Decision Weights

A decision weight reflects a person's subjective interpretation of an objective probability. Almost all medical decisions involve probabilistic outcomes. For example, there is some chance that a treatment will cure a disease and some chance that the treatment will have a side effect. Data are often available to help patients and providers know the probability that an outcome, such as a serious side effect, will occur. When people face decisions involving uncertain outcomes, how do they use these probabilities?

Theories of rational decision making recommend using the exact value of the probability in evaluating a decision. For example, in expected utility theory, a rational decision maker should evaluate the overall worth of an option by (a) multiplying the probability of each possible outcome by the utility of that outcome and (b) summing the products across all possible outcomes. However, people making actual decisions do not use the real, or “objective,” probability when making decisions; the subjective sense of a given probability p is not necessarily the same as p. This phenomenon is analogous to the psychophysics of light perception, in which the brightness a person perceives does not have a 1:1 relationship with the actual luminous energy in the environment.

In the most well-known descriptive theory of decision making, prospect theory, the subjective sense of a probability is known as the decision weight corresponding to that probability, denoted by π. Understanding how a person uses objective probabilities in decision making requires knowledge [Page 362]of that person's decision weight function, which describes how probabilities are related to decision weights.

Figure 1 shows a typical decision weight function and illustrates some typical findings from research on decision weights.

First, people tend to overweight small probabilities. Because people have difficulty conceptualizing small probabilities, they translate them into decision weights that are greater than the actual probabilities. This finding might help explain why, for example, both patients and investigators overestimate the small chances of benefit and harm associated with participation in early-phase oncology trials.

Second, people tend to be less sensitive to the differences among probabilities near the middle of the probability scale. Theories of rational decision making state that changes in objective probabilities should make a difference to people. However, actual decision weight functions are relatively flat for intermediate objective probabilities. Thus, a patient might appear to disregard information about the probabilities of success or failure when those probabilities are in the intermediate range (e.g., p = .25 to .75). In fact, the patient might be attending to the probabilities presented but assigning them similar decision weights.

Third, the decision weight function is usually steepest as it approaches 0 and 1.00. People tend to prefer changes in probabilities that will result in a state of certainty, something known as the certainty effect. Consider a patient deciding between medical and surgical therapies for a heart condition. If the probabilities of success are .80 and .90, respectively, there is a .10-point difference between the treatments. Now imagine that the .10-point difference arises from the probabilities of .90 and 1.00. In expected utility theory, these two scenarios should not be different, because the difference between the options is .10 in both. Yet people do not typically experience these scenarios as equivalent. The decision weight function shows that this is the case because people assign greater weight to the elimination of uncertainty.

...