The-Metacognitive-Paradox-of-OCD

Loosen et al. (in preparation) The metacognitive paradox of OCD: confidence is globally reduced but shows increased sensitivity to local evidence

Overview

This repository contains the data analysis and model implementation for a study conducted at the University College London (UCL) and Yale School of Medicine. Participants played a novel intra- and extra-dimensional shift task with integrated confidence ratings.

This repository contains code in Python for the behaviorala analysis (1) and the implementation of a Bayesian Observer model to capture how confidence ratings should develop under a Bayes-optimal framework (2), given task information and decision evidence. The resulting parameters from this model are analyzed in relation to the group status (3; patient vs. control).

---

Required

Please download the data on OSF.

Content of the Repository

1. Data Analysis:

   • Exploring group-level differences in task performance and confidence ratings.
     • PreprocBehavAnalysis

2. Bayesian Observer Model:

   • BayesianObserver: Simulates confidence trajectories of a Bayes-optimal observer, who knows task rules and is exposed to the specific task run of a given participant.
   • BayesianObserver-BehavAnalysis: As to be run after the BayesianObserver and links resulting parameters to behavioral data.

---