Sequential choices using signal detection theory can reverse classical predictions. (3rd November 2018)

• Record Type: Journal Article
• Authors: McNamara, John M; Trimmer, Pete C
• Is Part Of: Behavioral ecology. Volume 30:Number 1(2019)
• Page Start: 16
• Page End: 19
• Publication Date: 2018-11-03
• Subjects: error management theory -- evolution -- sequential decision-making -- signal detection theory -- state-dependence -- ROC analysis; Animal behavior -- Periodicals; Behavior evolution -- Periodicals; Ecology -- Periodicals; Psychology, Comparative -- Periodicals; 591.5
• DOI: 10.1093/beheco/ary132
• Languages: English
• ISSNs: 1045-2249

Abstract

Abstract : Intuition and past theory suggest that if dangers become more common then prey should be less bold. But this prediction should be reversed in many biological scenarios. We provide a simple demonstration that the old predictions of signal detection theory should be reversed, even when the same decision-making threshold is used by the animal regardless of its current state of reserves. In other words, more predators result in prey being *more* bold! Abstract: The behavioural sciences have been strongly influenced for several decades by signal detection theory (also known as "error management theory" and "ROC analysis"). The theory provides simple logic about how individuals should behave in order to maximize their expected payoff in a single decision; for instance, the theory predicts that as the probability of a particular danger increases, individuals should be more prone to avoiding that danger. However, such findings need not hold in situations where more than 1 decision will be made, and recent work has shown that when behavior is allowed to depend on the energy reserves of the individual, the classical predictions can be reversed. Here, we use a simple analytic technique to show that even when the same behavioral rules are used at every reserve level, the predictions of classical signal detection theory (as though only a single decision will be made) are often reversed in a world where autocorrelation is high. The classical predictions are reversed because individuals will often make multiple decisions to achieve a goal, and the outcome of an action can influence the number of subsequent decisions that are required. The finding suggests that fundamental predictions should be altered across many of the behavioral sciences, not only in relation to signal detection theory, but also other single-decision models.