Michelle Blomberg

The Model-Governance Review Board

AI Tools & Strategy · Learning Design

The graded summative for Module 7 of my graduate Advanced Data Science course. You convene a NIST AI Risk Management Framework review of a classifier that makes decisions about people, one whose subgroup table shows real disparate impact, assemble the governance decision, cast an approve, reject, or condition vote, and then defend it live.

What this is

A performance-based summative worth 100 points. This interactive walks you through assembling the governance decision, about 30 minutes. It then hands you to the recorded defense, where you answer the board’s randomized questions on camera and self-assess. Your full governance dossier (Model Card, Datasheet, and the RMF mapping) is submitted separately. Ethics is an engineering requirement here, not an afterthought.

What you are being assessed on

100 points total

15Disparate-impact analysisSubgroup rates and the ratio
15Model Card completenessUse, limits, disaggregated results
14Datasheet completenessProvenance and consent
14RMF mappingHarms to Govern, Map, Measure, Manage
14Mitigation proportionalityEach names what it degrades
14Defended voteApprove, reject, or condition
14Live defenseAnswers under questioning
What you submit: this interactive plus the recorded defense cover the analysis, the RMF mapping, the vote, and the live defense. In the instructor-led version of the course you also upload your full governance dossier, the completed Model Card, the Datasheet for the Dataset, and the written RMF harm-to-function mapping, to the Module 7 assignment in the LMS. Working on your own, the interactive here is self-scoring practice.
MODEL-GOVERNANCE REVIEW BOARD
CONVENE
Your job: your model helps some groups and harms others. Assemble the governance decision, vote, and defend it to the board.

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.