Thinking and Deciding

HEURISTICS AND BIASES IN PROBABILITY 153
still wanted to leave room for an intuitive attachment to a heuristic: “Even though
the probability of green coming up does not increase after several red — I always
have a feeling it will. Red is the safe bet but intuition will occasionally make me
choose green.... Iknowthatmyintuition has nothing to do with reality, but usually
they coincide.”
The availability heuristic
People often judge probability by thinking of examples. Consider the following
problem (Tversky and Kahneman, 1973, p. 211): Which is more likely, that a word
in English starts with the letter K or has K as its third letter? The sensible way to
try to answer that is to think of examples of words. This is the availability heuristic.
Most people find it easier to think of words that start with K than words with K as
their third letter, so they say that the former is more probable. Actually, the latter is
more probable. Tversky and Kahneman assert that we think otherwise because our
memory of spellings tends to be organized by initial letters.
Here is another example (Tversky and Kahneman, 1983, p. 295): What is the
probability that a seven-letter word randomly selected from a novel would end in ing?
What is the probability that such a word will have n as its sixth letter? When different
groups of subjects were asked these questions, the probability estimates were higher
for ing than for n. Of course, the former instances are included in the latter instances,
so the former instances cannot be more frequent. Here, the conjunction fallacy is
produced by the availability heuristic. The words ending in ing are more available
because the suffix ing is a better retrieval cue than a single letter; subjects think of
words more quickly when given this cue than when given the other.
Recall the study that Lichtenstein and her colleagues made (1978) of judgments
of the frequency with which different dangers caused death in the United States (Figure
6.1). We have already noted that subjects tended to overestimate the frequency
of low-probability events, and vice versa. Lichtenstein and her colleagues found,
however, that there was another important determinant of errors in estimation: the
frequency with which dangers were mentioned in newspapers. For example (taking
into account the effect of frequency itself), tornadoes and electrocutions — which
are almost always reported in the paper — are overestimated. Deaths resulting from
smallpox vaccinations and many common diseases such as asthma are underestimated;
these things are usually not reported.
Researchers have also demonstrated the effects of the availability heuristic in
subjects’ use of probability trees. When experts try to assess the probability of an
event’s occurring — let us say an engineer wants to predict the probability of a
catastrophe in a nuclear-power plant — they often use such tree diagrams as a way of
trying to ensure that they have considered all of the possibilities. By breaking such a
problem down in this way, we can obtain a more accurate probability assessment.
To take a more familiar example, however, consider the following list of categories
of the things that can prevent a car from starting: