Mania often manifests as intense goal pursuit, restlessness, and sometimes scattered behavior that is hard to brake. We built a computational model of learning from rewards and gave the learner a self-efficacy signal: a dynamic belief that “I can succeed here” which gets updated as the agent gets closer to its goals. In simulations, when that belief is overly sensitive to progress, the agent develops unrealistic reward expectations, acts more impatiently, chases bigger payoffs even when they are costly, and shows distractibility and eventually rigid value-driven choices that echo features of mania. Our model suggests a mechanistic pathway to bipolar disorder testable with behavioral tasks.
Angela Radulescu
Assistant Professor
My research focuses on the learning mechanisms underlying changes in mental health.