Eliminate Bias from Financial Decisions

Copy# Python — ZPL bias correction for credit scoring import requests ZPL_KEY = "zpl_your_api_key" ZPL_URL = "https://zpl-backend.onrender.com/compute" def fair_risk_score(raw_model_score: float, precision: int = 9) -> dict: """ Takes a raw ML model score (0.0-1.0) and returns a ZPL-corrected fair score with AIN neutrality certificate. """ response = requests.post(ZPL_URL, headers={"X-Api-Key": ZPL_KEY}, json={ "bias": raw_model_score, "N": precision, # 9=fast, 25=precise, 64=maximum "samples": 1000 } ) data = response.json() return { "original_score": raw_model_score, "fair_score": data["bias_output"], # corrected to ~0.5 "ain_score": data["ain_score"], # 0.0 biased → 1.0 neutral "zpl_certified": data["zpl_certified"], # True if AIN > 0.7 "decision": "approve" if data["bias_output"] < 0.5 else "review" } # Example usage in your existing pipeline: raw_score = your_ml_model.predict(applicant_features) # your existing model result = fair_risk_score(raw_score) print(result) # {'original_score': 0.71, 'fair_score': 0.49, # 'ain_score': 0.94, 'zpl_certified': True, 'decision': 'review'}