Your browser does not fully support modern features. Please upgrade for a smoother experience.
Generative Artificial Intelligence in Doctoral Supervision: Ethical Boundaries, Feedback Practices, and Supervisory Relationships: History
View Latest Version
Please note this is an old version of this entry, which may differ significantly from the current revision.
Contributor:
Carla Giuliana Guanilo Pareja
,
Lidia Ysabel Pareja Pera
, Andres Arias Lizares , Lupe Marilu Huanca Rojas , Henri Emmanuel Lopez Gomez ,
Zoila Rosa Díaz Tavera
,
Clara Patricia Almonte Andrade
, Fernando Martín RAMÍREZ WONG , Wanda Román Santana ,
Carmen Mata-de-Salcedo
, Judith Marcela Martinez Alonzo

Generative artificial intelligence is increasingly being adopted in doctoral education to support writing, literature synthesis, translation, feedback, research planning, and administrative work. While emerging scholarship has examined generative AI in higher education, academic integrity, and student writing, its specific implications for doctoral supervision remain comparatively underexplored. This review argues that generative AI should not be understood merely as a technical aid, but as a socio-technical force that may reshape ethical boundaries, feedback practices, doctoral student agency, and supervisory relationships. Drawing on literature from doctoral education, supervision studies, academic writing, feedback literacy, and AI-mediated learning, the review examines three interrelated thematic areas of AI-augmented doctoral supervision. First, it considers ethical issues related to authorship, originality, academic integrity, transparency, privacy, confidentiality, bias, data protection, and overdependence. Second, it analyses how generative AI may influence feedback preparation, interpretation, timing, and dialogic use in doctoral writing and research development. Third, it explores implications for supervisory relationships, particularly trust, power, responsibility, emotional support, and autonomy. The review also discusses institutional implications, including AI literacy, supervisor development, doctoral policy, and responsible use guidelines, and positions generative AI as a human-centred support mechanism rather than a substitute for supervisory judgement, ethical responsibility, or relational mentoring.

  • generative artificial intelligence
  • doctoral supervision
  • doctoral education
  • feedback practices
  • supervisory relationships
  • academic integrity
  • AI literacy
Doctoral supervision is a relational and developmental academic practice through which researchers learn to produce original knowledge, exercise methodological judgement, communicate within a discipline, and form a scholarly identity. It extends beyond monitoring progress or correcting drafts: supervisors provide intellectual direction, ethical guidance, critical feedback, emotional support, and professional socialisation [1][2][3][4][5][6]. Candidates, in turn, are expected to move from supported participation towards increasing independence, develop an academic voice, and defend an original contribution [7][8]. The central supervisory challenge is therefore to combine guidance with autonomy. Effective support should make the candidate more capable of reading, reasoning, deciding, and writing independently rather than becoming a continuing substitute for those activities.
Generative AI introduces new possibilities into this already complex relationship. Large language model-based systems can produce, revise, translate, summarise, classify, and reorganise text in response to prompts. In doctoral work, they may support literature searches, conceptual clarification, research planning, academic writing, coding, meeting preparation, translation, editing, administrative communication, and preliminary feedback [9][10][11]. For candidates, these systems can provide immediate assistance between supervision meetings and may reduce linguistic or organisational barriers. For supervisors, they may help prepare routine materials, formulate examples, or structure developmental activities. These uses make GenAI potentially valuable, particularly where supervisory time, language support, or access to informal academic assistance is limited.
Their significance, however, extends beyond efficiency. Reading difficult literature, developing an argument, making methodological choices, interpreting criticism, and revising a scholarly text are not merely tasks to be completed; they are formative processes through which doctoral researchers acquire expertise and an independent voice [12][13]. GenAI can intervene directly in each of these processes. Used reflectively, it may help candidates compare alternatives, identify uncertainty, or articulate a question. Used substitutively, it may bypass the intellectual struggle through which judgement is developed. The same capabilities therefore generate questions about academic integrity, authorship, originality, transparency, privacy, confidentiality, bias, epistemic authority, and the boundary between legitimate assistance and outsourced intellectual labour [14][15].
Much of the higher education debate has concentrated on assessment, misconduct, AI-generated student text, and institutional policy [16][17]. Doctoral education presents a distinct context. It is organised around original research, prolonged engagement with uncertainty, discipline-specific standards, and a sustained relationship in which the supervisor both supports development and contributes to judgements about progress and quality. GenAI may therefore alter more than the production of text. It can affect how knowledge claims are formulated, how feedback is interpreted, how candidate competence is inferred, and how authority and responsibility are distributed between candidates, supervisors, and institutions. These relational consequences require analysis that cannot be reduced to plagiarism detection or productivity gains.
AI use may also remain hidden or ambiguously disclosed. Candidates may use GenAI to interpret supervisory comments, revise arguments, generate literature summaries, rehearse explanations, or seek reassurance without discussing that assistance. In this sense, the system can operate as an unofficial third actor in the supervisory process [12][18]. Such use can be supportive, but opacity creates practical and ethical problems. Supervisors may be unable to determine whether a polished passage reflects the candidate’s current understanding, whether an argument has been substantially machine-shaped, or whether confidential data have entered an external platform. Candidates may likewise be uncertain about what is permitted, what should be disclosed, and where support becomes inappropriate substitution.
Feedback makes these tensions particularly visible. Doctoral feedback is dialogic and developmental: it helps candidates interpret disciplinary standards, refine arguments, evaluate evidence, and become increasingly self-regulating scholars [19][20]. GenAI can provide rapid comments on clarity, structure, coherence, or style and can help candidates formulate questions before a meeting. Yet generated feedback may lack disciplinary nuance, methodological context, ethical sensitivity, and knowledge of the candidate’s developmental trajectory. It can also appear authoritative despite being generic or wrong. The relevant issue is therefore not whether machine feedback is available, but how it interacts with supervisor feedback, feedback literacy, and the gradual development of independent evaluative judgement.
The supervisory relationship likewise depends on negotiated expectations, intellectual challenge, emotional attunement, and shared responsibility [12][18]. If GenAI is used to avoid difficult conversations, generate impersonal comments, or replace mentoring, it may weaken trust and relational accountability. Conversely, transparent use may improve preparation, enable candidates to articulate uncertainty, and make meetings more focused. AI can support a conversation, but it cannot assume responsibility for the quality of guidance, the wellbeing of the candidate, or the ethical consequences of a research decision. The central question is therefore how GenAI can be integrated without displacing human judgement or the formative role of the supervisor–candidate relationship.
Recent syntheses address important parts of this landscape. Reviews of AI and academic integrity examine broad higher education risks and policy responses [14][15]; a systematic review of generative-chat feedback focuses on academic writing outcomes [21]; and a systematic review of AI-powered tools maps applications in doctoral supervision [22]. These studies provide essential foundations, but their primary units of analysis are usually policy problems, feedback interventions, or technological tools. They do not fully integrate ethical accountability, feedback literacy, doctoral agency, supervisory trust and power, and institutional governance within one relational system. This leaves a need for a synthesis that treats GenAI as a mediating influence across the whole supervisory ecology rather than as an isolated application.
Table 1 positions the present review in relation to these adjacent syntheses. It shows that the distinctive contribution is not a catalogue of tools, but an integrated account of how AI-mediated practices affect the responsibilities and relationships through which doctoral researchers are formed.
Table 1. Positioning of the present review in relation to prior syntheses.
This review therefore shifts the unit of analysis from individual tools or isolated tasks to the socio-technical supervisory system comprising doctoral researchers, supervisors, institutions, and GenAI systems. It develops a human-centred framework connecting five dimensions: ethical accountability; feedback and learning; agency and autonomy; trust and supervisory relationships; and institutional governance. As a conceptual and integrative framework, it provides an organising structure for comparing future empirical findings and for translating broad principles into operational supervisory practices. Figure 1 represents these relationships.
Encyclopedia 06 00209 g001
Figure 1. Conceptual and integrative human-centred framework for generative AI in doctoral supervision. Four actors and five interdependent dimensions are situated within a recurrent cycle of expectation setting, evaluation, adjustment, and re-evaluation across doctoral progression.
Human judgement and scholarly responsibility form its governing core. GenAI may mediate writing, planning, feedback, and communication, but it cannot hold academic authorship, moral accountability, or supervisory responsibility. The four human and technological actors are therefore connected without being treated as equivalent: responsibility remains with candidates, supervisors, and institutions.
The framework is dynamic rather than a fixed configuration of actors and responsibilities. As doctoral work progresses, the relative intensity and form of interaction among the doctoral researcher, supervisor or supervisory team, institution, and GenAI system may change as researcher autonomy develops, project requirements evolve, and new methodological, data, publication, or technological conditions emerge. Responsible AI-augmented supervision therefore requires expectations and practices to be revisited at relevant milestones rather than established once at candidature entry. Human judgement and scholarly responsibility remain the governing core throughout this progression, while ethical accountability, feedback and learning, agency and autonomy, trust and supervisory relationships, and institutional governance operate as interdependent dimensions through which AI-related practices are evaluated and adjusted.
The five dimensions are analytically distinct but practically interdependent. Ethical accountability influences what information may be shared and who is responsible for a claim; feedback and learning concern how generated material is evaluated and used; agency and autonomy address whether candidates retain intellectual control; trust and supervisory relationships determine whether use is discussable and development can be judged; and institutional governance supplies the policy, infrastructure, and training that make responsible practice possible. A weakness in one dimension can undermine the others. For example, useful feedback generated from confidential data is not ethically defensible, and transparent use that nevertheless replaces candidate judgement does not meet the developmental purpose of doctoral education.
These relationships organise the review around four questions: (RQ1) How does GenAI reshape ethical accountability, authorship, transparency, and responsibility in doctoral supervision? (RQ2) How may GenAI influence feedback practices, doctoral learning, agency, autonomy, and independent scholarly judgement? (RQ3) How may GenAI reshape trust, power, and responsibility within doctoral supervisory relationships? (RQ4) What institutional governance, AI-literacy, data-protection, and support conditions are needed for responsible and human-centred GenAI use in doctoral supervision?
The article proceeds from the methodological approach and conceptual foundations to current uses and opportunities. It then examines ethical boundaries and research integrity, AI-mediated feedback and doctoral learning, trust and power in supervisory relationships, and institutional governance. Operational recommendations convert the synthesis into observable responsibilities for the principal stakeholders. The final sections identify research gaps, acknowledge the limitations of the review, and state the conditions under which GenAI may support rather than weaken doctoral formation.

References

  1. Vähämäki, M.; Saru, E.; Palmunen, L.-M. Doctoral Supervision as an Academic Practice and Leader–Member Relationship: A Critical Approach to Relationship Dynamics. Int. J. Manag. Educ. 2021, 19, 100510.
  2. Courtney, S.A.; Alexander, A.N. Doctoral Preparation in Mathematics Education: Graduates’ Perspectives on Program Features, Expertise Development, and Scholarly Learning Cultures. Int. J. Sci. Math. Educ. 2026, 24, 69.
  3. Li, Y.; Xu, W.; Chen, J. PhD Student-Supervisor Relationship and Its Impacts: A Perspective of the Interpersonal Relationship Model. Front. Educ. 2025, 10, 1570137.
  4. Kitano, N.; Aldous, C.; Race, D.; Clegg, K. The Taboo of Power Dynamics in Doctoral Supervision: Cross-Cultural Insights from Australia, South Africa and the South Pacific Region. High. Educ. Q. 2026, 80, e70155.
  5. Becker, S.; Jacobsen, M.; Friesen, S. Four Supervisory Mentoring Practices That Support Online Doctoral Students’ Academic Writing. Front. Educ. 2025, 10, 1521452.
  6. Huang, X.; Li, H.; Shcheglova, I. Exploring the Role of Generative Artificial Intelligence (GenAI) in Research Activities in Shaping Doctoral Students’ Well-Being. Euro J. Educ. 2026, 61, e70750.
  7. Guo, J. The Augmented Scholar: How Generative AI Usage Influences Research Self-Efficacy among Chinese Doctoral Students. High. Educ. Res. Dev. 2026, 1–17.
  8. Peng, Y. Toward an Understanding of Doctoral Students’ Feedback Literacy in Academic Publishing: An Ecological Perspective. Humanit. Soc. Sci. Commun. 2025, 12, 1819.
  9. Cowling, M.; Crawford, J.; Allen, K.-A.; Wehmeyer, M. Using Leadership to Leverage ChatGPT and Artificial Intelligence for Undergraduate and Postgraduate Research Supervision. Australas. J. Educ. Technol. 2023, 39, 89–103.
  10. Dai, Y.; Lai, S.; Lim, C.P.; Liu, A. ChatGPT and Its Impact on Research Supervision: Insights from Australian Postgraduate Research Students. Australas. J. Educ. Technol. 2023, 39, 74–88.
  11. Iatrellis, O.; Bania, A.; Samaras, N.; Kosmopoulou, I.; Panagiotakopoulos, T. ChatGPT in Doctoral Supervision: Proposing a Tripartite Mentoring Model for AI-Assisted Academic Guidance. Int. J. Dr. Stud. 2025, 20, 009.
  12. Lai, S.; Liu, S.; Dai, Y.; Lim, C.P.; Liu, A. The Impacts and Tensions of Generative AI on Doctoral Students’ Supervisory and Peer Dynamics: An Activity Theory Analysis. Australas. J. Educ. Technol. 2025, 41, 5.
  13. Brown, A.; Rossouw, J. Artificial Intelligence as a Reflexive Collaborator in Graduate Studies Supervision. Transform. High. Educ. 2026, 11, a657.
  14. Bittle, K.; El-Gayar, O. Generative AI and Academic Integrity in Higher Education: A Systematic Review and Research Agenda. Information 2025, 16, 296.
  15. Balalle, H.; Pannilage, S. Reassessing Academic Integrity in the Age of AI: A Systematic Literature Review on AI and Academic Integrity. Soc. Sci. Humanit. Open 2025, 11, 101299.
  16. Yusuf, A.; Pervin, N.; Román-González, M. Generative AI and the Future of Higher Education: A Threat to Academic Integrity or Reformation? Evidence from Multicultural Perspectives. Int. J. Educ. Technol. High. Educ. 2024, 21, 21.
  17. Wang, H.; Dang, A.; Wu, Z.; Mac, S. Generative AI in Higher Education: Seeing ChatGPT through Universities’ Policies, Resources, and Guidelines. Comput. Educ. Artif. Intell. 2024, 7, 100326.
  18. Qi, J.; Tuxworth, J.; Gomes, C. Navigating Supervisory Conversations About Gen AI: Taboo and Supervisory Power Dynamics. High. Educ. Q. 2026, 80, e70151.
  19. Jensen, L.X.; Bearman, M.; Boud, D.; Konradsen, F. Feedback Encounters in Doctoral Supervision: The Role of Generative AI Chatbots. Assess. Eval. High. Educ. 2025, 51, 849–862.
  20. Khuder, B.; Cervin-Ellqvist, M.; Vander Borght, M.; Wingrove, P. Academic Socialisation through AI and Peer Feedback in Doctoral Writing: Information and Orientation in Feedback Ecologies. Assess. Eval. High. Educ. 2026, 1–20.
  21. Urzúa, C.A.C.; Ranjan, R.; Saavedra, E.E.M.; Badilla-Quintana, M.G.; Lepe-Martínez, N.; Philominraj, A. Effects of AI-Assisted Feedback via Generative Chat on Academic Writing in Higher Education Students: A Systematic Review of the Literature. Educ. Sci. 2025, 15, 1396.
  22. Thong, C.L.; Atallah, Z.; Islam, S.; Lim, W.; Cherukuri, A.K. AI-Powered Tools for Doctoral Supervision in Higher Education: A Systematic Review. J. Info. Know. Mgmt. 2025, 24, 2530001.
More
© Text is available under the terms and conditions of the Creative Commons Attribution (CC BY) license; additional terms may apply. By using this site, you agree to the Terms and Conditions and Privacy Policy.
This entry is offline, you can click here to edit this entry!
Academic Video Service
Feedback