The contingent animal: does artificial innateness misrepresent behavioral development?

Abstract

While organisms are continually experiencing and interacting with their environments, the role and extent of experiences in behavioral development has been controversial. Some argue that adaptive behaviors are acquired through experiences, while others claim they are the result of innate programs that don’t require environmental input. Such controversies have historically occurred within animal behavior and psychology, but similar debates are emerging in the field of artificial intelligence. Here, the debate is centered on those who design experience-dependent systems that are trained to learn the statistical properties of “environmental” inputs, and those advocating the use of pre-packaged artificially “innate” responses tailored to prespecified inputs. Those favoring artificial innateness draw analogies with animal behavior to argue that innateness is necessary for the emergence of complex adaptive behavior. But does behavioral development in animals reflect the unfolding of innate programs? Here we highlight the widespread role of specifically causal experiences in the ontogeny of species-typical behaviors. All behaviors are an outcome of a chain of organism-environment transactions—called ontogenetic niches—that begin in the earliest periods of life. This challenges the notion that organisms come prepared with innate programs for behavior. We suggest that an artificial intelligence that matches the complexity of animal behavior should be based on principles of behavioral development, where experiences are necessary and specifically causal factors in the emergence of behavioral abilities.