Human-in-the-loop design principles
- Escalate by design
- Clear paths to a human for sensitive, complex, or emotional topics.
- Ground in verified sources
- Answers draw from current institutional policy, with sources shown.
- Disclose
- Students always know when they are talking to AI.
- Review outreach
- A person approves AI-drafted messages before they reach students.
- Measure connection
- Track handoffs to people and student-reported belonging alongside deflection rates.
Current applications
- Chatbots answering financial aid, registration, and deadline questions.
- Text-message nudges for milestones (e.g., Georgia State’s “Pounce”, which reduced summer melt among admitted students by 21% in a randomized trial).
- Generative assistants that draft advisor outreach or summarize notes.
- Degree-planning and course-recommendation tools.
Opportunities for belonging and integration
- Answer the questions students are embarrassed to ask a person.
- Reach students at the moments they disengage, including nights and weekends.
- Connect students to people: a mentor, a club, an office.
Risks
- Hallucinated policies or deadlines delivered with confidence.
- Over-reliance by students and staff; skills and judgment atrophy.
- Reduced human connection if AI replaces rather than routes to people.
- Unequal access and literacy across student groups.
Theories behind this section
In the Campus Enterprise Model
Where this work sits in campus systems, capabilities, and data.