Abstract: Artificial intelligence increasingly serves as an advisor in teaching and learning, yet it may be honest or pandering. Users cannot directly observe which type they face. This paper develops the Prince–Minister model to analyze how users should respond to advice from a potentially pandering AI. The user, treated as a prince, faces a binary decision. The AI, treated as a minister, either truthfully reports a private signal or always recommends intervention. The user updates beliefs, chooses an action to minimize expected loss, and may query an independent signal at a cost. The model derives the optimal action rule, query strategy, and equilibrium. It shows that positive AI messages are less trustworthy than negative ones, that querying is most valuable when the decision is uncertain, and that verification is worthwhile only when its benefit exceeds its cost. These results imply that students and teachers should be cautious toward affirmative AI feedback, take critical feedback seriously, verify selectively, and weigh verification costs against benefits. The model offers a decision-theoretic foundation for critical AI literacy, treating AI as an advisor to be evaluated rather than an authority to be obeyed.
Keywords: Prince–Minister model; AI-assisted teaching; pandering AI; advice-taking.
Title: A Study of the Prince–Minister Model for AI-Assisted Teaching
Author: Wenjie Chen, Xun Zheng
International Journal of Interdisciplinary Research and Innovations
ISSN 2348-1218 (print), ISSN 2348-1226 (online)
Vol. 14, Issue 4, October 2026 - December 2026
Page No: 7-13
Research Publish Journals
Website: www.researchpublish.com
Published Date: 15-October-2026
