Reconciling variability and optimal behaviour using multiple criteria in optimization models

A major objective in behavioural and evolutionary ecology is to understand how animals make decisions in complex environments. Examinations of animal behaviour typically use optimization models to predict the choices animals ought to make. The performance of animals under specific conditions is then compared against the predicted optimal strategy. This optimization approach has come into question because model predictions often do not match animal behaviour exactly. This has led to serious scepticism about the ability of animals to exhibit optimal behaviour in complex environments. We show that conventional approaches that compare observed animal behaviour with single optimal values may bias the way we view real-world variation in animal performance. Considerable insight into the abilities of animals to make optimal decisions can be gained by interpreting why variability in performance exists. We introduce a new theoretical framework, called multi-objective optimization, which allows us to examine decision-making in complex environments and interpret the meaning of variability in animal performance. A multi-objective approach defines the set of efficient choices animals may make in attempting to reach compromises among multiple conflicting demands. In a multi-objective framework, we may see variation in animal choices, but, unlike single-objective optimizations where there is one best solution, this variation may represent a range of adaptive compromises to conflicting objectives. An important feature of this approach is that, within the set of efficient alternatives, no choice can be considered to yield higher fitness, a priori, than any other choice. Thus, variability and optimal behaviour may be entirely consistent. We illustrate our point using selected examples from foraging theory where there is already an optimization program in place.