# Availability

Availability is the collective term used to describe the following range of fallacies people make when judging probabilities.

•  Retrievability of instances
• ·Illusory Correlation
• ·Biases due to the Effectiveness of a search set
• ·Biases due to Imaginability

We consider these in turn:

Retrievability of instances

If you have just witnessed a car accident, your estimate of the subjective probability of having a car accident will temporarily rise. The event is salient, and so more available. Similarly, in an experiment by Tversky and Kahneman, subjects who were asked whether a list of well-known personalities contained more men than women, responded positively if the men in the list were better known than the women, whereas the numbers of each gender were in fact the same. In this case it was increased familiarity which made the male names more available and hence caused the error of judgement.

Illusory Correlation

Here the co-occurrence of two events is judged on the strength of their association. This is a very important fallacy because subjective probability assessors are often asked to estimate joint or conditional probabilities that depend on the correlations between two events.

Chapman and Chapman first described this bias after performing an experiment in which information about hypothetical mental patients was presented to a selection of naive judges. For each patient there was a clinical diagnosis and a drawing of a person that the patient had drawn. Later, the judges were asked to estimate the frequency of co-occurrence of a diagnosis such as paranoia for example, with a feature of the drawing, such as peculiar eyes. The result was a marked overestimation of the frequency of co-occurrence, an effect that was found to be extremely resistant to contradictory data. It also interfered with the detection of other relationships that were in fact present.

The explanation: If the associative bond between two events is very strong, then it is easy to conclude that the events have more frequently occurred together than in reality, they have.

Biases due to the effectiveness of a search set

Think of any English text including words of three letters or more. Is it more likely that a word picked at random and containing the letter "r" would start with the "r", or that "r" would be the third letter?