Individuals in nature frequently face decision problems where the information available to them is uncertain and their reproductive success depends on the outcome of their decisions. In these cases natural selection should be expected to favour individuals whose behavioural strategies yield the best reproductive payoffs. It is accepted that decision-makers in nature should evolve to behave as if they were Bayesian learners when making decisions on uncertain information (Marshall et al., 2013a,b; McNamara et al., 2006; McNamara and Houston, 1980; Tenenbaum et al., 2006). In this thesis different decision problems from nature are modelled in order to determine the optimal strategies that should be expected to evolve in response to different parameters of the environment, under the assumption that decision-makers should also evolve to make decisions as if they were Bayesian.
One model is proposed to determine the conditions under which inducible defences, a type of phenotypic plasticity, should be expected to evolve as a defensive mechanism against predators. The model is used to predict when plasticity is the evolutionarily optimal strategy, given the decision-maker’s inherited uncertainty regarding predation risk. The model assumes that this inherited uncertainty has been shaped by natural selection in prior generations of the decision-maker’s species to reflect the uncertainty exhibited by the predation risk in the environment. It is shown that when this inherited uncertainty is high enough (and thus when the uncertainty exhibited by the predation risk is also high enough) then plasticity is the optimal strategy.
A second model is presented in order to test the hypothesis that decision-makers evolve their Bayesian priors in response to variation in the environment. The results confirm the assumption made in the model described above that the decision-maker’s inherited assessment and uncertainty of the predation risk can be shaped by natural selection. Finally, a third model is presented to determine when decision-makers should be expected to evolve self-deception biases in situations of conflict over resources, such as food, against other decision-makers. This model tests the theory proposed by Trivers (2011), which states that the most evolutionarily successful deceivers are those who self-deceive first. This is because self-deceiving deceivers do not have to pay the physiological costs paid by deceivers who are aware of their deception. In the model presented in this thesis it is shown that self-deception is more likely to be the optimal strategy as the information the decision-maker has access to becomes more uncertain and as the contested resource becomes more valuable.
