Ridgeline Renewables: audit a model before it ships
AI Tools & Strategy · Learning Design
The graded summative for Module 5 of my graduate Advanced Data Science course. A contractor left a “production-ready” failure-prediction model that scores 0.97 and says deploy now. Your job is to audit it, find every planted defect, re-estimate its honest performance, and decide whether it ships, then defend that call.
A performance-based summative assessment for this module, worth 100 points. The interactive below takes about 30 to 45 minutes and covers the diagnosis and the defense, and it is self-scoring, so you can run it on your own. In the instructor-led version of the course the full assessment, spread across the two-week module, also includes the corrected evaluation code and the written audit report you submit, which is where most of the hours go, and each student is issued their own version of the model, seeded from their roster identifier, so the specific defects differ from one submission to the next and the instructor can reproduce any of them.
Why you are doing this
A model that fits the data you already have is worthless; a model that generalizes to data you have not seen is the whole point. In the real world nobody hands you a clean model, they hand you a confident number and ask you to sign off. This assessment tests whether you can refuse to trust a number until you have interrogated how it was produced, the single most important habit this course teaches.
What you are being assessed on
100 points total
This simulation is an open educational resource by Michelle Blomberg, released under CC BY-NC-SA 4.0. Reuse, adapt, and share for noncommercial purposes with attribution, under the same license.