User:Amit mpib/sandbox

The priority heuristic is a mathematical model for choosing one of two risky gambles. It is meant as a descriptive model of what people do, not as a prescriptive or normative model of what people should do.

Let us say gamble A is to receive 50 euro with certainty, (symbolized by A = (50, 1)) and gamble B is to receive 100 euro with probability ½ and 0 euro with probability ½ (symbolized by B = (100, ½; 0, ½)). The heuristic makes a choice between A and B by comparing their attributes one-by-one, as we explain below:

The first attribute is the minimum outcome of each gamble, which is 50 for A and 0 for B. If the difference between the two minimum outcomes is large enough, no other attribute is inspected, and a choice is made for the gamble with the higher minimum outcome. More precisely, the difference is large enough if it exceeds 10% of the maximum outcome across both gambles. In our example, this maximum equals 100; and a choice is made for A because 50 – 0 = 50 > 10 = (0.1) × 100.

Let us now say that the choice is between B = (100, ½; 0, ½) and C = (95, 0.9; 5, 0.1). Now the minimum outcomes do not differ enough because 5 – 0 = 5 < 10 = (0.1) × 100. So, the priority heuristic will inspect a second attribute, which is the probability of the minimum outcome. If these probabilities differ by more than 0.1, then no other attributes are looked up, and the gamble with the lower probability of minimum outcome is chosen. Here, a choice is made for C because 0.5 – 0.1 = 0.4 > 0.1.

Finally, let us now say that the choice is between B = (100, ½; 0, ½) and D = (95, 0.52; 10, 0.48). Now the probabilities of minimum outcomes do not differ enough because 0.5 – 0.48 = 0.02 < 0.1. So, the priority heuristic will inspect a third attribute, which is the maximum outcome. If these outcomes differ, then no other attributes are looked up, and the gamble with the higher maximum outcome is chosen. In our case, a choice is made for B because 100 > 95. If the maximum outcomes were equal, then a choice is made randomly.

The heuristic has been also extended for choices between gambles which have more than two possible outcomes, or negative outcomes. In sum, the priority heuristic is a sequential model, which does not use all available information, and could even make a choice based on just one attribute. The heuristic belongs to the family of lexicographic semi-orders.

It has been analytically shown that the priority heuristic logically implies a number of common violations of expected utility theory, such as common consequence effects, common ratio effects, the reflection effect, the four-fold pattern of risk attitude (Katsikopoulos & Gigerenzer, 2008). This fact helps explain the empirical finding (Brandstätter et al, 2006) that, in a set of risky choices studied in the literature, the priority heuristic overall predicted peoples’ choices more accurately than other models such as cumulative prospect theory.

References: