Teaching With AI,Not for AI: Preserving Human Judgment and Surfacing Human Biases in Public Administration Education

Abstract

Artificial intelligence is reshaping public administration, yet public affairs curricula often stress tool proficiency and “human-in-the-loop” cautions without helping students examine how their own assumptions shape AI-assisted analysis. This paper uses learner-centered pedagogy that embeds AI as a “mirror” to surface human biases across problem framing, evidence use, alternative generation, and evaluation in a policy analysis assignment. Drawing on two cohorts in an MPA policy seminar, we compare no-AI-assisted and AI-collaborated cohorts and qualitatively analyze students’ policy analyses and reflections. The AI-collaborated cohort more consistently documented evidence verification, incorporated multiple perspectives, and produced more coherent justifications and comparative alternatives. A policy alternative assessment stage revealed students’ evaluative biases by contrasting their preferred options with AI-generated recommendations based on selected criteria and subsequent feasibility checks. Findings suggest that preserving human judgment does not require rejecting AI; reflective assignment design can build AI literacy and professional agency.