Tony2025CRRSGIMD

Title

Can Risk and Reward Sensitivity to Games Inform about Mood Dynamics and Upcoming Episodes in Bipolar Disorder?
View PDF | Save PDF

Authors

Lazar Tony, Liam Mason, Pragathi Priyadharsini Balasubramani

Abstract

Lack of information about the upcoming mood episode often hinders the prescription of mood-targeted medication in cases that experience extreme mood swings like in Bipolar disorder between depression and mania. Computational models are yet to seamlessly integrate into the clinical practices, especially for characterizing the phase, severity, and trajectory of mood oscillation dynamics, and personally intervening in people suffering from mood disorders and in particular-bipolar disorder. For our study, we used a dataset collected through the Happiness project initiative at University College London, that prompted the subjects to play two alternate choice task paradigm daily, one option having a higher probability of fetching a reward and the other lower. Mood changes were induced by a wheel of fortune presented in the middle of the game, giving the participants chance for a jackpot or a huge loss. Additionally, the subjects were also asked to rate their Elated, Irritable, Energetic, Sad, Anxious, Angry moods daily. Further, they were also asked to fill in information about their depression, mania symptoms a few times over 2 months. We specifically asked a few questions in our study: 1) Can games be sensitive to mood induction, and how does it inform bipolar disorder? 2) Can a model of subjective reward and risk sensitivity of subjects, along with their ecological momentary input and clinical history, predict the bipolar status of a subject and their precise mood disorder severity? 3) Can the predicted mania and depression severity inform about mood swings in future? Our results broadly suggest that change of happiness reported by Bipolar subjects in a mood induction game significantly differed to that of healthy controls, and the bipolar status along with mood swings for the next 7 days from any time point can be reliably predicted using a combination of extended risk based decision making model of the game and machine learning, statistical models.

KeyPhrases

Mood oscillations, computational model, reward sensitivity, risk sensitivity, bipolar disorder, predictive utility.

Dates

Created 2025-05-01, presented 2025-06-03, published 2025-07-18.

Citation

Brainiacs Journal 2025 Volume 6 Issue 1 Edoc S670B3834
DOI: 10.48085/S670B3834
PDP: Nexus/Brainiacs/Tony2025CRRSGIMD
URL: BrainiacsJournal.org/arc/pub/Tony2025CRRSGIMD