Unveiling moral development in generative AI chatbots

Read full text Summarise with VIRAT
Authors: Jiayu Han, Alton Y.K. Chua
Cited by: 0
References: 17

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Introduction. Guided by Kohlberg’s theory, this paper aims to investigate the moral development levels of GAI chatbots. Method. The Defining Issues Test Version 2 (DIT-2) was applied to assess the reasoning stages of four GAI chatbots, namely, Claude 4, Claude 4.1, ChatGPT 4o, and ChatGPT 5. Analysis. A total of 240 data points (6 indices × 10 runs × 4 chatbots) were analysed using the Coefficient of Variation (CV), Welch’s ANOVA, and Games-Howell test. Results. The results showed that Claude 4 was the most consistent in responding to moral dilemmas, whereas ChatGPT 4o was the least. Compared with Claude 4.1 and ChatGPT 4o, Claude 4 and ChatGPT 5 exhibited similarly higher levels of postconventional reasoning and moral differentiation. Conclusion(s). This paper advances the literature on AI ethics by shifting the focus from outcome-oriented evaluations to developmental levels of reasoning. Additionally, it extends Kohlberg’s theory of moral development into the domain of GAI. Practically, this study helps users understand the moral reasoning of the latest chatbots for more informed use. It also guides developers in improving models toward greater transparency and ethical alignment.

Fields of Study & Subjects

Ethics and Social Impacts of AI Psychology of Moral and Emotional Judgment Ethics in Business and Education

Publication details

Journal / source
Information Research
Volume
31
Issue
iConf
Pages
735–743
Publication date
20 March 2026
Language
EN