TRUSTWORTHY AI BEGINS WITH TRUSTWORTHY DATA

Higher education institutions are moving quickly to adopt AI for advising, admissions, and student success. But many are building these systems on top of institutional data that was never designed to work well with AI.

Take a term like “at-risk student.” The Registrar’s Office, Academic Advising, and Financial Aid may each define it differently. When an AI system pulls from these sources, those differences can lead to inconsistent answers. And when staff can’t understand where an answer came from or how a term was defined, trust in the technology quickly breaks down.

This talk makes the case that trustworthy AI in higher education starts with something simple: a clear, shared data dictionary.

We’ll look at a real-world example of how inconsistent definitions can undermine trust, what a shared data dictionary looks like in practice, and how basic data governance can create a stronger foundation for explainable AI. We’ll focus on a practical starting point: getting everyone to agree on the definitions of the ten data fields their institution uses most.

Attendees will leave with a practical framework they can use to review their own data definitions before their next AI rollout and a clearer understanding of why good AI starts with good data.