| Design | Implementation | Ongoing evaluation | |
|---|---|---|---|
| Transparency | Document purpose, data, and logic in plain language | Tell students what is used and why, at the point of use | Publish an annual summary of uses and changes |
| Equity | Test for unequal error across groups before launch | Check who receives outreach and who does not | Disaggregate outcomes; retire uses that widen gaps |
| Human oversight | Name an accountable person for every decision point | Staff review flags and AI drafts before action | Audit overrides, escalations, and response times |
| Privacy & consent | Minimize data; set retention limits | Provide notice and meaningful choices | Review access logs; purge on schedule |
| Bias mitigation | Exclude fixed traits as standalone risk drivers | Monitor live performance by group | Recalibrate or retrain when performance drifts |
| Student agency | Co-design with students | Let students see and correct their data | Collect student feedback and act on it |
Each principle needs an owner at every stage, from design through annual review.
- Privacy
- Comply with FERPA and state law, and go further: collect the minimum necessary and set retention limits.
- Consent
- Tell students what is collected and why; offer meaningful choices where possible.
- Bias detection
- Audit model performance and intervention outcomes across groups before and after launch.
- Transparency
- Explain flags and recommendations in plain language to staff and students.
- Institutional governance
- A cross-functional body (student affairs, IT, IR, faculty, students) reviews new tools and use cases.
- Alignment with values
- Ask whether a use serves students or primarily serves institutional metrics. Analytics creates an obligation to act on what is learned (Prinsloo & Slade, 2017).
Multi-pillar governance models
- The Campus AI Framework organizes institutional AI work into pillars, including governance, risk, compliance, and data governance, with an AI Strategic Compass for direction-setting.
- LEAGUE-style models for learning analytics name six pillars: Lawfulness, Equity, Agency, Governance, Utility, and Ethics by Design.
- Legal floor: FERPA, state privacy laws, and emerging AI guidance for student data.
In the Campus Enterprise Model
Where this work sits in campus systems, capabilities, and data.