Impaired adaptation of learning to contingency volatility in internalizing psychopathology

Chris Gagne, Ondrej Zika, Peter Dayan, Sonia J. BishopeLife 2020 · DOI

TL;DR

This study reveals that a core deficit in adapting to changing environments, common to both anxiety and depression, underlies impaired learning in individuals with internalizing psychopathology. This difficulty in adjusting learning rates to environmental volatility is domain-general, affecting both reward and punishment learning, and is particularly pronounced when learning from positive prediction errors.

What Problem Does This Paper Address?

Many real-world situations involve uncertainty, where the relationship between our actions and their outcomes isn't fixed. This "contingency volatility" means we need to constantly adapt our learning strategies. While intolerance of uncertainty is a known feature of anxiety and depression, the underlying cognitive mechanisms are not fully understood. This paper investigates whether a specific difficulty in adapting learning to changing probabilities (second-order uncertainty) is a transdiagnostic marker of internalizing psychopathology, and if this deficit is specific to certain types of outcomes (reward vs. punishment) or prediction errors (better vs. worse than expected).

Methods

The researchers employed a sophisticated approach combining computational modeling and statistical analysis.

  • Probabilistic Decision-Making Task: Participants engaged in a task where they learned to associate shapes with outcomes (reward or shock/loss) under conditions of stable or volatile contingencies.
  • Hierarchical Bayesian Modeling: A computational model was used to estimate individual participants' learning rates and other decision-making parameters. This model was hierarchical, allowing for population-level inferences about how these parameters varied with internalizing symptoms.
  • Bifactor Analysis: To disentangle symptom variance, a bifactor model was applied to questionnaire data. This separated a general "internalizing" factor (common to anxiety and depression) from specific anxiety and depression factors.
  • Model Comparison: Multiple computational models were compared using Pareto Smoothed Importance Sampling Leave-One-Out Cross-Validation (PSIS-LOO) to identify the best-fitting model of participants' behavior.

Datasets

The study utilized data from two experiments:

  • Experiment 1: Conducted in a lab setting with participants including those diagnosed with Major Depressive Disorder (MDD) or Generalized Anxiety Disorder (GAD), healthy controls, and a community sample.
  • Experiment 2: Conducted online via Amazon Mechanical Turk with a larger, more diverse sample.

Key Insights and Ideas

  • Shared Deficit: The core finding is that impaired adaptation to environmental volatility is not specific to anxiety or depression but is a shared deficit linked to the common underlying factor of internalizing psychopathology. This suggests a fundamental issue in how individuals with these conditions process and respond to changing environmental statistics.
  • Domain Generality: The learning deficit is not confined to a specific domain (e.g., only punishment learning). It holds across both rewarding and aversive outcomes, highlighting a broad impact on decision-making under uncertainty.
  • Asymmetric Learning: Individuals with lower levels of general internalizing symptoms show a distinct pattern: they boost their learning rate more significantly when outcomes are better than expected in volatile environments. This "positive prediction error boost" is absent in individuals with higher internalizing symptoms, suggesting a failure to capitalize on positive learning opportunities in uncertain situations.
  • Clinical Implications: The findings suggest that interventions targeting the ability to distinguish between different sources of uncertainty (noise vs. volatility) and to adjust decision-making accordingly could be beneficial for a broad range of individuals with internalizing disorders.

Key Results

  • Elevated scores on the general internalizing factor (common to anxiety and depression) were significantly linked to reduced adjustment of learning rate to contingency volatility.
  • This relationship was observed across both reward gain and aversive (shock/loss) tasks, indicating domain generality.
  • The effect was particularly pronounced in learning from better-than-expected outcomes in volatile environments. Individuals with low general factor scores showed a significant boost in learning rate under these conditions, which was absent in those with high general factor scores.
  • No significant relationships were found between anxiety- or depression-specific symptom factors and the observed learning deficits.

Limitations and Future Work

  • Generalizability to Other Psychopathologies: The study focused on internalizing psychopathology. Future work could investigate whether this deficit extends to externalizing or psychotic symptomologies.
  • Intervention Development: While the findings suggest potential targets for intervention, further research is needed to develop and test specific cognitive interventions aimed at improving decision-making under uncertainty.
  • Specific Anxiety Symptom Measures: While the study used a bifactor model, the authors note that the anxiety-specific factor's correlation with certain measures (like PSWQ) can vary, and future work could explore these nuances further.

Summary updated

Abstract

Using a contingency volatility manipulation, we tested the hypothesis that difficulty adapting probabilistic decision-making to second-order uncertainty might reflect a core deficit that cuts across anxiety and depression and holds regardless of whether outcomes are aversive or involve reward gain or loss. We used bifactor modeling of internalizing symptoms to separate symptom variance common to both anxiety and depression from that unique to each. Across two experiments, we modeled performance on a probabilistic decision-making under volatility task using a hierarchical Bayesian framework. Elevated scores on the common internalizing factor, with high loadings across anxiety and depression items, were linked to impoverished adjustment of learning to volatility regardless of whether outcomes involved reward gain, electrical stimulation, or reward loss. In particular, high common factor scores were linked to dampened learning following better-than-expected outcomes in volatile environments. No such relationships were observed for anxiety- or depression-specific symptom factors.
BibTeX
@inproceedings{gagne2020impaired,
  author = {Chris Gagne and Ondrej Zika and Peter Dayan and Sonia J. Bishop},
  title = {{Impaired adaptation of learning to contingency volatility in internalizing psychopathology}},
  booktitle = {eLife},
  year = {2020},
  doi = {10.7554/elife.61387},
}
Figure