Here’s a question you can try on yourself. Imagine a rare disease is spreading through a town of six hundred people. You have two treatment programmes available, and because of resource constraints, you can only use one. You have to choose.
Programme A will save two hundred people.
Programme B has a one-third chance of saving all six hundred people and a two-thirds chance of saving nobody.
Which do you choose?
Most people, asked this question, prefer Programme A. The certainty of saving two hundred lives feels better than the gamble that might save all or none. This is a reasonable preference; risk aversion in the face of life-and-death stakes is a defensible response.
Now consider a different pair of programmes.
Programme C will result in four hundred people dying.
Programme D has a two-thirds chance that six hundred people will die and a one-third chance that nobody will die.
Which do you choose?
Most people, asked this version, prefer Programme D. The certainty of four hundred deaths feels worse than the gamble that might mean all or none.
The strange thing, if you look at the numbers carefully, is that these are the same two options. Programme A saves 200 and lets 400 die; Programme C lets 400 die and saves 200. Programme B has a one-third chance of saving 600 and a two-thirds chance of saving none, which is identical to Programme D’s two-thirds chance of 600 deaths. The two framings describe exactly the same outcomes. But most people’s preferences reverse, depending on whether the numbers are presented as lives saved or lives lost.
This is called the framing effect, and it’s one of the most consequential findings in all of behavioural science.
The research that named it
The experiment described above, usually called the Asian disease problem, was designed by the Israeli psychologists Amos Tversky and Daniel Kahneman in 1981. Their finding — that identical information produces opposite decisions depending on framing — was striking enough to reshape several academic fields.
What Tversky and Kahneman showed, in this study and many that followed, was that humans don’t really evaluate options on their absolute merits. We evaluate them relative to reference points, and those reference points are easily manipulated by how information is presented. When outcomes are framed as gains relative to a baseline of harm (lives saved), we become risk-averse — we prefer the sure gain to the gamble. When outcomes are framed as losses relative to a baseline of safety (lives lost), we become risk-seeking — we prefer the gamble that might avoid loss to the certain loss.
This asymmetry is connected to the loss-aversion finding we’ve covered elsewhere in this series — losses hurt about twice as much as equivalent gains feel good. The same mechanism is operating here. A loss framing activates the psychological machinery that treats the situation as threatening, and people take bigger risks to avoid what feels like a loss than they would take to secure what feels like a gain.
The effect is not marginal. In the original Asian disease study, 72 per cent of participants chose Programme A under the saved-lives framing. When the same options were presented as Programme C versus D, 78 per cent chose D. The framing didn’t nudge people; it flipped them.
Why this matters in practice
The implications of this finding are significant because information almost never arrives unframed. Every statistic, every medical diagnosis, every policy proposal, every product description is delivered in some particular framing — and the framing, research now shows, substantially determines how people respond to the underlying information.
Medical communication. A surgery described as having a 90 per cent survival rate gets more patient consent than the same surgery described as having a 10 per cent mortality rate. The information is identical. The decisions, predictably, diverge. Doctors aware of this are sometimes careful to present statistics in both framings, or to use natural frequencies (“90 out of every 100 patients...”) rather than percentages that can be easily re-framed. But most medical communication, in ordinary practice, uses whatever framing happens to feel natural to the clinician, which is not necessarily the framing that serves the patient best.
Financial decisions. Investment products described in terms of potential gains attract different investors than identical products described in terms of potential losses. The phrasing “this fund has grown at 7 per cent per year” produces different behaviour from “this fund underperforms the market by an average of 1.5 per cent per year after fees”, even when both statements refer to the same fund. Marketing departments have known this for decades and choose accordingly.
Political communication. The political linguist George Lakoff has spent decades documenting how framing operates in political discourse. His argument, developed in books including Don’t Think of an Elephant, is that the terms used to describe policy debates often predetermine which side wins. A debate framed as “tax relief” has already taken a position — it implies taxes are a burden to be relieved. A debate framed as “investment in public services” has taken the opposite position. The specific policies under discussion may be identical; the framing determines how people respond. This is why political parties spend enormous resources trying to establish the framings in which issues are discussed, not just the positions within those framings.
Nudge-based policy. The behavioural economists Thaler and Sunstein, whose work on defaults we’ve covered, have extended the framing research into applied policy design. Their argument is that because framing is inescapable — there’s no neutral way to present information — policy-makers should deliberately choose framings that serve the public interest rather than accepting whichever framing happens to appear. Setting the default on organ donation to opt-out rather than opt-in is a framing choice. Presenting retirement savings as a normal deduction rather than a chosen contribution is a framing choice. These choices have measurable effects on behaviour, and they’re being made whether policy-makers are being deliberate about them or not.
The medical framing research in detail
Some of the cleanest evidence for framing effects comes from health communication. A classic study by Barbara McNeil and colleagues in the early 1980s presented medical students, patients, and physicians with information about two different lung-cancer treatments — surgery and radiation — framed in terms of either survival rates or mortality rates. The survival framings consistently produced more support for surgery; the mortality framings consistently produced more support for radiation. The effect appeared even in the physicians, who presumably were trained to evaluate the information objectively.
This finding has held up in multiple extensions. The framing effect appears in treatment preferences, in willingness to participate in screening, in vaccination decisions, in adherence to medication regimes, and in many other health behaviours. It appears in educated and uneducated populations, in different cultures, in different age groups. It’s not a quirk of naive lab participants. It’s a feature of how humans process probabilistic information about their own lives.
The counter-thread worth hearing
Before concluding that all decisions are hopelessly at the mercy of framing, a genuinely important caveat.
Framing effects are reduced substantially when the equivalence between framings is made explicit. Research by Craig Fox and Amos Tversky, and in other studies since, has shown that participants who are explicitly shown both framings side by side, and asked to notice that they describe the same options, respond much less strongly to the framing itself. They’re not immune — the effect doesn’t vanish entirely — but it shrinks considerably.
This matters because it means framing effects aren’t inescapable features of human decision-making. They’re features of decision-making under normal conditions, where people encounter one framing at a time and don’t usually notice that alternative framings would produce different responses. The practice of deliberately re-framing the information — asking yourself what the same facts would look like presented differently — substantially reduces susceptibility to the effect.
This is a trainable skill, though an unusual one. Most people don’t naturally ask, when presented with a statistic, how that statistic would read if the denominator or the frame were different. But the few seconds required to do so often changes the decision that follows.
A practical test
Here’s an exercise that builds resistance to framing effects over time. When you encounter a claim that carries a specific framing — a percentage, a statistic, a comparison — pause for a moment and ask what the alternative framing would look like.
Ninety per cent accuracy means one in ten is wrong. How does the claim look if you substitute the second?
This investment has a twenty per cent return is a gain framing. What does it mean about the losing side of the same investment? About the risk of loss that’s being under-emphasised?
Most doctors recommend is a social-proof framing. What does that mean in concrete terms? Fifty-one per cent? Ninety per cent? A group of ten doctors the marketer talked to?
Only one in a thousand side effects are serious is a small-numerator framing that implies safety. The reciprocal framing — of every thousand users, one has a serious side effect — produces different intuitions, even though the statistics are identical.
The goal isn’t to become suspicious of every framing you encounter. It’s to develop the habit of noticing that framings exist, and of occasionally re-framing to see whether your response holds up. Often it does. Sometimes it doesn’t. The ones where it doesn’t are usually the decisions worth slowing down on.
The question that remains
The deepest thing the framing research teaches is that the information you receive is never neutral. Every presentation of facts has been shaped, deliberately or not, by choices about what to emphasise, what to compare to, what baseline to assume. Those choices affect you more than you realise. The same underlying reality, described differently, produces different decisions — in you, in the people around you, in the political movements and marketing campaigns and health systems that shape collective life.
Knowing this doesn’t free you from framing effects. Nobody is free from them. But it makes you less at their mercy. The person who can casually say that’s one framing of the data — how does it look in the other? is harder to manipulate than the person who takes whichever framing arrives first as the real situation.
The question to carry, the next time you notice you’re responding strongly to a piece of information:
If the same information had arrived in the opposite framing — the same statistics, the same facts, different emphasis — would you be responding the same way?
Key research referenced: Amos Tversky and Daniel Kahneman, “The framing of decisions and the psychology of choice” (Science, 1981); Barbara McNeil and colleagues’ medical framing research (New England Journal of Medicine, 1982); George Lakoff, Don’t Think of an Elephant (2004); Thaler and Sunstein, Nudge (2008); Craig Fox and Amos Tversky on reducing framing effects through explicit comparison.