Supplementary Materials for: Agentic Explainable Artificial Intelligence (Agentic XAI) Approach To Explore Better Explanation

This dataset contains the supplementary materials for the manuscript "Agentic Explainable Artificial Intelligence (Agentic XAI) Approach To Explore Better Explanation."The repository documents the complete implementation of an iterative refinement process designed to generate actionable farmer recommendations from agricultural data.It includes all prompts, source code, visualization artifacts, and textual outputs across all refinement rounds.

Contents

"Figures" Folder Archives all visualization outputs and statistical artifacts generated throughout the iterative refinement process (Rounds 0–10). These files document the autonomous expansion of the system's analytical scope, progressing from fundamental bivariate relationships in early rounds to complex multi-way interactions, temporal pattern decompositions, and comprehensive correlation matrices in later iterations.

"Prompts" Folder Contains the three sequential prompts that control the agentic workflow, mapped to specific analysis stages:

Prompt 1 (Initial Interpretation): Used in Round 0 to request baseline recommendations based solely on SHAP visualizations.

Prompt 2 (Gap Analysis and Code Generation): Used in Rounds 1–10 to trigger self-reflection, instructing the model to identify analytical gaps and provide Python code for generating supplementary figures.

Prompt 3 (Synthesis and Update): Used in Rounds 1–10 to guide the synthesis of original insights with newly generated supplementary analyses for updated recommendations.

PythonCode.ipynb A Jupyter Notebook comprising the complete execution history, including the initial XAI analysis code and all supplementary Python scripts autonomously generated via Prompt 2 during Rounds 0–10. The notebook serves as evidence of escalating analytical complexity, demonstrating the system's ability to autonomously implement advanced statistical analyses and multi-dimensional data explorations.

"Recommendations" Folder Stores the textual recommendations for farmers generated at the baseline (Round 0, via Prompt 1) and throughout the refinement phase (Rounds 1–10, via Prompt 3). These documents illustrate the progressive enhancement of advice quality resulting from the agent's self-reflection and integration of supplementary analyses.

This organizational structure ensures full transparency and reproducibility of the Agentic XAI workflow. It traces the entire pipeline from initial explainability analysis to the final, farmer-oriented recommendations, while explicitly documenting the autonomous escalation of analytical complexity that characterizes the agentic refinement process.