Social information impairs reward learning in depressive subjects: behavioral and computational characterization

Lou Safra, Coralie Chevallier, Stefano PalminteribioRxiv 2018 · DOI

TL;DR

This study reveals that individuals with higher levels of depressive symptoms exhibit impaired reward learning specifically in social contexts, a deficit not observed in non-social settings. Using computational modeling, the researchers identified this as a "negative audience effect," where social cues paradoxically hinder learning, offering new insights into the cognitive mechanisms of depression.

What Problem Does This Paper Address?

Depression is often characterized by social withdrawal and reduced sensitivity to rewards. While non-social aspects of reward processing in depression have been extensively studied, the cognitive mechanisms underlying decision-making in social contexts remain poorly understood. This gap is significant because social interaction is a fundamental aspect of human life, and understanding how depression affects it is crucial for diagnosis and treatment. This paper aims to bridge this gap by investigating how depressive symptoms interact with reward learning in social versus non-social environments.

Methods

The researchers employed an online behavioral task designed to assess reward learning in three conditions:

  • Private: Participants learned reward probabilities based solely on their own choices and outcomes.
  • Social-Choice: Participants saw the choices made by a virtual demonstrator but not their outcomes.
  • Social-Choice+Outcome: Participants observed both the demonstrator's choices and their outcomes.

They used two independent cohorts (N=50 each) for discovery and replication. Participants also completed questionnaires assessing depression and anxiety.

Crucially, the study utilized a computational reinforcement learning model (a modified version of Burke et al., 2014) to analyze the data. This model allowed them to:

  • Distinguish between primary social learning deficits (e.g., impaired imitation) and secondary social learning deficits (e.g., a reduction in individual learning when social information is present).
  • Estimate parameters like learning rates and choice temperatures for both private and social contexts.
  • Employ structural equation modeling to analyze the influence of depression scores on these model parameters, while accounting for their inter-correlations.

Datasets

The study used data from two independent cohorts of 100 American participants each, recruited via Amazon Mechanical Turk. These participants completed the experimental task and psychometric questionnaires. No new benchmark datasets were introduced.

Key Insights and Ideas

  • Social Context is Key: The detrimental effect of depressive symptoms on reward learning is not a general deficit but is specific to situations where social information is present.
  • Negative Audience Effect: The computational model suggests that social cues in depression don't necessarily impair the ability to learn from others' outcomes directly, but rather lead to a reduction in one's own learning rate when social information is available. This is termed a "negative audience effect."
  • Anxiety vs. Depression: While depression significantly impacted social learning, anxiety symptoms did not show a similar effect in this task, suggesting distinct neural or cognitive underpinnings for these conditions' social deficits.
  • Computational Psychiatry's Power: The study highlights how computational modeling can dissect complex cognitive processes and reveal nuanced effects of clinical symptoms that might be missed by behavioral analyses alone.

Key Results

  • Behavioral Findings: Participants with higher depressive symptoms showed significantly lower correct choice rates in the 'Social-Choice' condition compared to the 'Private' condition. This effect was robust across different experimental setups and contingency types.
  • Model-Based Findings:
    • Depression severity was negatively correlated with the social context-specific learning rate (αs), indicating that social information reduced learning in depressed individuals.
  • This learning rate reduction in social contexts was specific and not due to a general impairment in private learning.
  • Diagnostic Value: The difference in correct choice rates between social and private conditions, and the difference in learning rates, could accurately identify participants with "severe" depressive symptoms with good accuracy, sensitivity, and specificity.

Limitations and Future Work

  • Medication and Diagnosis: The study did not control for participants' actual diagnosis or medication status, which can influence decision-making. Future work should replicate these findings in carefully controlled populations.
  • Self-Reported Symptoms: The study relied on self-rated depression scores. Future research could explore whether these behavioral and computational measures can improve existing self-assessments for clinically diagnosed depression.
  • Mechanisms of Social Impairment: While the study identified a "negative audience effect," the precise cognitive and psychological mechanisms underlying this remain to be fully characterized. Further investigation into how depression affects instrumental performance in the presence of others is warranted.
  • Personalized Treatments: The authors suggest that tools assessing reward learning in both social and non-social contexts could be valuable for improving diagnosis and personalizing treatments for depressive disorders.

Summary updated

Abstract

Abstract Depression is characterized by a marked decrease in social interactions and blunted sensitivity to rewards. Surprisingly, despite the importance of social deficits in depression, non-social aspects have been disproportionally investigated. As a consequence, the cognitive mechanisms underlying atypical decision-making in social contexts in depression are poorly understood. In the present study, we investigate whether deficits in reward processing interact with the social context and how this interaction is affected by self-reported depression and anxiety symptoms. Two cohorts of subjects (discovery and replication sample: N = 50 each) took part in a task involving reward learning in a social context with different levels of social information (absent, partial and complete). Behavioral analyses revealed a specific detrimental effect of depressive symptoms – but not anxiety – on behavioral performance in the presence of social information, i.e. when participants were informed about the choices of another player. Model-based analyses further characterized the computational nature of this deficit as a negative audience effect, rather than a deficit in the way others’ choices and rewards are integrated in decision making. To conclude, our results shed light on the cognitive and computational mechanisms underlying the interaction between social cognition, reward learning and decision-making in depressive disorders.
BibTeX
@misc{safra2018social,
  author = {Lou Safra and Coralie Chevallier and Stefano Palminteri},
  title = {{Social information impairs reward learning in depressive subjects: behavioral and computational characterization}},
  year = {2018},
  doi = {10.1101/378281},
  howpublished = {bioRxiv},
}

Loading related content...

Figure