def generate_text_explanation(grouped_importance, prediction, confidence):
    """
    grouped_importance : dict from grouping.py
        Example:
        {
            "Semantic": 0.92,
            "Stylometry": 0.21,
            "Statistical": 0.08
        }

    prediction : "AI" or "Human"
    confidence : float
    """

    # Sort groups by importance
    sorted_groups = sorted(grouped_importance.items(), key=lambda x: -x[1])

    explanation_lines = []

    explanation_lines.append(
        f"The model predicts the code is {prediction} with confidence {round(confidence,4)}."
    )

    explanation_lines.append("Key contributing factors:")

    for group, score in sorted_groups:

        if group == "Semantic":
            explanation_lines.append(
                f"• Semantic patterns strongly influenced the decision, indicating structural similarity with {'AI‑generated' if prediction=='AI' else 'human‑written'} code."
            )

        elif group == "Stylometry":
            explanation_lines.append(
                f"• Stylometric characteristics such as identifier usage and coding style contributed to the prediction."
            )

        elif group == "Statistical":
            explanation_lines.append(
                f"• Token‑level statistical patterns (entropy, repetition, vocabulary richness) affected the classification."
            )

        elif group == "AST":
            explanation_lines.append(
                f"• Abstract Syntax Tree structure and control flow patterns influenced the decision."
            )

        elif group == "Language":
            explanation_lines.append(
                f"• Programming language encoding provided contextual information for the classifier."
            )

    return "\n".join(explanation_lines)