12/In summary, we formalize the hypothesis that core symptoms in mania can emerge through sensitive self-efficacy beliefs by simulating a model-free RL agent in a sequential grid-world, and are in progress of testing our hypothesis with human behavioral data.
11/Our work makes several empirical predictions which we are currently testing in behavioral tasks.
10/This resembles some clinical characteristics of mania such as compulsivity and distractibility before goal attainment.
9/The peak of the learned value representation moved from the vicinity of the terminal state to more distant locations and continued until the learned best actions formed a closed loop around the states with peak values (attractor states).
8/We also simulated a possible symptom trajectory of mania emergence, showing that a stepwise increase in self-efficacy level was observed as self-efficacy level kept increasing until reaching an artificial limit.
7/This parallels clinical findings that individuals with vulnerability to hypomania are more likely to pursue more difficult goals and remain engaged longer as tasks become more difficult.
6/In a two-goal grid-world, agents with higher self-efficacy learning rates also exhibited stronger tendency to approach the more rewarding but also more costly goal.
5/This is in line with clinical observations that after small successes, individuals diagnosed with bipolar disorder (BD) exhibit a significant boost in confidence that extends beyond specific positive events to wider areas of their life.
4/We performed simulations in a sequential model-free grid-world environment and showed that agents with more sensitive self-efficacy beliefs develop optimistic overgeneralized future reward expectations.
3/In our dynamic self-efficacy model, the self-efficacy belief scales the highest possible future expected reward contingent on action, and updates according to reward prediction errors, which provide a continuous index of goal attainment.
2/How does goal-pursuit behavior become dysregulated in mania? In a reinforcement learning setting, we formalize the hypothesis that core symptoms of mania can emerge as a result of more sensitive self-efficacy beliefs.
1/Excited to share my first first-author paper with @angelaradulescu: ‘Dynamic self-efficacy as a computational mechanism of mania emergence’, on which I will present a poster this Friday 13:00 - 14:15 at #CogSci2024 escholarship.org/uc/item/78m0...
A short summary:
7/This parallels clinical findings that individuals with vulnerability to hypomania are more likely to pursue more difficult goals and remain engaged longer as tasks become more difficult.
6/In a two-goal grid-world, agents with higher self-efficacy learning rates also exhibited stronger tendency to approach the more rewarding but also more costly goal.
5/This is in line with clinical observations that after small successes, individuals diagnosed with bipolar disorder (BD) exhibit a significant boost in confidence that extends beyond specific positive events to wider areas of their life.
4/We performed simulations in a sequential model-free grid-world environment and showed that agents with more sensitive self-efficacy beliefs develop optimistic overgeneralized future reward expectations.
3/In our dynamic self-efficacy model, the self-efficacy belief scales the highest possible future expected reward contingent on action, and updates according to reward prediction errors, which provide a continuous index of goal attainment.