Weighting Bias
Weighting bias gives too much or too little influence to available cues. The authors distinguish this from cases in which corrective information is missing: those judgments may still be biased, but the problem is not how the person weighted evidence they had.
THE IDEA
Biases
Because the target attribute and the heuristic attribute are different, the substitution of one for the other inevitably introduces systematic biases. In this chapter we are mostly concerned with weighting biases, which arise when cues available to the judge are given either too much or too little weight. Criteria for determining optimal weights can be drawn from several sources. In the classic lens model, the optimal weights associated with different cues are the regression weights that optimize the prediction of an external criterion, such as physical distance or the GPA that a college applicant will attain (Brunswik, 1943; Hammond, 1955). Our analysis of weighting biases applies to such cases, but it also extends to attributes for which no objective criterion is available, such as an individual’s overall happiness or the probability that a particular patient will survive surgery. Normative standards for these attributes must be drawn from the constraints of ordinary language, and they are often imprecise. For example, the conventional meaning of overall happiness does not specify how much weight ought to be given to various life domains. However, it certainly does require that substantial weight be given to every important domain of life, and that no weight at all be given to the current weather, or to the recent consumption of a cookie. Similar rules of common sense apply to judgments of probability. For example, the statement “John is more likely to survive a week than a month” is a true statement in ordinary usage, which implies a rule that people would wish their judgments to follow. Accordingly, neglect of duration in assessments of survival probabilities would be properly described as a weighting bias, even if there is no way to establish a normative probability for individual cases (Kahneman & Tversky, 1996).
In some judgmental tasks, information that could serve to supplement or correct the heuristic is not neglected or underweighted, but simply lacking. If asked to judge the relative frequency of words beginning with k or r (Tversky and Kahneman, 1973) or to compare the population of a familiar foreign city with one that is unfamiliar (Gigerenzer and Goldstein, 1996), respondents have little recourse but to base such judgments on ease of retrieval or recognition. The necessary reliance on these heuristic attributes renders such judgements susceptible to biasing factors (e.g., the amount of media coverage). However, unlike weighting biases, such biases of insufficient information cannot be described as errors of judgment, because there is no way to avoid them.
When to use it
Human resources: Evaluating job candidates by giving appropriate weight to relevant experience and skills rather than interview charisma.
Medical diagnosis: Assessing patient prognosis by properly weighting base rates, symptoms, and test results without overemphasizing vivid cases.
Investment: Making financial estimates by weighting historical data and fundamentals rather than recent news or personal anecdotes.
Everyday life: Judging a friend's well-being by considering all important life domains, not just a recent positive or negative event.
FROM THEORY TO EVERYDAY LIFE
In practice
Overall Happiness Judgment
When judging someone's overall happiness, common sense requires substantial weight be given to important life domains and no weight to trivial factors like current weather or recent consumption of a cookie. Violating this rule constitutes a weighting bias.
Survival Probability Duration Neglect
In assessing a patient's survival probability, neglecting the duration (e.g., saying a patient is more likely to survive a month than a week) is a weighting bias, as ordinary language implies a rule that survival probability decreases over time.
GPA Prediction
In the lens model, optimal weights for predicting GPA are regression weights. A judge might overweight a single impressive interview or underweight past academic performance, introducing bias.
A MOMENT TO REFLECT
Put it to work.
Think of a recent judgment you made where an irrelevant factor (like mood or weather) might have influenced your evaluation of a person's overall happiness or quality of life. How could you adjust your weighting?
In medical or financial predictions, how can you determine the appropriate weights for different cues to avoid duration neglect or similar biases?
Can you identify a situation where a weighting bias led to a poor decision, and what normative standard was violated?
Often confused with
References
- The lens model of perception — Brunswik, E. (1943)
- The lens model of judgment — Hammond, K. R. (1955)
- On the reality of cognitive illusions — Kahneman, D. & Tversky, A. (1996)
- Availability: A heuristic for judging frequency and probability — Tversky, A. & Kahneman, D. (1973)
- Reasoning the fast and frugal way: Models of bounded rationality — Gigerenzer, G. & Goldstein, D. G. (1996)
CONTINUE THE CONVERSATION
Source texts
- Heuristics and Biases The Psychology of Intuitive Judgment (Thomas Gilovich, Dale Griffin, Daniel Kahneman)