Evolved to adapt: A computational approach to animal innovation and creativity

The production of novel behavioral sequences that gives rise to animal innovation and creativity is one of the most intriguing aspects of behavioral evolution. Numerous studies have recently documented the abundance and diversity of innovative and creative behaviors between and within species, yet the ability to innovate or to act creatively has mainly been described and quantified as a measure of animals' cognitive ability without explicit reference to cognitive mechanisms that may account for these behaviors. Here we discuss the creative process from a computational point of view and suggest such a mechanistic framework. In light of recent research on human creativity, animal learning, and animal problem solving, we suggest that animal creativity is best understood as the production of context-appropriate novel behavioral sequences, which may be facilitated by the ability to learn the regularities in the environment and to represent them hierarchically, allowing for generalization. We present a cognitive framework that we recently developed, which employs domain-general mechanisms and has been used in the modeling of a range of sequential behaviors, from animal foraging to language acquisition, and apply it to behavioral innovation. In a series of simulations, we show how innovation and creative behavior can be produced by this learning mechanism, as it constructs a network representing the statistical regularities of the environment. We use the simulations to demonstrate the role of particular cognitive parameters in this process and to highlight the effects of the learning dynamics and individual experience on creativity