Using Neural Style Transfer to study the evolution of animal signal design: A case study in an ornamented fish

Yseult Hejja-Brichard,Kara Million,Julien P Renoult,Tamra C Mendelson

DOI: https://doi.org/10.1101/2023.03.13.532060

2024-11-03

Abstract:The sensory drive hypothesis of animal signal evolution describes how animal communication signals and preferences evolve as adaptations to local environments. While classical approaches to testing this hypothesis often focus on preference for one aspect of a signal, deep learning techniques like generative models can create and manipulate stimuli without targeting a specific feature. Here, we used an artificial intelligence technique called neural style transfer to experimentally test preferences for color patterns in a fish. Findings in empirical aesthetics show that humans tend to prefer images with the visual statistics of the environment because the visual system is adapted to process them efficiently, making those images easier to process. Whether this is the case in other species remains to be tested. We therefore manipulated how similar or dissimilar male body patterns were to their habitats using the Neural Style Transfer (NST) algorithm. We predicted that males whose body patterns are more similar to their native habitats will be preferred by conspecifics. Our findings suggest that both males and females are sensitive to habitat congruence in their preferences, but to different extents, requiring additional investigation. Nonetheless, this study demonstrates the potential of artificial intelligence for testing hypotheses about animal communication signals.

Animal Behavior and Cognition

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to test fish's preference for color patterns by using neural style transfer technology, specifically the preference of Rainbow Darter (*Etheostoma caeruleum*) for color patterns that match their natural habitats. The research is based on the sensory drive hypothesis, that is, animals' communication signals and preferences evolve as a result of adaptation to the local environment. The researchers used the Neural Style Transfer (NST) algorithm to manipulate the body patterns of male Rainbow Darter, making them closer to or farther from the characteristics of their habitats, in order to test whether conspecifics prefer males whose body patterns are more similar to their natural habitats.

The core problem of the research is to verify the processing bias hypothesis, that is, signal designs that mimic the spatial statistical properties of natural scenes are more easily processed efficiently by the brain and may therefore be more popular among conspecifics. Through experiments, the researchers hope to understand whether such a preference exists and whether there are differences in the manifestation of this preference among different genders. In addition, the research also explores whether background matching itself will affect an individual's preference, that is, whether individuals prefer patterns with a high degree of background matching, whether in natural habitats or heterogeneous habitats.