Week 14: Trustworthy AI and Responsible AI in Practice

Dates: Apr 12-16 · Reading: Handout 12: Trustworthy AI and Responsible AI

Everything we’ve studied - security, fairness, privacy - rolls up into one business question: can people trust this AI? Monday: the four pillars of trustworthy AI (transparency, explainability, robustness, accountability), a real “why the model decided this” explanation, and how organizations build responsible AI by design. Wednesday: a hands-on tabletop incident-response exercise (CISA scenario) with an AI/ML system, then a debrief on AI + human judgment. Lab 12 released; Quiz 11 Wednesday.

📄 Read Handout 12 (PDF)

Learning Objectives

Explain the four pillars of trustworthy AI: transparency, explainability, robustness, accountability
Describe how organizations build and deploy responsible AI (“trust by design”)
Apply the incident-response lifecycle to an AI/ML system failure
Connect governance, transparency, and business decision-making

Lecture Slides

⬇ Monday: Trustworthy & Responsible AI (PPTX) ⬇ Wednesday: Tabletop Incident Response (PPTX)

Two full 75-minute decks, CSU-branded. Monday: the four-pillars framework, a transparency-vs-explainability diagram, a real model-explanation figure, robustness and accountability visuals, and a "trust by design" pipeline. Wednesday: a facilitated three-round tabletop with the NIST incident-response lifecycle and a debrief - detailed speaker notes throughout.

Monday Session

Why trust is a business (and legal) requirement, then the four pillars: transparency (be open about the system; model cards), explainability (WHY a specific decision - shown with a real phishing-classifier explanation), robustness (holding up under attack, data drift, and edge cases), and accountability (named owners, audits, human override, appeals). Finish with “trust by design” and real case studies of trust won and lost.

Wednesday Session

A hands-on tabletop incident-response exercise (adapted from CISA’s free packages): teams work an AI fraud-detector crisis through the NIST lifecycle - detect & analyze, contain & decide, recover & communicate - then debrief on how AI (fast detection) and human judgment (deciding, communicating, accountability) work together.

Lab

Lab 12: Tabletop Incident Response Exercise. A team, talk-through crisis simulation (no real systems touched) using CISA’s free scenarios. Assign roles, work three rounds of “injects,” and debrief. Graded on participation and reasoning.

⬇ Lab 12 Guide & Worksheet (PDF)

Complete the worksheet (roles, three decision logs, five debrief questions) and submit in Canvas. AI tools are not permitted for the write-up.

Quiz / This Week

Quiz 11 (Wednesday, in class). The four pillars of trustworthy AI; transparency vs. explainability; model cards; data drift; responsible AI / “trust by design”; the incident-response lifecycle; and AI + human judgment.

In-Class Activity

AI Incident Response: Biased Hiring System — CISA-style incident-response tabletop (small groups). Graded, 20 points.

⬇ Activity slides (PPTX) 📄 Student worksheet (PDF)

Graded in-class activity with a worksheet. Full rotation on the Activities page.

← All lecture notes · Assessments unlock on the Quizzes and Labs pages when the instructor releases them.