Description-experience gap
People treat rare risks very differently when they're described versus learned from experience.
What it means
The description-experience gap is the systematic divergence between choices made from explicit descriptions of probabilities and choices made by sampling outcomes from experience. From description, people overweight rare events, as prospect theory predicts; from experience, they tend to underweight them, behaving as if rare events are even less likely than they are. Two mechanisms drive the experiential pattern: small samples often fail to contain the rare event at all, and recently sampled outcomes loom larger in memory. The gap is consequential because many real-world risks — financial, medical, environmental — are learned through experience rather than read off a label, so warnings stated as probabilities may not reproduce experiential behavior. It reframed a long-standing assumption that probability weighting is a single, format-independent phenomenon.
What the evidence shows
The gap first appeared in Hertwig, Barron, Weber and Erev's 2004 experiments, where a rare outcome that description-based choosers avoided was chosen by experience-based samplers as if its probability were near zero. Wulff, Mergenthaler-Canseco and Hertwig's 2018 meta-analysis, pooling more than a hundred experiments, put the average gap at roughly nineteen percentage points when a risky option is set against a safe one, and about seven points when two risky options compete. The direction is consistent across studies: sampling outcomes shifts choices as though rare events matter less than their stated odds. It is one of the more reproducible findings in decision research, but the size of the gap depends heavily on how many outcomes a person draws before committing, which is why individual studies vary so widely.
How much is a genuine bias
Not all of the gap reflects a real change in how people weight probabilities. Because experiential learners take small samples, they often never encounter the rare event, or meet it less often than its true rate, which is pure statistical undersampling rather than distorted judgment. Fox and Hadar (2006) reanalysed the original data and argued that once this sampling error is removed, experiential choices are broadly consistent with prospect theory rather than a reversal of it. Later work complicates that verdict: underweighting tends to persist even when samples are large or the rare event has actually been seen, pointing to a second, memory-based mechanism in which recent and vivid draws crowd out older ones. The honest summary is that the gap is part sampling artefact, part genuine underweighting.
Where it shows up
Most consequential risks are not read off a label but accumulated through years of uneventful experience: driving without a crash, skipping backups without losing a file, living in a floodplain through dry seasons, taking a drug without side effects. In each case the base rate is low and personal sampling is sparse, so lived experience teaches that the hazard is effectively nil while the stated probability says otherwise. This is why safety warnings, insurance disclosures and public-health messages framed as probabilities often fail to move behaviour: they speak the language of description to people whose beliefs were formed by experience. Cybersecurity complacency, vaccine hesitancy and slow climate adaptation all fit the pattern, where a long run of non-events steadily erodes the felt likelihood of a real but rare threat.
Using it in practice
Because the two modes pull in opposite directions, the format of a risk message can matter as much as its content. When the aim is to make a rare hazard feel real, letting people sample simulated outcomes — an interface that occasionally serves the bad draw — can raise its felt likelihood more than a bare percentage does. When the aim is to stop people overreacting to a vivid but tiny risk, plain description may cool it down. Presenting both together tends to narrow the gap. Practitioners should also resist reading revealed choices as fixed preferences: the same person can look risk-seeking from experience and risk-averse from description over an identical prospect, so the elicitation method partly manufactures the answer you get.
Examples
Told a medication carries a 0.1% chance of a severe side effect, patients fret; people who have taken it for years without incident barely consider the risk.
Workers told the harness prevents a one-in-a-thousand fall take it seriously at induction; ten years of never seeing anyone fall teaches them the risk is zero, and the harness stays in the locker.
Homeowners quoted a 1% annual flood risk buy the cover. Those who have lived there through twenty dry years let it lapse — the rare event has simply never appeared in their sample.
An investor who started during a long bull market has never sampled a crash, so the prospectus warning that shares can fall sharply reads as boilerplate and heavy leverage feels safe.
A team that has shipped to production for years without an outage quietly drops its rollback drills; the documented risk of a bad release never turned up in their short run of luck.
First described in Hertwig, Barron, Weber & Erev (2004).
Key references
- Hertwig, R., & Wulff, D. U. (2022). A description-experience framework of the psychology of risk. Perspectives on Psychological Science, 17(3), 631-651. doi.org/10.1177/17456916211026896
- Wulff, D. U., Mergenthaler-Canseco, M., & Hertwig, R. (2018). A meta-analytic review of two modes of learning and the description-experience gap. Psychological Bulletin, 144(2), 140-176. doi.org/10.1037/bul0000115
- Hertwig, R., & Erev, I. (2009). The description-experience gap in risky choice. Trends in Cognitive Sciences, 13(12), 517-523. doi.org/10.1016/j.tics.2009.09.004
- Fox, C. R., & Hadar, L. (2006). "Decisions from experience" = sampling error + prospect theory: Reconsidering Hertwig, Barron, Weber & Erev (2004). Judgment and Decision Making, 1(2), 159-161. www.cambridge.org/core/journals/judgment-and-decision-making/article/decisions-from-experience-sampling-error-prospect-theory-reconsidering-hertwig-barron-weber-erev-2004/A3059BE69934789B636A7E43F1E6AD82
- Hertwig, R., Barron, G., Weber, E. U., & Erev, I. (2004). Decisions from experience and the effect of rare events in risky choice. Psychological Science, 15(8), 534-539. doi.org/10.1111/j.0956-7976.2004.00715.x