Simulations
Case study · AI Tools & Strategy · Learning Design
Each module of this course ends in a summative simulation of a real professional task, engineered so a generative model cannot complete it for the student. This is the high-level map; the full write-ups and the playable simulations live in the Authentic Assessment section.
The interactive simulations below are self-scoring practice anyone can run. In the graded, instructor-led version of the course, core datasets are seeded per student from a data-generating process only the instructor knows, so the graded artifact is unique to each student. Incidents (leakage, bias, missingness, a data-drift event) are planted, so the student has to detect, diagnose, and adapt. The consulting deliverables are defended in a simulated stakeholder meeting where a role-played executive, compliance officer, and engineer ask unscripted questions. The graded object is live reasoning about a private artifact, which is the durable answer to AI: change what students are asked to do, rather than trying to police it.
Modules 5 through 8 have playable interactive simulations, linked below. The Module 1 to 4 simulations are described in full within the course itself, so there are no separate M1 to M4 widgets to hunt for here.
The suite at a glance
- Module 5 · Model evaluationRidgeline Renewables: audit a model before it ships
Catch target leakage, contamination, a mismatched metric, and a subgroup collapse hiding behind a 0.97 score, then defend a sign-off, condition, or reject.
- Module 6 · Clustering and PCAHow Many Segments, Really?
Read bootstrap stability and defend the number of customer segments the evidence actually supports.
- Module 7 · Ethics and governanceThe Model-Governance Review Board
Take a model with real disparate impact before a NIST AI RMF board, choose a mitigation that names its cost, and defend your vote.
- Module 8 · CapstoneRiverCity: engagement and defense prep
Confirm the engagement is complete, then rehearse the live stakeholder defense on camera.
- Reusable engineDefense practice: record, respond, self-assess
On-device rehearsal that speaks randomized questions one at a time. Powers the Module 7 and 8 defenses.
These simulations are released under CC BY-NC-SA 4.0, free to reuse, adapt, and share for noncommercial purposes with attribution, under the same license. The course draws on open resources (An Introduction to Statistical Learning, MLU-Explain, the NIST AI RMF, and others), each credited in the module where it appears.
Read the full case for this approach and run each simulation in the Authentic Assessment section →