Counterfactual Explanations for an Insurance Pricing Model
Overview
What this challenge is about.
You receive a trained LightGBM regression model (premium in GBP), the feature schema (28 features, 14 mutable from the customer's side), and 500 sample quotes. Use DiCE (Diverse Counterfactual Explanations) to generate up to 3 diverse, actionable counterfactuals per quote that respect mutability and reasonable bounds (no 'be 22 years younger'). Build a strict JSON contract the web frontend will consume. Write a one-page copy guideline for the customer-experience writer and a 2-page methodology memo. Deliver a Python service skeleton (FastAPI) that returns the JSON contract.
The Brief
What you'll do, and what you'll demonstrate.
Generate actionable, customer-facing counterfactual explanations for motor-insurance pricing that respect feature mutability and consumer comprehension.
Earning criteria — what you'll demonstrate
- Generate diverse, actionable counterfactual explanations
- Encode mutability and reasonable-bound constraints in the search
- Design a JSON contract that a frontend team can implement against
- Translate model-explanation outputs into consumer-grade copy
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
AI Product Designer
Designing a JSON contract + copy guideline on top of a counterfactual generator is exactly the day-one work of an AI product designer at an insurtech or fintech scale-up.
This challenge sharpens
- product-design
- counterfactual-explanations
- interpretability
AI Engineer
Shipping a low-latency explanation service behind a FastAPI contract is core AI-engineer work at any product-led AI company.
This challenge sharpens
- fastapi
- dice-ml
- lightgbm
Data Scientist
Generating actionable model-output explanations that customer-facing teams trust transfers directly to data-science roles in customer-facing analytics teams.
This challenge sharpens
- counterfactual-explanations
- interpretability
- lightgbm