Peacemaking among inconsistent rationalities?

Kacelnik, Schuck-Paim and Pompilio (this volume, p. 377) show that rationality axioms from economics are neither necessary nor sufficient to guarantee that animal behavior is biologically adaptive. To illustrate that biological adaptiveness does not imply conformity with the consistency axioms of economics, Kacelnik et al describe animals that sensibly experiment with actions yielding sub-maximum levels of short-term energy intake to monitor their environments for change, leading to apparently intransitive patterns of choice that are nevertheless biologically adaptive. Invalidating the converse claim that economic rationality implies biological adaptiveness is Kacelnik et al’s example of female ruffs that are worse off when they conform to the constant-ratio rule, frequently interpreted as a normative consistency requirement of economic rationality. Together, the two examples demonstrate that axiomatic norms are both unnecessary and insufficient for determining whether a particular behavior is biologically adaptive. Additionally, Kacelnik et al call into question what has been reported in the animal behavior literature as preference reversals, such as risk attitudes among wild rufous hummingbirds or the food-hoarding propensities of grey jays. Kacelnik et al attribute apparent reversals to state-dependent fitness functions modulated by subtle differences in the training phase of animal experiments. For example, animals trained on menus that include a strictly dominated option will tend to have lower accumulated energy reserves and therefore exhibit systematically different patterns of choice––not because they fail to maximize, but because their training has induced systematically different nutritional states. Another possible explanation for preference reversals in animal studies with strictly dominated, or “decoy” options is that menus containing dominated items may convey valid information about future opportunities (Houston and McNamara, 1999). If menus are correlated through time, then menus with inferior options today predict scarcity in the future and imply a distinct optimal course of action, in violation of regularity assumptions that posit invariance with respect to the inclusion of strictly dominated alternatives. In environments with payoff structures that can be modeled as cooperative games, a family’s best response sometimes requires individual family members to behave suboptimally as part of a diversification strategy that reduces the risk of reproductive failure (Hutchinson, 1996). Futhermore, theoretical biologists have documented the fragility of expected fitness maximizing behaviour with respect to the assumption of stable environments. Once the model allows for shocks to the environment’s stochastic structure, simple behavior rules that are suboptimal (in terms of expected fitness) when viewed narrowly from the perspective of unchanging payoffs in a fixed environment may outperform rules based on maximazation within a static small world (Bookstaber and Langsam, 1985).