Here’s an updated README for the repository, including details about the additional advice_inversion.m script:

---

README

Note: This repository is currently under development.

Do not use this code in its current state. Features and functionality are incomplete and subject to change.

---

Model Fitting for Advice Task with Trust Learning

This repository contains MATLAB and Python scripts for modeling, fitting, and simulating behavior in the Advice Task. These models incorporate trust learning, decision-making with social advice, and multi-level parameter estimation.

---

Repository Files

MATLAB Scripts

1. main_advise.m

   • Primary script for running model fitting and simulations.
   • Configurable for subject-level or simulated data.
   • Outputs results in .mat and .csv formats.

2. Advice_fit_prolific.m

   • Fits Advice Task models to behavioral data collected from Prolific.
   • Processes task-specific trial information and estimates model parameters.
   • Outputs individual fits and summary statistics.

3. Simple_Advice_Model_CMG.m

   • Core script for the Advice Task model.
   • Simulates belief updates and action probabilities using task-specific parameters (eta, omega, p_right, etc.).
   • Supports various trust and learning dynamics.

4. advice_inversion.m

   • Performs model inversion using Variational Bayes.
   • Estimates posterior distributions and log evidence (free energy) for the model parameters.
   • Supports two core models: Simple_Advice_Model_CMG and Simple_Advice_Model_CMG_same_num_choices.

---

Python Scripts

1. runall_advise_fit.py
   • Automates model fitting across multiple subjects using Slurm.
   • Organizes results into directories and logs.
   • Configurable for batch processing and high-performance computing.

---

Key Features

• Simulation: Generate synthetic data for validation and testing.
• Model Inversion: Use Variational Bayes to estimate model parameters.
• Multi-Level Analysis: Handle individual and group-level behavior with hierarchical models.
• Dynamic Trust Learning: Models epistemic and pragmatic value computation with advice.

---

Usage

1. MATLAB Workflow:

   • Configure paths and settings in main_advise.m or Advice_fit_prolific.m.
   • Run the script to fit models or simulate data, saving outputs automatically.

2. Batch Processing with Python:

   • Update runall_advise_fit.py with subject lists and output directories.
   • Execute the script to submit jobs for large-scale model fitting.

---

Dependencies

• MATLAB: Required for core modeling and analysis.
• Python 3.x: Used for batch processing and Slurm integration.
• Slurm: Needed for high-performance job scheduling.