In this session, I argue that educators should build custom, purpose-built AI tutors rather than rely exclusively on enterprise AI tools. I will discuss my own effort to build a transparent and responsible AI tutor for my Fall 2024
Ethics and Artificial Intelligence course (see
here and
here for articles that describe the project). I will present my views about how bounded, deterministic instructional logic can be used to constrain AI tutor behavior, such that these tutors enhance student reasoning rather than replace it. I will also describe my IRB-backed plans to rigorously assess the impact of this tutor on student learning outcomes in my Fall 2026 version 2.0 of the course.
