Risky Choice Vs’ Constrained Rate Maximisation: 3 models, 3 species, 3 means

AW, J.; MONTEIRO, T.; VASCONCELLOS, M.; REBOREDA, J.C.; CHAPPEL, J.; KACELNIK, A.

Perth

Congreso; 13th International Society of Behavioral Ecology Congress; 2010

It is a matter of conjecture and empirical test whether natural selection has shaped mechanisms adapted specifically to handle differences in environmental variance (as assumed by Risk Sensitivity Theory) or if the effects of variance follow from mechanisms evolved to maximise expected payoff but subject to computational non-linearities. We study risky choice with European starlings (Sturnus vulgaris), magpies (Pica pica) and Shiny cowbirds (Molothrus bonariensis). Our subjects faced choices between one cue predicting fixed delays to reward, and another predicting either of two equiprobable outcomes. We identify the “certainty equivalent” or point of indifference between the poor and rich variable outcomes and compare them to the predictions of several putative choice mechanisms. The certainty equivalent should be at the arithmetic mean of the variable outcomes under unconstrained rate maximisation, at the harmonic mean under rate maximisation computing only the times between action and outcome (Associative learning, Expectation of the Ratios), and at the geometric mean under Weber’s Law (Scalar Utility Theory). Results across the three species clearly reject unconstrained maximisation, and fall in between the predictions of the last two models. Given these (and other) results it is unparsimonious to assume the presence of specific adaptations for risky choice.