can we computationally define what makes individuals differ in how they approach uncertain environments?
Cognitive and computational neuroscience has made incredible progress in defining computational models and parameters that can influence reinforcement learning and decision making. However, these still rely on assumptions about what an agent is trying to accomplish (what goals they are optimizing for) and how they should accomplish it (what strategy makes sense in a particular environment). However, we know there is wide variation in individuals, and in the real-life situations they employ decision making. To fulfill the promise of computational psychiatry, these questions have to be explored across mechanisms which are major drivers of individual differences.
Our lab studies sex differences as a major biological factor tuning the function of numerous brain regions involved in reinforcement learning and decision making. Both male and female mice are able to learn to associate actions with outcomes, inhibit impulsive and habitual actions, and make adaptive decisions. However, we find there are substantial differences in the strategies male and female mice use to accomplish these tasks, and thus how effectively they meet their goals. By understanding how the multiple, dissociable mechanisms of sex differences influence decision making, we can begin to define which parameters and/or computational models are most sensitive to these factors, with the goal of computationally defining the axes of interindividual variability in this complex cognitive domain.
Recent data from our lab shows enhanced acquisition of a two-armed bandit task in female mice. More specifically, we find evidence that different strategies drive reward learning in male and female mice in the bandit task.