Many preferences are assembled at the moment of choosing rather than looked up

The assumption inside almost all advice about choosing

Most guidance about decisions assumes that you have preferences and the task is to find the option that best matches them. Work out what you want, compare the candidates, pick the closest fit. The procedure is sound whenever the assumption holds, and the assumption holds far less often than the procedure implies.

For familiar, frequently repeated choices, a stable preference plausibly exists: you know which of two foods you prefer because you have compared them many times. For anything rare, complex or unfamiliar — a job, a policy, an unusual purchase, a medical trade-off — there may be no stored answer at all, and what feels like consulting a preference is closer to constructing one on demand from whatever the situation makes salient.

The reversal that made the point unavoidable

The clearest evidence comes from a set of findings in which people prefer one option when asked to choose between two, and place a higher monetary value on the other when asked to price them separately. The same person, the same two options, opposite orderings depending only on the response format. That is not a small anomaly at the edge of a theory; it is a direct contradiction of the idea that a single underlying ranking is being read out.

These reversals have been reproduced repeatedly, including in settings with real money and with experienced participants, and economists spent a considerable period trying to make them disappear before accepting them. The standard explanation is that different formats make different attributes prominent: choosing draws attention to one feature of the options, pricing draws attention to another, and the ranking follows whatever was in focus.

Context effects, and which ones actually survived

A large family of findings shows that adding an irrelevant option changes preferences between the original ones. The best known is the case where a clearly inferior version of one option raises that option’s share. This effect is real in some settings and it has had a genuinely difficult replication history: several careful attempts have found it weak or absent, particularly with realistic multi-attribute goods and with participants making consequential choices.

That mixed record is itself informative. It suggests that constructed preference is not a licence to believe every published context effect, and that the robust core is narrower than the popular version. The response-format reversals are on much firmer ground than the decoy manipulations, and it is worth keeping the two apart rather than citing them as one body of evidence.

What this does not mean

It does not mean preferences are arbitrary or that people can be moved to want anything. The influences operate within limits, and they mostly bite when the options are genuinely close on the dimensions the person cares about. Nobody is talked out of a strongly held value by an option order.

It also does not mean the constructed answer is wrong. A preference assembled thoughtfully in the moment can be an excellent one; the problem is only that it was produced partly by the presentation, which means a different presentation would have produced a different answer with the same feeling of correctness attached.

Where the practical damage lands

It lands wherever somebody else controls the format. The order of options, the reference point supplied, the attributes shown in the comparison table and the ones omitted, and whether you are asked to choose or to rate — all of these are decisions made by whoever built the interface, and all of them shape the output. This is not necessarily manipulation, since some format has to be chosen, but it does mean the presentation is never neutral.

It also lands in your own reasoning, in a subtler way. A preference formed while looking at a set of options is anchored to that set. Ask yourself what you want after seeing three flats and the answer will describe a compromise among those three, not what you would have said before.

Working with a preference you had to build

The most useful single habit is to write down what matters before looking at the options, in whatever crude form you can manage. It does not have to be a weighting scheme. A short list of what the decision is for, made in advance, gives you something the presentation did not supply.

The second is to re-ask the question in a different format when the decision is important. If you chose, then price. If you ranked, then pick one and ask what would have to be true for the other to be better. Agreement across formats is weak evidence that a stable preference exists. Disagreement tells you that you are constructing, and that the construction is currently being done by whoever laid out the page.