PsyPlan

# AI-Driven Behavior Planning for Enhancing Psychosocial Health Outcomes This repository contains the official implementation of the framework proposed in the paper *"AI-Driven Behavior Planning for Enhancing Psychosocial Health Outcomes"*. The framework introduces a novel approach for interpretable, adaptive, and robust behavior planning using structured intentions and counterfactual reasoning in complex psychosocial environments. ## 🧠 Overview This project proposes an AI-driven planning system combining: - **Intent-Structured Planning Transformer (ISPT):** A goal-conditioned transformer that learns joint latent representations of belief states and symbolic intentions to generate interpretable behavioral trajectories. - **Hierarchical Counterfactual Planning (HCP):** A strategy that simulates multiple counterfactual futures to select robust action plans under uncertainty. ### Key Features - Interpretable and symbolic goal-aligned planning - Multimodal hierarchical encoding of belief-intention dynamics - Robust to uncertain and partially observable environments - Applications in digital mental health, human-AI collaboration, and social robotics ## 📁 Project Structure ```bash.├── models/ # ISPT and HCP model components├── datasets/ # CelebA, LSUN, FFHQ, MNIST datasets├── utils/ # Helper functions and shared modules├── configs/ # YAML configuration files├── scripts/ # Training and evaluation scripts├── checkpoints/ # Pretrained model weights├── notebooks/ # Visualization and analysis notebooks├── results/ # Experimental outputs and metrics└── README.md