Lower confidence and increased error sensitivity in OCD patients while learning under volatility

Decision-making relies on beliefs and confidence in those beliefs. High confidence leads to minimal belief updates upon encountering new information, while low confidence motivates seeking additional evidence. Obsessive-compulsive disorder (OCD) is characterized by intrusive obsessions and compulsions, and is often associated with lower confidence compared to healthy controls. This lower confidence might contribute to the doubts and uncertainty driving compulsive behaviors. Previous research using a predictive inference task yielded inconsistent results regarding the interaction between confidence and behavior in OCD patients during volatile learning. Some studies showed a decoupling between confidence and behavior, correlating with OCD symptoms, suggesting that compulsive behavior stems from inaccurate confidence judgments in volitional actions. Others showed excessive action in response to small prediction errors, indicating increased error sensitivity. Studies using general population samples with high obsessive-compulsive (OC) symptoms showed conflicting results, with some showing increased confidence, and others showing decoupling between action and confidence, depending on the sample characteristics and methodologies. The current study aimed to address these inconsistencies by investigating action, confidence, and their coupling in a medication-free OCD sample without comorbid diagnoses, comparing them to healthy controls and a matched sample of high and low compulsive individuals from the general population using the same predictive inference paradigm.

Existing literature on confidence and behavior in OCD during volatile learning is inconsistent. Some studies using predictive inference tasks reported a decoupling between confidence and behavior in OCD patients, linked to symptom severity, while others found increased error sensitivity, with excessive adjustments to minor prediction errors. Research on general population samples with high OC symptoms showed contradictory findings, some indicating increased confidence and others a decoupling between action and confidence. These inconsistencies highlight the need for further investigation, particularly considering the influence of medication and comorbid diagnoses. A recent study challenged the assumption that highly compulsive individuals' metacognitive abilities mirror those of OCD patients, finding that their behavior is driven by different mechanisms.

This study included 38 OCD patients and 37 healthy controls (HCs), matched for age, gender, and education. Participants performed a web-based predictive inference task, predicting the landing location of a particle and rating their confidence. Action was operationalized as the absolute difference in bucket position between trials. Data from a previous study (Seow & Gillan, 2016) provided a comparison group of 76 high-compulsive (HComp) and 73 low-compulsive (LComp) individuals from the general population, matched to the HC and OCD groups based on OCD symptoms and demographics. Questionnaires assessed additional psychiatric disorders (MINI), OC symptoms (OCI-R), OCD symptom severity (Y-BOCS), and anxiety/depression symptoms (DASS, STAI, Zung's scale). Task-based exclusions were applied based on pre-registered criteria. Linear mixed-effects models analyzed group differences in action and confidence, and the strength of action-confidence coupling. Learning rates were calculated to assess how participants used prediction errors to update their actions. Model-based analyses compared participants' behavior to a Bayesian model. Correlations examined the relationship between task behavior and symptom severity.

Compared to HCs, OCD patients showed significantly lower confidence (β = −18.9, p < 0.001) but no differences in action. No group differences were found in the coupling between action and confidence. However, OCD patients displayed higher learning rates for small prediction errors. In the comparison between HComp and LComp individuals, HComp individuals exhibited higher confidence (β = -7.72, p = 0.030) than LComp individuals, and a weaker action-confidence coupling (β = -2.27, p = 0.033). In both OCD and HComp groups, confidence in HComp individuals correlated significantly with symptom severity (OCD: r = 0.52, p < 0.001; HComp: r = 0.50, p < 0.001), whereas no such correlation was observed in OCD patients for either confidence or action-confidence coupling. Model-based analyses showed no significant differences in how participants used Bayesian parameters to inform their actions and confidence judgments between the groups. Learning rates increased as a function of prediction error magnitude across all groups, but the effect was less pronounced in HComp individuals compared to LComp individuals, suggesting a less steep increase of learning rate in high compulsive individuals.

This study's findings challenge previous inconsistent results, demonstrating that medication-free OCD patients without comorbid diagnoses show lower confidence but no significant differences in action-confidence coupling compared to healthy controls. The increased error sensitivity in OCD patients, particularly for small prediction errors, may reflect hyperreactive error signaling, a known endophenotype of OCD. The contrast between OCD patients and highly compulsive individuals from the general population highlights the heterogeneity of obsessive-compulsive symptoms. OCD patients exhibited lower confidence, while highly compulsive individuals showed increased confidence, both correlating with symptom severity. These different (meta)cognitive profiles suggest distinct underlying mechanisms, potentially explained by different ways individuals use past experiences to inform future actions. Future research should investigate the ecological validity of the task and use more comprehensive assessments of OCD symptoms to better understand the impact of compulsive symptoms on daily life.

This study clarifies the relationship between action, confidence, and their coupling in OCD patients compared to healthy controls and a matched general population sample. OCD patients showed lower confidence and increased error sensitivity, while high compulsive individuals displayed increased confidence and decoupled action-confidence. These findings highlight the heterogeneity of obsessive-compulsive symptoms and emphasize the need for more research to identify distinct underlying mechanisms associated with different behavioral manifestations of obsessive-compulsive symptoms.

The online testing environment, although with extensive instructions, may limit the ecological validity. The HComp and LComp groups potentially had more experience in online research, potentially influencing their task performance. The model-based analyses, while consistent with previous studies, should be interpreted cautiously due to their lower prospective validity. Future research should consider tasks with greater ecological validity and more comprehensive symptom assessments. The exclusion of trials based on learning rate and prediction error magnitude necessitates careful interpretation of the results.