Evaluating the impact of a socratic AI mentor on the ethical decision-making of undergraduate students: a quasi-experimental case study at Assumption University of Thailand

Background: Generative AI is being quickly used in higher education, which has created a "AI Learning Paradox" where students may rely on automated short-cuts that make them lose interest in learning and make "hollow decisions." Most of the time, traditional ethics education does not help people build the inner strength they need to fight these self-serving biases. Aims and Objectives: This study looked at how well an empathetic, Socratic AI agent that was meant to help students improve their morals worked. The main goals were to look at how student achievement changed when educators tried to reduce bias, how dialog changed students' morals, and how students felt about the AI interface. Theoretical Framework: The research was based on two different theories: the Western Human-AI Collaboration Model, which sees AI as a Socratic dialog partner, and the Eastern Buddhist Psychology of Intention, which looks at how people can change their desires from cravings to good desires through thoughtful reflection. Method: This study followed 33 Business Administration students at Assumption University of Thailand for four weeks using a quasi-experimental classroom action research design with a one-group pre-test and post-test. An automated AI interaction log and a 5-point Likert scale perception questionnaire were used to gather data. Results: The results of a paired-samples (t)-test showed that students did statistically better on the post-test than they did on the baseline tests (t (32) = -4.81), (p =.0048), Cohen's (d = 0.40), and 66.7% of them were able to choose the ethical choice. The dialog logs showed that the main ways to lower bias were through structural ethical prompts (51.5%) and deep personal thought (30.3%). Items on the questionnaire scored higher than the 3.00 baseline, and the caring tone (M = 4.30) created a strong sense of psychological safety. Conclusion and Recommendation: The Socratic AI agent worked well as a digital moral friend, changing actions that were based on fear into actions that were based on integrity. Socratic and empathetic AI models should be built into core curriculums and teachers should be trained on how to use AI technologies to help students learn and grow morally and academically.