Biases Relating to the Representativeness Heuristic

Biases associated with the representativeness heuristic can and frequently do distort decision makers’ perceptions of probability. We rely on consistency and superficial resemblances with stereotypic notions that we all carry around in our heads. Therefore, certain types of occurrences appear to be more or less likely in our perceptions than they are in reality. Bazerman and Moore identify five such biasing influences: “insensitivity to base rates, insensitivity to sample size, misconceptions of chance, regression to the mean, and the conjunction fallacy.” Following is an illustration for each type in Table 2.1:

Biases on Perceptions of Probability

Insensitivity to Base Rates	A person starts a new business because of a success story in a related area. The individual feels the description of the other business founder matches his or her personality (both are intelligent, committed, and go-getters). Without due consideration of failure rates among start-ups in general, the person launches a business and later fails. The stereotype for what contributes to success in the example is overpowering the judgment of likelihood and obscures the relevance, as well as utility, of other potentially helpful information.
Insensitivity to Sample Size	A store manager receives a complaint about a given employee on three consecutive days and decides that something has to be done about said employee because “there seems to be a pattern emerging” even though over, say, a 30-day period, the likelihood of a “pattern” of upsetting customers might be quite small. The smaller the sample on which one bases a probability estimate, the more likely we are to see systematic events that are, in fact, random and well within the realm of chance.
Misconceptions of Chance	A gambler who is losing badly decides to continue gambling because his/her odds of winning “just have to improve.” Chance is not self-correcting. For instance, the odds of coming up with heads in flipping a coin on any given occasion are 50% each time. Those odds do not change simply because one might flip tails five or six times in a row.
Regression to the Mean	A firm fires a former top salesperson because his/her figures for the last quarter are down, and the person must, therefore, “be slipping.” However, it gives a bonus to another person whose performance had been repeatedly poor, but is “surprisingly good” because he or she recently improved. What those involved fail to realize is that extreme deviations from the average tend not to remain at the extremes over time, but instead revert toward the average, as when a basketball player who typically scores, say, 5 to 10 points a game seems to catch on fire and scores 20 points in one contest. It would be much too soon to begin celebrating the “dramatic change” in the player’s performance.
The Conjunction Fallacy	A champion athlete receives an offer to take a position as a coach on the assumption that being good in athletic performance makes it likely that s/he will also be good in coaching. The odds that one can be good in both coaching and athletic performance are worse than the odds that one will be good in either activity alone. Joint probabilities cannot exceed the value of what is the most probable of the items being conjoined. For instance, if the odds that an employee at a firm will be happy working there are 80% and the odds that the same person will be productive are 60%, the odds that he or she will be both happy and productive are only 48%.  Joint probability is the product of multiplying one probability by the other--in this case .8 x .6.

Table 2.1. Types of Representativeness Biases

To learn more about the representativeness heuristic, visit: Changing Minds' Representativeness Heuristic.