Neuronal mechanisms for nutrient-sensitive reinforcement learning in primates

Reinforcement learning (RL) is a dominant framework in neuroscience and artificial intelligence (AI) that formalises how agents learn from rewards. Nutrients, including fat and sugar, are critical, noninterchangeable reward components that must be separately processed by neural reward systems. Although direct neuronal implementations of key RL mechanisms, including reward prediction errors, have been discovered, how these mechanisms process biologically critical nutrient rewards remains unclear. Here, I combine monkey single-neuron recordings with recently developed nutrient-sensitive RL models to investigate how neurons in primate taste and reward systems process nutrients during RL. First, I will record single-neuron activity in orbitofrontal cortex, amygdala, anterior cingulate cortex, and insula while macaques learn and adapt their choices to changing visual-nutrient associations. These data will uncover how neurons encode learning and decision variables for nutrient rewards. Second, I will apply focused transcranial-ultrasound stimulation to these brain structures to test their causal roles in nutrient-sensitive RL. Third, I will determine how nutrients, as fundamental reward dimensions, shape the geometry of neural-population codes underlying nutrient-sensitive RL. The rare data and new concepts from this project have potential to deepen our understanding of neural reward systems, inspire AI innovations, and uncover vulnerabilities for maladaptive reward-processing in human mental disorders.