nlp/evidence_transparency.py

from typing import Tuple

def evaluate(response: str) -> Tuple[float, str]:
    """
    Evaluate Evidence & Transparency Fit (Metric 1) with optional heuristic.
    Community note: Provides a basic heuristic; primary scoring done by GPT-4o/Claude via evaluators.py.
    """
    transparency_keywords = ["limitation", "consult", "doctor", "evidence", "uncertainty"]
    score = sum(1 for kw in transparency_keywords if kw.lower() in response.lower()) / len(transparency_keywords) * 5
    comment = "Basic transparency detected." if score > 2.5 else "Lacks transparency elements."
    return float(score), comment