Two Models of Predator Decision Making Under Uncertainty

Predators are faced with an uncertain world. The presence of anti-predator defences means that any potential prey item may actually be unpalatable, outright toxic, difficult to catch, or cause harm. In order to deal with this uncertainty regarding profitability, predators need strategies to make good decisions on what to attack based on the information about potential prey available to them. I develop two models of optimal decision making for predators. The first deals with generalizing from experience to novel prey types: I develop a Bayesian model framework that treats generalization as a process of learning about the distribution of prey in the environment, and apply it to a problem in generalization. The second deals with startle displays: I develop an extension of signal detection theory to cases where continued examination is possible, and apply it to predators faced with startle displays.