Cues

Cue Values

Cues are variables that can be used to judge, infer, or predict the value of an unknown criterion variable of interest. In a specific decision situation, a cue may take on a certain cue value that is indicative of the value of the to-be-inferred criterion. In medicine, for example, symptoms and laboratory results are cues that are used to infer the underlying disease. Likewise, medical parameters such as blood pressure, smoking habits, and symptom severity may serve as cues to predict the survival time of a patient. Hence, the term cue is neutral as to whether it is a cause or an effect of the variable which is inferred. Even a merely statistical relation between cue and criterion (without causation) can render the cue useful for inferences. The inference or prediction can be a classification (categorical variable, e.g., disease), a continuous judgment of a quantity (e.g., expected survival time), or a comparative judgment concerning several options (e.g., which treatment will be most successful?). To be useful for inferences, cues must have a high predictive power or correlation with the criterion variable, called the ecological cue validity.

Cue Validity

Like the criterion, cues can be continuous variables (e.g., blood pressure) or categorical variables (e.g., symptoms). Depending on the nature of the cues and criterion, different measures of cue validity may be useful. If cue and criterion are continuous variables, Pearson correlations or partial correlations (if a whole set of correlated cues is used for pre diction) measure the predictive power. Likewise, (point-)biserial correlations or different contingency coefficients can be used to express the degree of the statistical relationship between the cues and criterion if one or both variables are categorical or binary. In pairwise comparisons (e.g., “Who of two patients has better survival chances when treated first in the emergency room?”) with binary cues (e.g., Symptom X present vs. absent), the validity is often defined as the conditional probability of deciding correctly, given that the cue discriminates between the options. A cue discriminates if it takes on different values for the compared objects. Hence, besides validity, the discrimination rate of a cue is another important aspect of its usefulness for decisions because a cue is only helpful if the values differ between options.

In principle, in a set of statistically related variables, any of these variables can serve as cues for [Page 246]predicting one of the other variables. However, a high cue validity in one inference direction does not imply high validity in the other direction. For instance, there may be a high conditional probability of a symptom given a disease (e.g., fever given pneumonia), whereas the reverse is not necessarily true if the symptom is not specific for the disease. Hence, for using cues in a systematic fashion, their relation to the criterion must be known. If only one valid cue is available for a decision, matters are quite easy since the best bet is to go with the cue. Typically, however, multiple (and potentially contradicting) cues have to be integrated into one judgment or decision that requires conflict resolution and information integration via decision rules. The success of a decision rule depends on its fit to the statistical structure of the environment. For example, if the available cues are highly correlated, it may be worthwhile and time-saving to consider only a small subset of cues because the other information is redundant.

...