Mapping moral reasoning research among college students: a bibliometric analysis of traditional foundations

Synthesis: Jia, Parahakaran, Wider, Zhao and Wang (2026) map 677 Web of Science documents published between 1983 and 2024 to describe how research on college students' moral reasoning is actually organized. Co-citation analysis returns three clusters — moral judgment and its measurement, moral behavior and ethical decision-making, and moral disengagement — while keyword analysis places digital ethics, Academic Integrity, and Generative AI in a small, self-contained periphery. Their protocol is reported in enough detail to repeat: PRISMA-based screening, VOSviewer 1.6.20, association-strength normalization, fractional counting, and thresholds rerun at adjacent values to confirm that the cluster solution was not an artifact of one cutoff. The authors read the pattern as theoretical continuity with contextual expansion rather than a paradigm shift: Kohlberg and Rest still hold the field's center, and AI-related ethics is being assimilated through the vocabulary of cheating and misconduct. For Higher Education, the practical claim is that digital ethics belongs inside moral reasoning instruction rather than in a separate compliance module.

Key Findings

1. The corpus comprised 677 documents from the Web of Science Core Collection covering 1983 to 2024, which had received 16,426 citations excluding self-citations, an h-index of 59, and an average of 24.26 citations per article.
2. Co-citation analysis at a threshold of 20 citations identified 54 citing references and three clusters: moral judgment and development, moral behavior, and moral disengagement.
3. Rest (1992) recorded the highest total link strength at 86, followed by Hanlin and Zucker (1986) and Rest et al. (1999) at 84 each — evidence that the Defining Issues Test, not shared theory alone, binds the developmental cluster together.
4. Keyword analysis retained 57 terms appearing at least five times; "ethics" led with 41 occurrences, ahead of "moral reasoning" at 35, "moral judgment" at 34, and "college students" at 24.
5. Four keyword clusters emerged, and the digital one — academic dishonesty, plagiarism, integrity, piracy — was the smallest at nine keywords, with members carrying low link strength relative to their frequency.
6. Digital-ethics terms co-occur mainly with one another rather than across the network, which the authors interpret as a coherent specialty that has not diffused into the field's core vocabulary.

Three clusters, one shared instrument

The co-citation structure explains more than it lists. Cluster one is psychological and developmental, its cohesion supplied less by theoretical agreement than by a common measurement technology: the Defining Issues Test, whose variants dominate the top of the co-citation ranking. Cluster two comes from business and organizational ethics and is organized around predictive models — the theory of planned behavior, Hunt and Vitell's general theory, Jones's moral intensity — asking what turns judgment into action. Cluster three, moral disengagement, is the most recently consolidated and functions as a bridge, retaining the cognitive orientation of the first while addressing the judgment–action gap that motivates the second. The authors argue these are parallel strategies for one unresolved problem rather than successive stages, and that the field's structure was generated by Blasi's observation that moral judgment and moral conduct relate weakly and inconsistently. Sharing an instrument, they suggest, can bind a cluster as tightly as sharing a theory.

A periphery, not a replacement

The keyword map shows where the digital arrives, and where it stops. Cluster one covers moral education and ethical practice, including virtue ethics, ethical self-examination, and simulation-based instruction; cluster two covers student dilemmas and the cognitive and emotional shaping of judgment; cluster three covers causes of unethical behavior across personal, social, and cultural factors. Cluster four, on academic integrity in digital settings, contains nine keywords and sits at the edge. Two explanations are offered: recency, since generative AI entered the literature only in the window's final years, and structure, since the cluster is organized around behaviors to be detected and deterred rather than a developmental construct. That second reading carries the sharper warning. When digital ethics is absorbed as a compliance problem, the authors argue, students' online moral experience stops being treated as a domain in which reasoning itself develops.

The judgment–action gap as organizing problem

Across both analyses, the field's persistent question is why ethical recognition does not reliably produce responsible action. Each cluster is a different way of approaching it, and the digital cluster is the weakest-served by that inheritance. Anonymity, algorithmic mediation, rapid diffusion, and AI-assisted authorship redistribute agency in ways the field's theories and measures were not built to capture, since those instruments assume bounded decision settings with identifiable consequences and authors. The authors call this a theory-context and voice gap: the conceptual vocabulary is being extended to environments whose affordances it did not anticipate, and bibliometric categories privilege published scholarly language over students' own culturally situated definitions of acceptable and transgressive conduct. Their prescription is not a new model but a research agenda — test Kohlberg and Rest against digitally native dilemmas, study responsibility attribution when work is co-produced with AI, and include students as participants in defining the problems.

What this means for practice

• Instructors. Teach digital ethics as a reasoning domain with dilemmas, justification, and competing obligations, not as a rule list about detection, since the map shows the field's compliance framing has crowded out developmental framing.
• Curriculum designers. Add dedicated modules on digital ethics, responsible AI use, academic integrity, social media ethics, and algorithmic decision-making to existing ethics or general education courses, as the authors recommend.
• Instructors. Use dilemma discussion and simulation deliberately, because the moral-education cluster already treats interactive and immersive formats as promising vehicles for reasoning practice.
• Researchers. Cite across the three clusters rather than within one, since each addresses a different segment of the same judgment–action problem and the digital specialty remains lexically isolated.

Limitations

• The corpus draws on a single database, the Web of Science Core Collection, excluding work indexed only in Scopus, ERIC, or other sources and likely underrepresenting non-English and regionally published scholarship.
• Retrieval depends on one search string; alternative vocabulary would have produced a different corpus.
• The networks are aggregate structures over 1983–2024 with no longitudinal overlay or period-by-period decomposition, so no chronological transition is demonstrated.
• Cluster labels are interpretive rather than algorithmic, assigned from the highest-weight items and agreed between authors, though the underlying item lists are published for independent review.

Citation

Jia, J., Parahakaran, S., Wider, W., Zhao, J., & Wang, Y. (2026). Mapping moral reasoning research among college students: a bibliometric analysis of traditional foundations and emerging digital ethics. Cogent Education, 13(1).