ABS 2023
Search
Sensory drive in the era of artificial intelligence: new tools for new experiments
Yseult Hejja-Brichard1,2, Kara Million1,3, Iain Moodie2,4, Julien P Renoult2, Tamra C Mendelson1. 1University of Maryland, Baltimore County, Baltimore, MD, United States; 2Centre d'Ecologie Fonctionnelle et Evolutive, Montpellier, , France; 3University of North Alabama, Florence, AL, United States; 4Lund University, Lund, , Sweden

Sensory drive describes how animal communication signals and preferences evolve as adaptations to local environments. Classical approaches to testing this hypothesis focus on preference for one component or feature of a signal, such as color. We use artificial intelligence (AI) to overcome the limits of a one-trait paradigm. With three unique studies conducted in Etheostoma, a diverse genus of freshwater fish, we demonstrate (1) how AI can generate new holistic phenotypes without focusing on a single feature, (2) how AI algorithms can be used to determine how similar two stimuli appear to the eye of Etheostoma, and (3) how artificial neural networks can represent the brain as an agent of selection. These studies represent three different applications of AI that test predictions of sensory drive and “processing bias,” asking whether sexual signals in Etheosoma mimic the underlying patterns of their habitat and whether the fish prefer these patterns. We illustrate how AI can be leveraged to test predictions of sensory drive further while overcoming some of its limitations, here the one-trait approach, thereby demonstrating its generalisability.