03 · What You Need to Know
Why AI Governance Should Reflect the Purpose of Academic Work
Academic Research Activities Do Not All Serve the Same Purpose
A research methods assignment is primarily an educational activity, even when it resembles authentic research.
A thesis is both a scholarly investigation and evidence that a student has achieved the competencies expected by a degree program.
A journal article communicates research findings to the scholarly community and creates responsibilities for the authors, editors, and publisher.
These purposes overlap but are not identical.
For example, an instructor may prohibit AI-generated statistical interpretations during an assessment because the objective is to measure students' independent reasoning.
The same student may appropriately use AI to generate preliminary analytical code during thesis research, provided that the procedure is permitted, verified, and understood.
Different rules can therefore be justified without abandoning common standards of scholarly integrity.
What Should Remain Consistent Across All Three Contexts?
Differences in permitted assistance should not imply that research integrity changes according to the academic setting.
The European Code of Conduct for Research Integrity, revised by ALLEA in 2023, identifies reliability, honesty, respect, and accountability as fundamental principles.
These principles provide a common foundation for AI-assisted academic work.
Students and researchers should not fabricate evidence, misrepresent procedures, disclose confidential information without authorization, or present unsupported claims as established findings.
They should also follow the applicable rules governing attribution, documentation, and disclosure.
Permitting AI assistance does not transfer responsibility for the resulting work to the system.
How Should AI Rules Differ in Coursework?
Coursework rules should follow the intended learning outcomes.
If an assignment assesses independent literature searching, students may be required to construct search expressions without generative AI assistance.
If an activity teaches AI verification, students may be expected to examine and critique generated references or summaries.
Thus, AI may be prohibited in one activity and required in another within the same course.
The distinction is pedagogically defensible when the instructor explains what each activity measures.
The question of whether AI should be banned during research methods training should therefore be answered through learning objectives rather than a universal prohibition.
How Should AI Rules Differ in Thesis Research?
Thesis research combines authentic scholarship with the assessment of individual research competence.
Students may use software, consult experts, collaborate with research teams, and receive supervisory feedback.
Generative AI may also provide legitimate assistance, depending on institutional requirements.
However, the student must demonstrate the intellectual competencies expected by the degree program.
For example, a thesis student who uses AI to suggest a research design should still be able to explain why the selected methodology is appropriate.
A student who uses AI-generated statistical code should understand the analytical procedure sufficiently to evaluate or obtain qualified verification of its implementation.
Thesis policies should therefore address both the research process and the student's independent contribution.
How Should AI Rules Differ in Published Research?
Published research is governed by journal policies, publisher requirements, disciplinary conventions, and relevant institutional or funder obligations.
The International Committee of Medical Journal Editors states that AI-assisted technologies should not be listed as authors and that human authors remain responsible for the accuracy and integrity of submitted work.
Its recommendations also address disclosure of AI assistance.
Other publishers may distinguish among language editing, manuscript generation, image production, and analytical uses.
These policies can differ and may change over time.
Researchers should consult the current instructions of the intended publication venue rather than assume that permission granted during thesis research automatically extends to journal submission.
Comparing AI Governance Across Academic Contexts
| Governance consideration |
Coursework |
Thesis research |
Published research |
| Primary purpose |
Learning and assessment |
Original scholarship and degree competence |
Communication of accountable research findings |
| AI permissions |
Determined by activity objectives and course rules |
Determined by institutional and supervisory requirements |
Determined by journal, publisher, and other applicable policies |
| Independent competence |
Assessed according to course outcomes |
Central to thesis examination and scholarly contribution |
Relevant to author responsibility and expertise |
| Documentation |
May include learning reflections or assignment declarations |
May require consequential AI-use and research decision records |
Should satisfy applicable disclosure and reproducibility requirements |
| Verification |
Supports learning and accurate submissions |
Protects methodological validity and thesis integrity |
Supports accuracy, integrity, and author accountability |
| Human oversight |
Instructor and student responsibilities |
Student, supervisor, and relevant committees |
Authors, editors, reviewers, and research institutions |
This comparison is a conceptual framework rather than a universal policy statement. Institutions may establish additional requirements or restrictions.
Why Assessment Validity Matters in Coursework
The Standards for Educational and Psychological Testing emphasize that assessment interpretations require appropriate supporting evidence.
In research education, an instructor must determine whether an assignment actually measures the intended competency.
For example, an AI-generated literature synthesis may not adequately demonstrate that a student can independently evaluate original publications.
However, an assignment requiring students to identify inaccuracies in an AI-generated synthesis may validly assess verification skills.
AI rules should therefore follow the construct being assessed.
This is central to assessing students' own research abilities when AI assistance is allowed.
Why Thesis Research Requires More Than Coursework Permission
Permission to use AI in a course does not necessarily authorize the same use in a thesis.
Thesis research may involve participant information, unpublished institutional data, research ethics approvals, and obligations to collaborators.
These considerations may restrict the use of external AI services even when comparable tools are permitted for classroom exercises.
For example, uploading synthetic interview transcripts during a methods class differs from uploading identifiable participant interviews collected under a confidential research protocol.
Students should consult the actual research ethics approval, consent arrangements, institutional policies, and data-handling requirements.
Educational permission and research authorization are distinct.
Why Publication Permission Does Not Automatically Authorize Student Use
A journal may permit certain forms of AI-assisted language editing.
That permission does not override an instructor's assessment rules or a university's thesis requirements.
For example, a publisher may accept a manuscript edited with AI while a research methods instructor prohibits such assistance during an independent writing assessment.
The difference arises because the journal evaluates a scholarly submission, while the instructor may be assessing the student's writing competence.
Students should therefore avoid treating publication policies as universal permissions for academic coursework.
How Should Institutions Resolve Conflicting AI Rules?
Conflicts may arise when an instructor, supervisor, research ethics committee, institution, and publisher establish different expectations.
Researchers should identify which requirements govern the specific activity.
For example, a publisher's permission to use a particular AI service does not eliminate institutional confidentiality restrictions.
Similarly, a supervisor's informal approval may not override a formally adopted university policy.
Where requirements are ambiguous, students should seek clarification from the responsible authority before undertaking consequential AI-assisted work.
Conflicting requirements should be documented and resolved through appropriate institutional processes rather than through assumptions about which policy is more convenient.
Should AI Disclosure Requirements Be Identical?
Not necessarily.
In coursework, a brief declaration may help instructors understand how assistance affected an assignment.
In thesis research, documentation may need to explain how AI influenced methodological decisions, analytical procedures, or interpretation.
In publication, disclosure must follow the applicable journal or publisher policy.
The distinction between documenting AI use during student research and formally disclosing assistance is important.
Documentation supports reconstruction and evaluation of the research process, while disclosure communicates relevant information to a specified audience.
Neither automatically substitutes for the other.
How Should AI-Generated Code Be Governed?
AI-generated analytical code illustrates why context matters.
In an introductory programming assessment, students may be required to write code independently to demonstrate foundational competence.
During thesis research, AI-generated code may be permitted if the student understands the procedure, tests the implementation, and preserves appropriate computational records.
For publication, the final analysis must be accurately reported and meet relevant reproducibility and disclosure requirements.
Across all contexts, successful code execution does not establish scientific correctness.
Researchers should verify that the implementation matches the intended analytical method.
Should AI Be Allowed to Draft Scientific Arguments?
AI-generated prose can range from routine language editing to substantive argument construction.
These activities raise different educational and scholarly concerns.
During coursework, allowing AI to construct the central argument may undermine an assessment of scientific reasoning.
During thesis research, extensive AI drafting may make it difficult to establish the student's independent intellectual contribution.
In publication, authors remain accountable for the accuracy and integrity of the submitted text, regardless of permitted assistance.
Researchers should distinguish improving expression from delegating the development of evidence-based claims.
Where AI contributes substantively, appropriate verification and disclosure become particularly important.
Why a Single University-Wide Rule May Be Insufficient
Institutions may reasonably establish common minimum standards concerning confidentiality, transparency, and research integrity.
However, different disciplines and assessment activities may require more specific rules.
A clinical research project involving sensitive participant information may require stricter controls than a classroom exercise using publicly available literature.
A programming course may restrict generated code during particular assessments, while an advanced research seminar may explicitly evaluate AI-assisted code review.
A useful institutional framework can establish shared principles while allowing justified, clearly communicated task-specific requirements.
How Can AI Rules Remain Fair to Students?
Students should know which assistance is permitted before completing the relevant work.
Rules should define the activity, the restricted or permitted AI functions, documentation expectations, and applicable consequences.
Policies should also account for accessibility and approved accommodations.
Ambiguous or retrospectively imposed requirements can create avoidable disputes about academic integrity.
Fair governance requires clear expectations and consistent application of institutional procedures.
What Should Happen When AI-Assisted Coursework Becomes Published Research?
Student work may evolve into a thesis chapter, conference paper, or journal article.
When this occurs, the researchers should review the work against the requirements of its new context.
For example, AI assistance permitted during an assignment may need to be disclosed differently in a manuscript.
AI-generated references and interpretations should be independently verified before publication.
Authorship and contribution decisions should reflect applicable scholarly standards rather than the original classroom arrangements alone.
Moving from coursework to publication changes the responsibilities associated with the work, even when the underlying research material remains similar.
