Support Theory

Support theory is a descriptive model of probability judgment that posits that judgments of probability are made based on descriptions of events rather than on events themselves.

Probability theory provides a normative framework for determining the probability of a combination of disjoint events. If two events A and B are exclusive, and have probabilities p(A) and p(B) of occurrence, the probability of one or the other occurring is exactly p(A) + p(B) = p(A or B). “A or B” is referred to as the disjunction of A and B. For example, if A is “The patient's heart rate is between 60 and 70 beats per minute” and B is “The patient's heart rate is between 70 and 80 beats per minute,” the disjunction might be expressed as “The patient's heart rate is between 60 and 70 beats per minute or between 70 and 80 beats per minute” (an explicit disjunction) or as “The patient's heart rate is between 60 and 80 beats per minute” (an implicit disjunction).

Support theory attempts to explain the observation that people often judge the probability of implicit disjunctions to be lower than the sum of the probabilities of the constituent events. This property is referred to as subadditivity and is contrasted with the normative probability theory model's additivity property (and with superadditivity, in which the probability of a disjunction is judged to be higher than the sum of the probabilities of the constituent events).

In support theory, descriptions of competing events are evaluated by assessing the relative support (s) for each description, characterized by a nonnegative real number associated with the strength of evidence for that description. Formally,

Support for a description may be a function of the strength of memories for events matching the description. As the name suggests, it may also be related to the ability to provide reasoned justifications for the described event. A stochastic extension of support theory, random support theory, extends the basic support theory model by representing support for events as a random variable. That is, people are assumed not to assign a fixed level of support to a given description but to sample support at random from a distribution of support. As a result, it is possible to speak of the variance and expectation of support associated with a given description. This enables random support theory to model the calibration of judgments (whether the objective frequencies of events are correctly predicted by their subjective probabilities).

Support theory assumes that the support for an implicit disjunction is less than or equal to the support for an equivalent explicit disjunction. It also holds that the support for an explicit disjunction is less than or equal to the sum of the support for the two descriptions. Formally, if A describes an implicit disjunction of B and C,

Two cognitive processes, unpacking and repacking, operate on descriptions of events and determine whether they are considered to be implicit or explicit disjunctions. Unpacking an implicit disjunction into its constituents increases the overall support, and thus the judged probability, of the event. A classic illustration is that judgments of “the likelihood of death by any vehicle accident” are often lower than judgments of “the likelihood of death by car accident, death by plane accident, death by bicycle [Page 1100]accident, or death by any other vehicle accident.” Moreover, the latter is often lower than the sum of the judgments of “the likelihood of death by car accident,” “the likelihood of death by plane accident,” “the likelihood of death by bicycle accident,” and “the likelihood of death by any other vehicle accident” when elicited individually.

...