The Frame

A choice has to arrive in some form. Someone names the alternatives, selects a point of comparison and decides which consequences to place beside one another. If the description changes what the person learns about the alternatives, a changed decision is readily intelligible. The harder case holds the alternatives still. The same consequences, with the same chances of occurring, are put into different words, and the choice moves. What, then, has the choice revealed? The preceding plate tested the coherence of preferences across different gambles. This plate asks what happens before that comparison begins: how the decision-maker identifies the outcome to be valued. The frame is the account of the problem that enters the judgment, and its relation to the underlying alternatives is itself something to examine.

Tversky and Kahneman gave the question a memorable form in 1981. Respondents considered a hypothetical disease expected to kill six hundred people. One program would save two hundred with certainty; the other offered a one-third chance of saving everyone and a two-thirds chance of saving nobody. Seventy-two per cent of the first group selected the certain rescue. A different group received the equivalent description in deaths: four hundred would die for certain, or there would be a one-third chance of nobody dying and a two-thirds chance of all six hundred dying. Seventy-eight per cent of that group selected the gamble. The identities are exact: two hundred saved means four hundred dead, and saving all or none means losing none or all. The reported result is a difference between groups under these descriptions, rather than proof that every individual would reverse his answer. Its significance lies in the change in the pattern of choice while the stated outcomes remain equivalent.¹

Prospect theory supplies one way to organize the difference. Outcomes are evaluated as changes from a reference point: a gain relative to one baseline can be a loss relative to another. The proposed value function usually bends downward over gains and upward over losses, expressing diminishing sensitivity as the change grows larger; it also assigns a larger impact to a loss than to an equal gain near the reference. These are distinct features. A person can become less willing to exchange a sure gain for a gamble because of curvature, while the greater weight on losses governs comparisons that cross the reference point. Decision weights introduce a further departure from ordinary probability weighting. The model therefore has several moving parts, and one observed reversal cannot measure all of them. The fuller Prospect distinguishes their roles in risk. Here the consequential step is the placement of the reference from which the outcome is read.²

A small cash example isolates what a movable reference can do. Compare a receipt of two hundred for certain with a one-third chance of receiving six hundred and a two-thirds chance of receiving nothing. Keep the person's actual resources and both options unchanged. Suppose gains are valued by their square root and losses by minus twice the square root of their size, and use the stated probabilities directly as weights. Relative to a reference receipt of zero, the sure option scores about fourteen, against about eight for the gamble. Relative to a reference receipt of six hundred, the sure option is a loss of four hundred and scores minus forty; the gamble scores about minus thirty-three, so it is preferred. The shift changes the evaluation without changing the receipts. If value were linear with the same slope on both sides, both options would remain tied under either reference. The reversal in the illustration needs the specified shape and reference change; simply renaming a number is not an explanation on its own.³

The frame can also decide which consequences enter the account together. Imagine two ways of describing the same posted terms: a purchase costs a hundred in cash and a hundred and five by card. One description starts from the higher price and offers a cash discount; the other starts from the lower price and adds a card surcharge. The final menu of payments is unchanged, but the reference suggested by the description differs. A claim that the descriptions affect choice must therefore separate that suggestion from any actual difference in fees, convenience or protection attached to the payment methods. The same discipline applies to separating an expense from the larger purchase that gives it meaning. A saving can look substantial beside a small bill and negligible beside a large one, even when its amount and the effort required are unchanged. Framing directs attention to a comparison. Establishing what it changes requires keeping the other comparisons available.⁴

The evidence leaves room for several accounts of the mind behind the answer. A response may express a preference assembled in the moment; it may also reflect attention, interpretation or a failure to recognize that two descriptions coincide. Those possibilities matter to the inference. A speaker who emphasizes a particular outcome can sometimes convey information about its significance, so physical equivalence on the analyst's page does not automatically establish informational equivalence for the listener. Conversely, a convincing demonstration of equivalent alternatives with different choices prevents us from treating the answer as a transparent reading of one fixed, description-independent ordering. The appropriate test compares what was offered, what was understood and what the person chose after the relation between the descriptions was made clear. The frame gives the theory of choice a further object to explain. It does not license an observer to declare every inconvenient preference a product of manipulation.⁵

The frame therefore reaches the place where the volume's descriptive and evaluative questions meet. A theory may explain why one presentation makes an option more attractive and still leave open whether that presentation helps the chooser live better. Coherence can show that two judgments cannot both express the same ordering of identical consequences; it cannot always identify which judgment ought to survive. The chooser's considered reasons, the experience that follows and the freedoms preserved can become relevant to that further assessment. Even if the frame were held stable and each person's ordering were fully known, another problem would remain. My ordering supplies no unit with which to weigh the strength of your want. The next plate, The Comparison, crosses that boundary: from the account by which a person values an outcome to the attempt to place two persons' valuations on one scale.