A multi-dimensional transfer learning framework for studying reward-guided behaviors across species

Reward-guided behaviors, essential for survival and adaptation, exhibit both conserved and species-specific features across humans and other animals. Translating findings across species is often hindered by limited cross-species reliability in measurements and uncertain translational validity regarding functional or mechanistic relevance. This Perspective proposes a multi-dimensional transfer learning framework that integrates artificial intelligence (AI) to enhance cross-species research of reward-guided behaviors. By leveraging AI techniques, our framework connects behavioral neuroscience insights from animal models, especially land-based mammals, with functional outcomes in humans, enabling concept- and parameter-level transfer to identify universal principles, clarify mechanisms and optimize experimental paradigms. Using example expression components ofbehaviors, including locomotion trajectories and facial expressions, we highlight how multi-dimensional transfer learning can reveal conserved neural circuits while accounting for species-specific variations and contextual dynamics. This AI-powered framework offers a promising path to deepen our understanding of reward-guided behaviors and their relevance to mental health disorders.