Good audits connect evidence to governance decisions rather than accumulating principles.
Mid-Course Review
The architecture of ethical analysis
Can you turn seven weeks of concepts into one repeatable audit practice?
A full workshop week: teams integrate technical evidence, ethical frameworks, stakeholders, law, uncertainty, pre-mortems, and an ethical matrix into the capstone method.
Join the live roomBefore class
Bring a capstone candidate and a one-page system boundary: purpose, actors, data, model, workflow, affected parties, and current controls.
After class
Submit the revised system boundary, first ethical matrix, and top five pre-mortem failures.
The promise
By the end of this room…
- 01Run the course's ethical-analysis sequence end to end.
- 02Build a stakeholder × principle ethical matrix.
- 03Conduct a pre-mortem and prioritize failures by severity and likelihood.
- 04Distinguish a claim, its evidence, uncertainty, owner, and remedy.
Why this week now
Signals, not scene-setting.
A pre-mortem protects against the success narrative becoming the only scenario the team can imagine.
The capstone is an adversarial defense, so uncertainty and counter-evidence are first-class artifacts.
Run of show
Provoke → frame → work → argue → synthesize.
- 0:00provocation
Commit before the concepts
Which layer most often makes an AI audit fail?
Live activity · week 7 opening - 0:15frame
From principle list to audit trail
Good audits connect evidence to governance decisions rather than accumulating principles.
- 0:45discussion
Reading tension
Student leaders present the assigned readings as a clash of defensible positions, then moderate questions that expose the hidden assumptions.
- 1:10frame
SMACTR, ethical matrix, pre-mortem
A pre-mortem protects against the success narrative becoming the only scenario the team can imagine.
- 1:35break
Break
Ten minutes. Leave the room's unresolved question visible.
- 1:45forensics
Clearview synthesis speed round
Role-based groups work the anchor case through technical, legal, stakeholder, and normative lenses.
Deliverable · A two-minute finding with evidence, uncertainty, and an actionable remedy.Live activity · week 7 forensics - 2:25controversy
Capstone red-team clinic
Assigned positions, side-switch, and a joint recommendation that names the value or stakeholder it leaves exposed.
Live activity · week 7 controversy - 2:50synthesis
Re-vote and leave a trace
Repeat the opening vote, inspect what moved, and submit the strongest argument you still reject.
Live activity · week 7 exit
Case room
Evidence before opinion.
Clearview synthesis
Can one audit connect data, privacy, bias, accountability, and enforcement?
Amazon hiring reconstruction
Which audit artifact would have surfaced the failure earliest?
Reading stack
Read the tension, not the bibliography.
- 01CoreClosing the AI Accountability Gap ↗
Raji et al.
- 02ReferenceAI Risk Management Framework ↗
NIST
Evidence ledger
Every case has a receipt.
4 primary, scholarly, or first-party sources
Closing the AI Accountability Gap
Raji et al.'s end-to-end internal algorithmic auditing framework and the spine of the capstone audit.
Audit framework ↗Artificial Intelligence Risk Management Framework 1.0
The primary source for Govern–Map–Measure–Manage and lifecycle risk ownership.
Risk-management framework ↗Clearview AI ordered to delete facial recognition data
A cross-jurisdictional case about scraping, consent, biometric inference, and enforcement limits.
Regulatory finding ↗A Framework for Understanding Sources of Harm throughout the Machine Learning Life Cycle
Separates historical, representation, measurement, aggregation, evaluation, and deployment harms.
Research framework ↗