03 · What You Need to Know
Navigating Supervisor-Required AI Use, Research Ethics, and Academic Authority
Why Supervisor-Required AI Use Can Create Difficult Decisions
Research supervisors influence methodological choices, thesis progression, research opportunities, and scholarly development.
Students may therefore experience pressure to follow supervisory recommendations even when they have legitimate concerns.
However, supervisors and students may interpret the benefits and risks of AI differently.
A supervisor may view AI as a useful analytical tool, while the student may question its compatibility with the selected methodology or approved data-handling arrangements.
Neither perspective should automatically be dismissed.
The appropriate response depends on the actual research activity, relevant evidence, and applicable institutional requirements.
Distinguish Disagreement From an Actual Violation
A student may consider an AI application undesirable without it being prohibited.
For example, a student may prefer manually developing preliminary qualitative codes because this approach supports their interpretive engagement with the data.
The supervisor may prefer exploring AI-assisted coding as a supplementary procedure.
This may be a methodological disagreement that can be resolved through scholarly discussion.
By contrast, a proposed upload of confidential participant information to an unauthorized external service may raise a compliance concern.
The distinction matters because a personal preference and a binding research obligation do not carry the same authority.
Methodological disagreement
Different defensible judgments about whether AI assistance is useful or suitable for the research task.
Compliance or integrity concern
A proposed AI activity that may conflict with research ethics approval, confidentiality, institutional rules, accurate reporting, or other binding obligations.
Students should identify which category best describes the concern before determining how to respond.
Why Supervisory Authority Has Limits
Supervisors have legitimate responsibilities for guiding research quality and student development.
However, supervisory authority operates within institutional and research governance frameworks.
A supervisor cannot ordinarily authorize processing prohibited by binding institutional requirements merely through an informal instruction.
Similarly, a supervisor should not require a student to fabricate research findings or misrepresent procedures.
The European Code of Conduct for Research Integrity emphasizes reliability, honesty, respect, and accountability.
These principles apply to supervisors as well as students.
Where authority is uncertain, the responsible institutional office should clarify which requirements govern the proposed activity.
What If the Supervisor Requires Uploading Confidential Research Data?
Confidentiality concerns should be evaluated before information is transferred.
Research material may include identifiable interviews, unpublished institutional records, restricted datasets, or confidential collaborative manuscripts.
Using an external AI service can introduce questions about provider access, retention, contractual conditions, and permitted processing.
Whether the proposed use is authorized depends on the actual system and governing arrangements.
Students should examine the approved research protocol, participant consent, institutional data-handling requirements, and relevant provider conditions.
Where authorization has not been established, the student should seek clarification before proceeding.
Watch Out
A supervisor's request to use AI does not automatically authorize uploading confidential participant data. Students should confirm the required approvals and processing safeguards before transferring restricted information.
What If the Supervisor Requires AI-Generated Research Methods?
A supervisor may recommend using AI to compare methodological alternatives or generate preliminary analytical code.
These activities are not necessarily inappropriate.
However, the student should understand how the recommended procedure relates to the research question and data.
For example, a supervisor may suggest asking AI to select a statistical model.
The student may reasonably request an opportunity to evaluate the recommendation against methodological literature or consult a qualified analyst.
The concern should be framed around scientific validity rather than a general objection to AI.
AI-generated methodological advice should not be implemented solely because either the supervisor or the system presents it confidently.
What If AI Use Conflicts With the Research Methodology?
Some research approaches assign particular importance to the researcher's interpretive role.
For example, Braun and Clarke (2022) emphasize active researcher engagement and methodological coherence in reflexive thematic analysis.
A student using this approach may question whether automatically generated themes are compatible with the intended analytical process.
The supervisor may nevertheless consider AI useful for exploring alternative interpretations.
The disagreement should examine the actual function of AI within the methodology.
Using AI to suggest questions for reflection is not necessarily equivalent to delegating the development of final themes.
Students should support their concerns through relevant methodological literature and propose alternatives consistent with the research design.
What If the Supervisor Requires AI to Draft Substantive Thesis Content?
AI-assisted writing may range from language editing to generating theoretical arguments, methodological explanations, and interpretations.
Students may reasonably question whether extensive generated drafting is compatible with degree requirements concerning independent scholarly contribution.
The appropriate response depends on institutional policy and the nature of the task.
A student may propose using AI for language feedback while independently constructing the scientific argument.
If the supervisor requires substantive AI drafting, the student should seek clarification about how intellectual contribution, verification, and disclosure will be handled.
The distinction between legitimate AI assistance and intellectual outsourcing is relevant to this discussion.
What If the Supervisor Requires AI Use During an AI-Prohibited Assessment?
A research methods assessment may explicitly prohibit generative AI because it measures independent competence.
If a supervisor encourages AI use despite that restriction, the student should identify which rules apply and who has authority to modify them.
An informal instruction from one academic staff member may not override a formal course or graduate school requirement.
The student should request clarification before undertaking the disputed assistance.
Where necessary, the course coordinator or responsible graduate program authority should resolve the inconsistency.
Students should avoid assuming that a supervisor's approval automatically protects them from consequences under another applicable policy.
How Should Students Evaluate the Supervisor's Rationale?
Students should consider whether the supervisor has identified a legitimate educational or methodological purpose for AI use.
For example, AI-generated code may help automate repetitive operations while allowing the student to concentrate on interpretation.
AI-assisted literature exploration may suggest terminology that improves a database search.
These benefits may be genuine.
However, benefits should be evaluated alongside verification requirements, data risks, and methodological suitability.
The relevant question is whether the proposed AI use improves the research under acceptable conditions.
Neither technological enthusiasm nor generalized skepticism provides a sufficient justification.
Why Students Should Raise Concerns Specifically
A statement such as "I do not trust AI" may not communicate the precise research concern.
A more useful explanation identifies the particular risk or requirement.
For example, the student might explain that the approved protocol does not clearly authorize external processing of identifiable transcripts.
Alternatively, the student may explain that the generated statistical model does not account for repeated observations.
Specific concerns can be evaluated against evidence.
They also make it easier to identify whether the problem can be resolved through verification, additional safeguards, or an alternative method.
What Alternatives Can Students Propose?
Students may be able to address the supervisor's objective without undertaking the disputed AI activity.
| Supervisor's proposed use |
Student concern |
Possible alternative |
| Upload confidential interviews to an external chatbot |
Unauthorized processing and confidentiality |
Use direct analysis, synthetic practice data, or an approved processing environment |
| Accept AI-selected statistical methods |
Unverified methodological suitability |
Compare recommendations with methodological literature and qualified review |
| Generate final qualitative themes automatically |
Conflict with interpretive methodology |
Use AI for supplementary reflection while retaining researcher-led interpretation |
| Generate substantive thesis arguments |
Independent intellectual contribution |
Develop the argument independently and use permitted AI feedback |
| Use AI during an explicitly restricted assessment |
Conflict with assessment rules |
Seek clarification from the responsible academic authority |
| Use AI-generated references without checking them |
Fabricated or misrepresented sources |
Verify original publications before incorporating claims |
Alternatives should reflect the actual research context. An approved local system, for example, is not automatically authorized for every category of data.
How Should Students Communicate Their Concerns Professionally?
Students should explain the concern without assuming that the supervisor intends to violate research requirements.
A constructive discussion can identify the proposed task, relevant uncertainty, and an alternative way to achieve the research objective.
For example, a student might explain that they are willing to explore AI-assisted coding but would first like confirmation that the proposed data processing is consistent with the approved protocol.
Where possible, the discussion should refer to actual policies or methodological sources.
The objective is to resolve the research problem rather than personalize the disagreement.
However, professional communication does not require students to undertake an activity they reasonably believe may violate binding requirements while authorization remains unresolved.
Should Students Document the Supervisor's Instruction?
Documentation may be useful when a disagreement concerns consequential research procedures or compliance requirements.
A concise record can identify the proposed AI activity, the concern raised, the guidance received, and any subsequent clarification.
For example, after a supervisory meeting, a student may summarize the agreed data-processing approach in an appropriate project record.
Documentation should be factual and proportionate.
It should not involve unauthorized recording, disclosure of confidential communications, or collection of unrelated personal information.
The purpose is to preserve an accurate understanding of research decisions and responsibilities.
When Should Students Seek Institutional Guidance?
Institutional consultation may be appropriate when the disagreement cannot be resolved through ordinary supervisory discussion or involves obligations outside the supervisor's authority.
Relevant offices may include the graduate program, research ethics committee, data-protection office, research integrity office, or designated student adviser.
The appropriate route depends on the issue and the institution's governance structure.
For example, uncertainty about participant consent and external data processing may be best addressed by the responsible research ethics or data-governance authority.
A disagreement about independent thesis competence may require graduate program guidance.
Students should use established channels rather than assume that every disagreement requires a formal misconduct complaint.
What If the Supervisor Continues to Insist?
If the supervisor continues to require an activity that may violate a binding obligation, the student should seek authoritative clarification before proceeding.
The student should identify which requirement is potentially affected and what authorization would be necessary.
Where the institution provides confidential advice or dispute-resolution mechanisms, these may be appropriate initial channels.
If the matter involves an immediate risk to participant confidentiality or safety, prompt contact with the relevant institutional authority may be necessary.
The student should not assume that supervisory insistence resolves an unresolved compliance issue.
Equally, the student should remain open to the possibility that the proposed AI use can be made appropriate through verified safeguards or revised procedures.
Why Power Imbalances Matter
Supervisors may influence thesis approval, research opportunities, recommendation letters, and academic progression.
Students may therefore find it difficult to challenge supervisory instructions.
This power relationship can affect how concerns are communicated and whether students feel able to seek independent guidance.
However, the existence of a power imbalance does not establish that a particular supervisor has acted improperly.
Institutions should provide accessible procedures for resolving legitimate disagreements without requiring students to rely exclusively on the individual whose instruction is being questioned.
Research governance should support both supervisory authority and the student's ability to raise evidence-based concerns.
What If the Student Fears Academic Retaliation?
Students may be concerned that questioning an instruction could affect their academic progress or supervisory relationship.
Where institutional policies provide confidential advice, grievance procedures, or protections against retaliation, students should consult the relevant provisions.
The existence and scope of these protections vary by institution and jurisdiction.
Students should avoid assuming that every concern will automatically qualify for a particular legal or institutional remedy.
Documenting relevant communications accurately and obtaining guidance from an authorized office may help clarify available options.
Any allegation of retaliation should be assessed through the institution's applicable procedures.
What If the Supervisor's AI Recommendation Is Scientifically Reasonable?
A student may initially oppose AI use because the technology is unfamiliar or because they prefer a traditional workflow.
These preferences deserve consideration, but they do not necessarily establish that the proposed assistance is inappropriate.
For example, using AI to generate preliminary code from synthetic data may be methodologically acceptable when the implementation is tested and the student retains analytical understanding.
The supervisor may have a legitimate reason to introduce the tool as part of research training.
Students should evaluate whether their concern can be addressed through instruction, verification, or safeguards.
A defensible objection should identify a substantive educational, methodological, ethical, or compliance issue rather than rely solely on unfamiliarity.
Can Students Refuse Supervisor-Required AI Use?
The answer depends on the circumstances and the applicable institutional framework.
Students should not undertake processing that is prohibited or unauthorized merely because a supervisor requests it.
However, a student may not have unrestricted discretion to reject every permitted research activity or educational requirement.
Where the disagreement concerns academic expectations rather than an apparent compliance violation, the student should seek clarification and use appropriate supervisory or program procedures.
The objective is to resolve the dispute through legitimate authority and evidence rather than unilateral assumptions about either party's rights.
What If the Supervisor Requires AI-Generated Fabrication or Misrepresentation?
A request to fabricate research data, invent participant responses, or describe procedures that were never performed raises a serious research integrity concern.
The problem is the proposed misrepresentation of research evidence, regardless of whether AI is involved.
The student should not knowingly submit fabricated material as genuine research.
Relevant concerns should be reported through appropriate institutional research integrity channels.
The European Code of Conduct for Research Integrity emphasizes honesty and accountability in scholarly conduct.
Formal determinations of misconduct should follow applicable procedures rather than be presumed from an allegation alone.
How Does This Relate to Meaningful Human Oversight?
Human oversight is not meaningful merely because a person approves an AI-generated recommendation.
Reviewers must have sufficient expertise, access to evidence, and authority to question or reject unsuitable outputs.
This applies to supervisors as well as students.
A supervisor's endorsement of an AI recommendation does not automatically establish its methodological validity.
The principles of meaningful human oversight in AI-assisted research require evaluation of the actual evidence and consequences.
Where necessary, qualified independent expertise should be consulted.
How Can Student-Supervisor Agreements Prevent These Disputes?
Early agreement on AI use can reduce uncertainty about permissions, restrictions, and responsibilities.
Students and supervisors should identify which activities may involve AI, which require authorization, and what verification is expected.
The agreement should also explain how disagreements or new AI-related questions will be resolved.
Such arrangements should remain consistent with institutional rules and research ethics approvals.
The broader practice of establishing clear supervisory expectations for AI use can support transparent and defensible research decisions.
However, a supervisory agreement cannot authorize conduct prohibited by binding requirements.
Why Responsible Research Sometimes Requires Questioning an Instruction
Research integrity involves more than following directions.
Researchers must evaluate whether their methods, evidence, and reporting practices satisfy relevant scientific and ethical obligations.
Students are developing this professional judgment while working within supervisory relationships.
Appropriately questioning an AI-related instruction can therefore be part of research competence.
The challenge is to distinguish justified concern from unsupported resistance and to use institutional procedures when authority is uncertain.
Respectful disagreement and scholarly accountability are not inherently incompatible.
