference value. First indicate whether the true answer is higher or lower than the reference, then give your final estimate. Your previous answers are not shown — answer each fresh.
Each problem shows two premises followed by a conclusion. Your job: determine whether the conclusion logically follows from the premises — regardless of whether it happens to be true in real life. This is about logical structure, not factual accuracy.
Each problem gives you two pieces of information: a general base rate (how common something is in the population) and a specific signal (evidence about a particular case). Your job: combine them into a probability estimate. This measures how well you integrate base rates with individuating information.
Each scenario describes someone facing a choice about whether to continue or stop. Rate how likely you would be to continue in each situation. There are no right answers — scenarios are designed to probe how past investment influences future decisions.
Each case describes a decision made under uncertainty, followed by how things turned out. Rate the quality of the decision itself — the reasoning, preparation, and choice — independent of the outcome. Decisions can be well-reasoned but produce bad outcomes, or poorly reasoned but produce good ones.
Now we'll reveal what actually happened in the four scenarios from Task 2. Without looking at your earlier answers, try to recall what probability you assigned when you first saw each scenario. The gap between your recall and your actual earlier rating is the hindsight bias measurement.
These are LBL-original self-report statements, written to mirror the seven biases the tasks above measured. They capture how you perceive your own tendencies — which often differs from how you actually behaved on the tasks above. The gap between the two is the Bias Blind Spot, and it's a featured finding in your results.
A cognitive bias test measures how far your judgments drift from a normative standard under specific, well-studied conditions. This one does it with behavioral tasks rather than self-report alone: you make estimates, judge probabilities, evaluate arguments, and rate decisions, and the scoring compares what you actually did against what the research says an unbiased responder would do. It covers seven biases — anchoring, hindsight, overconfidence, belief bias, base-rate neglect, sunk cost, and outcome bias — plus your Bias Blind Spot: the gap between how biased you are and how biased you think you are. It takes about 15 minutes, runs entirely in your browser, and stores nothing.
Asking "how susceptible are you to anchoring?" measures your theory of yourself, not your behavior. The methodological move that makes bias measurement possible is to put you in the situation and observe the drift. Anchoring is measured by giving one group a reference number and another none, then comparing estimates. Hindsight bias is measured by asking for a prediction, revealing the outcome, and then asking what you originally predicted. Both designs are built into this assessment, which is why some questions come back a second time.
An arbitrary number pulls your estimate toward it, even when you know it is arbitrary. How it is measured here: you give six numerical estimates with no reference point (Phase A). Later, the same six questions return with a reference number attached (Phase B). Your anchoring score is the movement between them. Research anchor: Kahneman and Tversky's heuristics-and-biases program.
Once you know how something turned out, your memory of what you expected shifts toward the outcome. How it is measured here: you rate how likely four historical outcomes were (Phase 1). Much later, after the actual outcomes are revealed, you are asked to recall your own earlier ratings (Phase 2). The distance between your real rating and your recalled rating is the bias.
Confidence outruns accuracy. How it is measured here: eight general-knowledge questions, each followed by a confidence rating. Calibration is the relationship between the two — a well-calibrated person is right about 70% of the time when they say 70%.
You judge an argument as valid because you agree with its conclusion, not because the logic holds. How it is measured here: eight short logic problems in which validity and believability are deliberately crossed, so agreement and correctness come apart.
Vivid, specific information swamps the underlying prior probability. How it is measured here: four probability problems where the base rate is stated plainly and the individuating detail pulls the other way.
Money, time, or effort already spent — and unrecoverable — keeps you committed to a failing course. How it is measured here: four decision scenarios in which the rational choice is to stop, and the prior investment varies.
You rate a decision as good because it worked out, ignoring what was knowable at the time. How it is measured here: eight short decision cases, each rated for how well-reasoned it was, with outcomes varied independently of reasoning quality.
Twelve self-report statements ask how susceptible you believe you are. Your Bias Blind Spot is the discrepancy between that self-assessment and your measured task performance. Scopelliti and colleagues (2015) demonstrated that most people underestimate their own susceptibility while readily spotting bias in others — and that this blind spot is itself measurable and varies between people.
Twelve steps, seven tasks, roughly 15 minutes. The order is not arbitrary. Two of the tasks are split in half and separated by other work, because anchoring and hindsight can only be measured across a gap: you must answer once without the manipulation, do something else, and answer again with it. If you skip ahead or look back, you break the measurement — which is why the test does not let you.
Some questions will feel repetitive. They are not. The second pass is the experiment.
Each of the seven biases is scored on its own task, then expressed as a susceptibility percentile relative to the scoring model. Those scores are summarized in a two-factor interpretation rather than a single number, and your responses also place you in one of six archetypes — The Anchored, The Contrarian, The Escalator, The Certain, The Calibrated, or The Variable — describing the shape of your profile, not its quality.
There is deliberately no single "rationality score." The research is genuinely divided on whether one exists: Stanovich (2016) argues for a rationality factor with caveats; Teovanović and colleagues (2015) find the biases largely independent. Publishing one number would mean picking a side of an open scientific question and hiding that from you.
Three things worth holding onto.
Susceptibility is not intelligence. Twenty-plus years of work by Stanovich, West, and Toplak shows bias susceptibility is largely independent of IQ. Highly intelligent people can be highly susceptible; the two abilities dissociate.
Everyone has all seven. These are features of ordinary human cognitive architecture, not defects. A high score is an invitation to try the interventions listed for that bias, not a diagnosis.
Your score is a snapshot. Susceptibility shifts with practice, mood, time of day, and item set — research suggests movement on the order of 10–15 percentile points. Retest in a month and expect a different number. That is measurement working, not failing.
The literature on debiasing is more modest than the popular literature suggests, but a few strategies have support. Consider the opposite: before committing to a judgment, explicitly generate reasons it could be wrong — one of the better-replicated interventions, effective against overconfidence and hindsight. Take the outside view: ask what usually happens to projects like this one, rather than reasoning from the specifics of yours; this directly counters base-rate neglect. Pre-commit to a stopping rule before you invest, which is the only reliable defense against sunk cost, since the bias operates after the money is spent. Separate decision quality from outcome quality in reviews: ask what was knowable at the time, not what happened.
This is an educational instrument, not a clinical or diagnostic one. It uses a fixed item set, so repeated testing with the same items measures partly your memory of them. Scores are percentile positions within a scoring model, not clinical norms. The Bias Blind Spot is a discrepancy measure and inherits the noise of both components. Bias tasks in general show modest test-retest reliability — a well-documented property of the field, not a flaw in this implementation, and a reason to treat any single score as provisional.
The instrument draws on the heuristics-and-biases and rational-thinking literatures: Kahneman and Tversky on heuristics and biases; Frederick on cognitive reflection; Bruine de Bruin and colleagues on adult decision-making competence; Stanovich, West, and Toplak on the independence of rationality and intelligence; Scopelliti and colleagues on the bias blind spot; and Teovanović and colleagues on the structure — or absence of structure — among individual bias measures. Full references and scoring detail are in the methodology.
Your susceptibility to 7 well-studied cognitive biases — anchoring, belief bias, overconfidence, hindsight bias, base rate neglect, sunk cost, and outcome bias — using actual behavioral tasks rather than self-report alone. The tool also computes your Bias Blind Spot (the gap between how biased you actually are and how biased you think you are).
About 15 minutes. Some people finish in 12, some take 20. The tasks require genuine thinking, so rushing produces noisier results.
No. The research is divided on whether a single 'rationality factor' exists. Stanovich 2016 argues yes with caveats; Teovanović 2015 argues no. LBL-CBS gives you per-bias scores plus a two-factor interpretation, explicitly framed as one perspective among several.
No. 20+ years of research (Stanovich, West, Toplak) shows cognitive bias susceptibility is largely independent of IQ. Smart people can be highly susceptible; less-formally-educated people can be well-calibrated. They are different cognitive capacities.
The gap between how biased you actually are (behavioral tasks) and how biased you rate yourself (self-report). Scopelliti et al. 2015 demonstrated that most people underestimate their own susceptibility. LBL-CBS shows you your specific gap.
No. All computation happens in your browser. Nothing is transmitted to any server. No accounts. No tracking. Close the tab and all data is gone.
Yes, by design. Research shows bias susceptibility can shift 10-15 percentile points with deliberate practice, mood, time of day, and retesting with different items. Your score is a snapshot, not a fixed trait.
Primary foundations come from the heuristics-and-biases and rational-thinking literature — Frederick (cognitive reflection), Kahneman and Tversky (heuristics and biases), Bruine de Bruin (decision competence), Pronin and Scopelliti (bias blind spot), Stanovich, West, and Toplak (rational thinking), and Baron (decision theory), among others. All items are LBL-original; the full reference list is at the bottom of the results page.
Yes, with a privacy-respecting share card that includes your archetype label and Bias Blind Spot score. No personal identifiers are included.
No. Cognitive biases are universal human cognitive architecture. Every person has all 7 biases to some degree. High scores are an invitation to try the research-backed interventions listed for each bias, not a character judgment.