Group 1 · framed as lives saved
A 200 people will be saved.
B ⅓ chance 600 are saved, ⅔ chance none are.
72% chose the certain option — risk-averse
01The Problem
Every airline multiplies a guess
by another guess.
ICAO requires every operator in its member states to run a Safety Management System. The Safety Management Manual defines a risk as probability × severity: pick a likelihood from 1 to 5, pick a severity from A to E, combine them into a cell like 3A, and read off whether the risk is acceptable, tolerable, or unacceptable.
When there is enough data, that probability is a measurement. When there is not, it is a judgement call — and the manual offers no method for the low-data, high-uncertainty case. It quietly assumes the assessor is an unbounded rational decision-maker. Business aviation lives in exactly that gap: small fleets, low flight hours, too little history to infer from.
So the number that drives the whole system — whether a hazard gets mitigated, monitored, or waved through — is often an intuition. And intuitions are known to be movable.
02The Theory
Losses feel bigger than gains.
So wording moves the answer.
Kahneman and Tversky's prospect theory — the most influential account of decision-making under uncertainty — found that the value function is not symmetrical. It is steeper for losses than for gains. Four consequences follow, and the fourth is the one that matters here.
- Reference dependenceValue is judged from a reference point, not in absolute terms.
- Loss aversionA loss hurts more than an equivalent gain pleases.
- Diminishing sensitivityFacing likely gains, people turn risk-averse.
- Probability weightingExtreme probabilities are systematically over-weighted.
The framing effect, in its original form
Tversky and Kahneman put the same problem to two groups: an outbreak is expected to kill 600 people, and two programmes are proposed. The arithmetic is identical in both versions. Only the wording differs.
Group 2 · framed as lives lost
C 400 people will die.
D ⅓ chance nobody dies, ⅔ chance 600 do.
78% chose the gamble — loss-averse
Same expected value, opposite preference. The effect has been replicated for decades, traced to specific neural activity, and even observed in monkeys — but before this paper it had not been tested on safety risk assessment.
03The Study
Two groups. Same hazards.
Different sentences.
Business aviation safety professionals — safety managers and key SMS personnel drawn from the Turkish DGCA operator list — were split at random into two groups. Group A rated hazards as they are normally written. Group B rated the same hazards rewritten as the desired outcome. Everyone used ICAO's 1–5 probability scale.
Study 1 · pilot
39 respondents
One event, plus a free-text question asking respondents to explain their reasoning. Group A n = 22, Group B n = 17.
Study 2 · main
49 respondents
Five events spanning human, technical and environmental hazards. Run seven months later so nobody remembered the pilot. Group A n = 24, Group B n = 25.
The five statements
Framing was done three ways: negating the verb, swapping in its antonym, or restructuring the sentence to express the complement of the event. Mathematically the pairs are equivalent — if an event has probability P, its complement has 1 − P.
| # | Unframed — negative | Framed — positive | Type |
|---|---|---|---|
| E1 | a pilot or technician missing an oil leak during pre-flight check | a pilot or technician finding an oil leak during pre-flight check | Human |
| E2 | having a critical engine failure during take-off | taking off without a critical engine failure | Technical |
| E3 | someone being injured during severe turbulence | getting over severe turbulence without any injuries | Environmental |
| E4 | encountering birds during final approach | completing final approach without encountering birds | Environmental |
| E5 | a pilot or technician missing a checklist item while experiencing fatigue | a pilot or technician completing a checklist precisely while experiencing fatigue | Human |
04Results
Every hazard scored higher.
The human ones, significantly.
Flip the wording below. The gauge moves from what Group A said to what Group B said about the very same hazard — their answers inverted so the two are directly comparable.
Human factors
The probability of a pilot or technician missing an oil leak during pre-flight check…
1.58 mean assessed probability — Group A, unframed
To test that sentence an assessor looks for reasons it won't happen — and lists the controls already in place. The risk looks small.
All five events
Only the human-factors events moved
E1 and E5 — missing an oil leak, missing a checklist item under fatigue — shifted significantly. The purely technical event did not. Manufacturers publish failure statistics for engines; nobody publishes them for tired people.
Uncertainty is the ingredient
That pattern is what prospect theory predicts: the greater the uncertainty, the stronger the loss and risk aversion. Where a "known risk" has hard numbers behind it, wording has little left to move.
And the reasoning changed too
In the pilot study, 77% of Group A explained their answer through existing safeguards. Among the positively framed group only 44% did — the rest reasoned from weaknesses.
05Why It Happens
People try to falsify
the sentence in front of them.
This is the paper's explanation, and it is worth sitting with. Faced with a claim, an expert instinctively tries to knock it down. What counts as knocking it down depends entirely on which way the claim points.
Negative statement
“The leak will be missed.”
To disprove it, find reasons it won't be missed
You inventory your defences — training, checklists, double-checks
Risk rated low
Positive statement
“The leak will be found.”
To disprove it, find reasons it won't be found
You inventory your weaknesses — fatigue, haste, poor light, complacency
Risk rated higher
Both assessors are being rigorous. Both are doing exactly what a good safety professional should do. The sentence decides which half of the picture they get to see.
06Positive Safety
The practice:
assess it both ways.
The paper's own recommendation is deliberately conservative, and it is the whole of the method. Put both wordings of a hazard in front of your assessors, and carry forward whichever produces the higher risk value.
- Write the hazard as you always would The undesired event, in the negative — exactly what your hazard log holds today.
- Write its complement The same event as the outcome you want: negate the verb, use its antonym, or restructure the sentence to describe success instead of failure.
- Assess both, then take the higher Invert the framed rating to put it on the same scale, compare, and carry the more cautious of the two into the matrix.
Why one scale point is not a rounding error
A shift of about one point on a five-point scale sounds modest until it crosses a boundary in the risk matrix. Take E1's measured shift — 1.58 to 2.56, so probability 2 becomes probability 3 — and choose a severity to see where it lands.
Severity of consequence
| Probability ╲ Severity | A | B | C | D | E |
|---|
Adapted from ICAO's Safety Management Manual. The six cells the paper names as unacceptable — 3A, 4A, 4B, 5A, 5B, 5C — are marked; the paper does not enumerate the boundary between tolerable and acceptable, so the remaining cells are left undivided here. The 2 → 3 move is this site's illustration of the measured shift, not an analysis from the paper.
07Limits
What this does not
yet establish.
The authors are explicit about the boundaries of the result, and any honest reading of it should be too.
- Small samples. 39 and 49 respondents, drawn from the limited population of key safety personnel in Turkish business aviation. Replication at scale is needed.
- Three of five events showed no significant shift. All five rose, but only the two human-factors events did so significantly.
- More cautious is not automatically better. Whether framing improves real safety outcomes — or tips into over-protection and wasted mitigation — can only be answered by long-term monitoring in live use.
- Open questions. Does the human-factors effect come from uncertainty specifically? Do experience, familiarity or demographics change how strongly an assessor responds to framing?
What the study does establish is narrower and still consequential: prospect theory's predictions about decision-making under uncertainty are not confined to money. They reach into safety risk assessment, where the stakes are not measured in currency.
08Cite & Access
Cite this work
Uyar, T., & Paksoy, M. (2020). Framing effect in safety risk probability assessment: A prospect theory approach. The International Journal of Aerospace Psychology, 30(3–4), 119–129. https://doi.org/10.1080/24721840.2020.1774379
BibTeX
@article{Uyar2020Framing,
author = {Uyar, Tevfik and Paksoy, Mahmut},
title = {Framing Effect in Safety Risk Probability Assessment:
A Prospect Theory Approach},
journal = {The International Journal of Aerospace Psychology},
year = {2020},
volume = {30},
number = {3--4},
pages = {119--129},
doi = {10.1080/24721840.2020.1774379},
url = {https://doi.org/10.1080/24721840.2020.1774379}
}
Authors
Tevfik Uyar
Faculty of Economics and Administrative Sciences
Istanbul Kültür University, İstanbul, Turkey
Corresponding author
[email protected] ORCID 0000-0003-0124-6910 ↗Mahmut Paksoy
Faculty of Economics and Administrative Sciences
Istanbul Kültür University, İstanbul, Turkey