Three paradigms for viewing decision biases

Decisions can, and do, go wrong: a doctor misdiagnoses a patient's illness; a new product fails in the marketplace; a military commander mistakenly engages a civilian aircraft. Undesirable outcomes, however, do not necessarily imply faulty decision making; consider General Hod's decision to shoot down a Libyan airliner (described at the end of the last chapter). Even though the aircraft turned out not to be on a hostile mission, his conclusion that it was hostile might have been justified by the information he had at the time, by his efforts to gather further relevant data, and by the costs of a terrorist incident. It can also happen, of course, that a bad decision works out well. It is natural, then, for psychologists to look for a way of evaluating the decision itself, or the process that led to the decision, as distinct from its outcomes: to point, for example, at false prior beliefs, inappropriate priorities, shaky inferences from data, or even logical inconsistencies, rather than simply a bad outcome. A widely accepted research paradigm in psychology (e.g., Kahneman, Slovic, and Tversky, 1982) has taken Bayesian decision theory as the standard by which reasoning is to be judged, and has identified pervasive patterns of error, called biases, in laboratory performance. According to these researchers, unaided decision processes employ rules of thumb (or "heuristics") that under many (but not all) conditions lead to "severe and systematic errors" (Tversky and Kahneman, 1974). The bottom line of this research has been a rather pessimistic view of human reasoning. To illustrate, let us extrapolate some of the laboratory results on biases and their interpretation to a hypothetical physician: ! In assessing the probability that she will encounter cases of diseases A, B, and C among her patients, the physician may rely on the ease with which she can recall or imagine instances of each disease. This is the so-called "availability heuristic", postulated by Tversky and Kahneman (1973). Availability may be influenced by factors like the recency or salience of the physician's own experiences, which do not reflect the true relative frequencies of the diseases in the relevant population. ! In estimating a quantity, such as the required length of treatment for disease B, the doctor may first generate her best guess, then adjust it upwards and downwards to allow for uncertainty, e.g., 10 days plus or minus 2. This is the "anchoring and adjustment" heuristic. According …