# Bayesian modeling of mood-driven decision biases for predicting clinical outcome

> **NIH VA IK2** · VA SAN DIEGO HEALTHCARE SYSTEM · 2021 · —

## Abstract

Depressive symptoms, including negative self-focused emotions (e.g., sadness, guilt), make trauma-
exposed Veterans vulnerable to developing depressive pathology and substance use disorders (SUD). Both
conditions are in turn linked to persistent executive and functional impairments. A common marker of both
depression severity and addictive behavior is anhedonia, i.e., reduced ability to seek and experience pleasure/
healthy rewards, with evidence of reduced neural activity in the reward centers of the brain. Such reward
sensitivity alterations are thus likely to play a critical role in promoting depressive and addictive pathology in
trauma-exposed Veterans. A precise understanding of the neurocognitive mechanisms supporting reward-
based decision-making, and how mood is integrated into these processes, is therefore needed to: a)
understand the contribution of these risk factors to current and future psychopathology, b) help detect and
predict treatment needs/outcomes for Veterans with such complex clinical profiles. Bayesian models can offer
a powerful quantitative account of these intricate mechanisms by disentangling: a) learning processes
(prediction of reward likelihood in the environment), and b) action selection strategies (choosing an action
based on those predictions). This computational framework can help better delineate how low mood impacts
both the prediction of rewards, and the integration of these predictions into decision strategies. Moreover,
given that existing treatments for trauma-exposed Veterans focus on reducing anxiety symptoms, there is a
significant need for developing behavioral treatments that can more thoroughly target depressive symptoms
associated with low reward responsiveness, such as anhedonia and substance use. Computationally based
assessment of reward processing alterations may be particularly useful and timely for providing new treatment
targets and directions for improving such treatment outcomes for trauma-exposed Veterans.
 To address these questions, we propose to use computational modeling and neuroimaging to identify
precise affective neurocognitive predictors of psychopathology and behavioral treatment response in recently
deployed young (age 20-40) trauma-exposed Veterans. This project will use Bayesian modeling applied to the
analysis of reward-based decisions, with baseline dependent measures of brain circuit activity derived from
functional magnetic resonance imaging (fMRI), and experimental manipulation of mood, to delineate precisely
a) how sad mood may bias the learning or strategic adjustments guiding reward-based decisions in Veterans,
b) the degree to which these affect-driven computational biases relate to depression and problem substance
use symptoms in this population, and c) the degree to which these computational markers can predict
Veterans' response to a cognitive behavioral treatment targeting depressive symptoms associated with trauma,
including guilt and anhedonia. In other words, ca...

## Key facts

- **NIH application ID:** 10060726
- **Project number:** 5IK2CX001584-04
- **Recipient organization:** VA SAN DIEGO HEALTHCARE SYSTEM
- **Principal Investigator:** Katia Harle
- **Activity code:** IK2 (R01, R21, SBIR, etc.)
- **Funding institute:** VA
- **Fiscal year:** 2021
- **Award amount:** —
- **Award type:** 5
- **Project period:** 2017-10-01 → 2022-09-30

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/10060726

## Citation

> US National Institutes of Health, RePORTER application 10060726, Bayesian modeling of mood-driven decision biases for predicting clinical outcome (5IK2CX001584-04). Retrieved via AI Analytics 2026-05-24 from https://api.ai-analytics.org/grant/nih/10060726. Licensed CC0.

---

*[NIH grants dataset](/datasets/nih-grants) · CC0 1.0*