AI evidence extraction

Generative AI as a ‘Precipitant’ of Challenges in Doctoral Supervision: A Dialogue Among Supervisors

Luca Morini, Israel Kariyana, Nompilo Tshuma · Open Books and Proceedings · 2026

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

9/10

Relevance

2/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.38140/obp4-2026-02

Methodology & findings

Study design

Collaborative autoethnography (CAE) involving four experienced doctoral supervisors from the UK and South Africa.

Sample

N = 4, 3 groups

Primary method

No formal statistical methods were applied. The study employed qualitative thematic analysis through collaborative meaning-making, reflection, and self-examination of critical incidents and dialogue transcripts.

Main result

The study found that generative AI acts as a "precipitant" of challenges in doctoral supervision by highlighting existing structural dysfunctions rather than creating entirely new issues. The authors concluded that "GenAI technology is quickly bringing these pre-existing issues into much sharper contrast" including "the growing, output-oriented productivity emphasis in academia, the lack of formalisation in the role of the supervisor, and the complexities of modelling an evolving role for PGRs." Additionally, the research revealed that "AI can provide a lot of information, but it can't take a stand. It always goes for 'neutrality', which means it upholds the status quo," suggesting that GenAI tools perpetuate epistemic injustice by replicating stereotypical academic discourse rather than enabling diverse scholarly voices.

Reports effect sizes.

Research paradigm

Social constructionism

Author conclusions

The authors conclude that "there appears to be a potential way to positively address and mitigate the heightened challenges brought by GenAI in doctoral supervision: create relational spaces where we, as a community of supervisors and postgraduate researchers, acknowledge the thesis and 'doctorateness' as an emergent property of conversations marked by shared understanding and co-creation of a joint narrative." They further emphasize that supervisors should focus on "recovering a more process-oriented thinking around the PhD" and emphasize "the relational dimension of doctoral study" through "actual, back-and-forth verbal conversations and questioning with supervisees."

Risk of bias

Small sample size (n=4 supervisors only); Lack of anonymity introduces social desirability bias and vulnerability to researcher identity; Self-selection bias: participants were brought together by shared professional pathways and concern about GenAI; Single epistemological perspective (social constructionism) may not capture alternative viewpoints; Autoethnographic method inherently subject to researcher bias and selective narrative construction; Selection bias: authors self-selected based on shared professional pathways and concerns about GenAI; Researcher positionality bias: all authors are experienced supervisors with specific epistemological orientations that may influence interpretation; Small sample size (n=4): limited generalizability across different supervisory contexts and geographies; Lack of anonymity: self-disclosure may affect candor or create pressure toward consensus; Selection bias: small self-selected group of four supervisors with shared interests; Confirmation bias: authors already concerned about GenAI disruption before study commencement; Lack of anonymity: authors disclosed identities, potentially influencing how they presented experiences; Positionality bias: all authors are experienced supervisors in privileged UK/South Africa institutions, limiting representation of diverse global perspectives; Narrative construction bias: collaborative nature may lead to consensus-seeking rather than genuine disagreement capture; Recall bias: reliance on critical incidents from memory

Limitations

The study accepted "the limitations that come with this methodology
Firstly, the number of participants in a collective autoethnographic study is limited compared to other qualitative methods, starting from as few as two." Additionally, "CAE makes it difficult to safeguard anonymity, and we made a conscious decision to disclose our identities when reporting on our individual experiences and beliefs." The authors acknowledge that "by foregoing anonymity, we have created a certain vulnerability for ourselves as researchers, but also for others who may be implicated in our narrative."

Open questions raised

The authors identify several gaps: (1) The need for clearer institutional and disciplinary guidelines on GenAI use in doctoral education; (2) The need for supervisors to develop awareness of GenAI affordances and limitations; (3) The absence of consensus on how knowledge attribution and authorship should function in an AI-mediated research environment; (4) Lack of established practices and normative standards for GenAI use in academic writing and research; (5) The need to develop 'assessment literacy' among doctoral students regarding GenAI tools.
Lack of clarity and strategies for how supervisors should embrace or ban GenAI technology
Absence of established normative practices for GenAI use in doctoral education
Need for formalisation of the supervisor's role in the era of GenAI
Lack of consensus on authorship and ownership principles when GenAI is involved in research
Unclear responsibilities for educating postgraduate researchers about GenAI affordances and limitations

Extracted from: pdfAgreement 68%

Explore related topics

Responsible AI Use Doctoral Training Qualitative