AI Usage Level Rubric
This rubric defines five levels of AI involvement in research tasks, from no AI usage to complete delegation. Research teams should use this framework to discuss and establish appropriate boundaries for their specific context, considering factors such as disciplinary norms, methodological requirements, and the nature of the doctoral contribution.
1
No AI Usage
The researcher completes the task entirely independently without any AI assistance. This is appropriate for tasks where unassisted human performance is essential to the integrity of the research or the development of the researcher, or where AI involvement would be inappropriate or impossible.
Human-only
Unassisted
Independent
2
Light Assistance
AI serves as a reference tool or provides minor support. The researcher maintains full intellectual ownership and control, using AI only for explanations, clarifications, checking work, or overcoming minor obstacles. The AI's role is comparable to consulting a textbook or dictionary.
Reference
Clarification
Checking
Explanation
3
Moderate Collaboration
AI provides substantive suggestions, structures, or content that the researcher evaluates, adapts, and integrates. The researcher remains the decision-maker and author, but AI contributions meaningfully shape the output. This is comparable to receiving detailed feedback from a knowledgeable colleague.
Suggestions
Structuring
Feedback
Collaboration
4
Substantial Delegation
AI produces significant portions of the work product, which the researcher reviews, verifies, and takes responsibility for. The researcher's role shifts toward direction, curation, and quality assurance rather than primary production. Intellectual ownership becomes shared or ambiguous.
AI-generated
Human-reviewed
Curation
Verification
5
Full Offload
AI completes the task with minimal or no human input, verification, or intellectual engagement. The researcher accepts AI outputs without substantive evaluation or contribution. This level raises significant questions about authorship, learning, and the integrity of the doctoral contribution.
Automated
Unverified
Hands-off
Delegated
How to Use This Framework
This framework is designed to facilitate discussion within research teams about appropriate AI use. It is not prescriptive—what is appropriate will vary by discipline, methodology, institution, and the specific nature of the research.
For Supervisors
Use this framework to:
- Establish clear expectations with students early
- Discuss AI use as part of regular supervision
- Identify tasks where AI assistance is encouraged vs. discouraged
- Consider disciplinary and methodological implications
For Researchers
Use this framework to:
- Reflect on your current AI usage patterns
- Discuss boundaries proactively with supervisors
- Document AI use for transparency
- Ensure you are developing necessary skills
For Institutions
Use this framework to:
- Develop discipline-specific guidance
- Inform policy development
- Support training and development programmes
- Address assessment and integrity concerns
Phase 1
Orientation and Scoping
This initial phase establishes the intellectual foundations and direction of the project, including understanding
the research area, identifying gaps, and formulating research questions.
Reviewing and understanding the inherited or initial research proposal
+
1
Read and analyse the proposal independently without AI assistance
2
Using AI to explain unfamiliar terminology or concepts in the proposal
3
Asking AI to summarise the proposal's key arguments and identify assumptions
4
Having AI critically analyse the proposal's strengths, weaknesses, and unstated implications
5
Asking AI to produce a comprehensive critical evaluation with recommendations
Reframing the research proposal to align with researcher's interests and perspective
+
1
Develop reframing entirely through personal reflection and supervisor discussion
2
Using AI as a sounding board to articulate half-formed ideas
3
Asking AI to suggest angles or framings that connect the inherited proposal to the researcher's interests
4
Having AI draft alternative framings or revised problem statements
5
Asking AI to rewrite the proposal from scratch based on stated interests
Negotiating modifications to scope or direction with supervisors and funders
+
1
Prepare for and conduct negotiations entirely independently
2
Using AI to rehearse arguments or anticipate counterarguments
3
Having AI draft talking points or a briefing document for meetings
4
Asking AI to draft formal communications proposing changes
Level 5 not applicable: negotiation inherently requires human participation and relationship management
Conducting preliminary literature searches to map the field
+
1
Conduct all searches manually using databases and citation tracking
2
Using AI to suggest search terms or databases to try
3
Asking AI to identify key authors, papers, or debates in an area
4
Having AI produce annotated reading lists with summaries of each source
5
Asking AI to produce a comprehensive map of the field with no human searching
Identifying key theorists, debates, and schools of thought
+
1
Identify these through independent reading and discussion with experts
2
Using AI to explain the positions of specific theorists
3
Asking AI to outline major debates and where scholars are positioned within them
4
Having AI produce a detailed intellectual genealogy of the field
5
Relying entirely on AI's characterisation of the field without independent reading
Recognising gaps, contradictions, or underexplored areas in existing knowledge
+
1
Identify gaps through independent critical reading and analysis
2
Using AI to check whether a perceived gap is genuine or already addressed
3
Asking AI to identify commonly noted gaps in the literature
4
Having AI analyse a body of literature to surface contradictions and omissions
5
Asking AI to identify the research gap for the project without researcher input
Formulating or refining research questions or hypotheses
+
1
Develop questions entirely through researcher reflection and supervisor discussion
2
Using AI to check whether questions are clearly worded and appropriately scoped
3
Asking AI to suggest refinements or sub-questions
4
Having AI generate alternative formulations of the research question
5
Asking AI to propose the research questions based on identified gaps
Developing the theoretical or conceptual framework
+
1
Develop framework through independent reading and intellectual work
2
Using AI to explain theoretical concepts or compare frameworks
3
Asking AI to suggest which theories might be applicable to the research problem
4
Having AI draft a conceptual framework showing relationships between concepts
5
Asking AI to select and fully articulate the theoretical framework for the study
Identifying relevant disciplinary and interdisciplinary connections
+
1
Identify connections through broad reading and discussion with scholars
2
Using AI to explain how a concept is used differently across disciplines
3
Asking AI to identify adjacent fields that address similar questions
4
Having AI map interdisciplinary connections and suggest literatures to explore
5
Relying on AI to determine which disciplines are relevant without researcher judgment
Writing or substantially revising the formal research proposal
+
1
Write the proposal entirely independently
2
Using AI to proofread or improve clarity of prose
3
Asking AI to suggest structural improvements or identify weak arguments
4
Having AI draft sections of the proposal based on researcher notes
5
Asking AI to write the complete proposal from a brief description
Presenting and defending the proposal at confirmation or upgrade reviews
+
1
Prepare and deliver entirely independently
2
Using AI to anticipate likely questions from reviewers
3
Having AI help prepare responses to potential challenges
4
Asking AI to draft the presentation script or slides
Level 5 not applicable: presentation and defence require human presence and understanding
Phase 2
Research Planning and Design
This phase develops the practical and methodological architecture of the project. Note that ethics approval
often creates a chicken-and-egg situation: approval requires a finalised design, but design decisions
sometimes need to be tested in ways that would require approval.
Research Planning
Developing a multi-year project timeline with key milestones
+
1
Develop timeline entirely based on own judgment and supervisor input
2
Using AI to check whether proposed timelines are realistic based on typical PhD patterns
3
Asking AI to suggest milestones commonly used in similar projects
4
Having AI draft a complete timeline based on project parameters
5
Asking AI to produce the project plan with no researcher input on priorities or constraints
Breaking the project into discrete work packages or task clusters
+
1
Decompose work entirely through researcher analysis
2
Using AI to check whether task breakdown is logical and complete
3
Asking AI to suggest how similar projects typically decompose work
4
Having AI produce a full work breakdown structure from project description
5
Relying entirely on AI-generated task structure without researcher validation
Creating forecasts for each phase (time, resources, access requirements)
+
1
Develop forecasts based on own experience and expert consultation
2
Using AI to sanity-check time estimates
3
Asking AI what resources are typically needed for specific methods
4
Having AI produce detailed resource forecasts for each phase
5
Accepting AI forecasts without adjustment for local context or researcher circumstances
Identifying dependencies between tasks
+
1
Map dependencies through own analysis of project requirements
2
Using AI to check whether identified dependencies are complete
3
Asking AI to highlight dependencies commonly overlooked
4
Having AI map all dependencies and produce a critical path analysis
5
Relying on AI to determine project sequencing entirely
Building in contingency time for setbacks and iteration
+
1
Plan contingencies based on own risk assessment
2
Using AI to suggest what typically goes wrong in similar projects
3
Asking AI to recommend contingency allowances for different task types
4
Having AI produce a risk-adjusted timeline with built-in buffers
5
Accepting AI contingency planning without researcher judgment about likely risks
Planning for skills development and training needs
+
1
Identify training needs through self-assessment and supervisor guidance
2
Using AI to explain what skills are needed for specific methods
3
Asking AI to identify skill gaps based on project requirements
4
Having AI produce a personalised training plan
5
Following AI training recommendations without considering existing competencies
Research Design
Selecting and justifying the overarching methodological approach
+
1
Select methodology through independent study and expert guidance
2
Using AI to explain different methodological traditions
3
Asking AI to compare approaches and their fit with research questions
4
Having AI recommend and justify a methodological approach
5
Adopting AI's methodological recommendation without independent evaluation
Designing data collection instruments
+
1
Design instruments entirely based on methodological expertise
2
Using AI to check instrument wording for clarity or bias
3
Asking AI to suggest questions or items used in validated instruments
4
Having AI draft complete instruments based on research questions
5
Using AI-generated instruments without researcher refinement or piloting
Developing sampling or recruitment strategies
+
1
Develop strategy based on methodological principles and local knowledge
2
Using AI to explain different sampling approaches
3
Asking AI to suggest appropriate sample sizes or recruitment channels
4
Having AI design the complete sampling strategy
5
Implementing AI sampling design without consideration of access realities
Planning data management, storage, and security procedures
+
1
Plan procedures based on institutional guidance and expert consultation
2
Using AI to explain data management requirements and best practices
3
Asking AI to identify risks and suggest security measures
4
Having AI draft a complete data management plan
5
Implementing AI data management plan without institutional consultation
Ethics and Approvals
Identifying ethical considerations relevant to the research
+
1
Identify considerations through ethical training and reflection
2
Using AI to explain ethical principles applicable to research
3
Asking AI to identify ethical issues commonly arising in similar studies
4
Having AI produce a comprehensive ethical analysis of the project
5
Relying on AI identification of ethical issues without researcher reflection
Completing institutional ethics application forms
+
1
Complete forms entirely based on own understanding
2
Using AI to clarify what is being asked in form questions
3
Asking AI to suggest how to articulate ethical safeguards
4
Having AI draft responses to ethics application questions
5
Submitting AI-drafted ethics application without researcher verification
Preparing participant information sheets and consent forms
+
1
Draft documents based on templates and ethical principles
2
Using AI to check readability and clarity of documents
3
Asking AI to suggest standard content for these documents
4
Having AI draft complete information sheets and consent forms
5
Using AI-generated documents without checking accuracy or appropriateness
Addressing data protection and GDPR requirements
+
1
Address requirements through training and institutional support
2
Using AI to explain data protection requirements
3
Asking AI to identify which requirements apply to the project
4
Having AI draft data protection documentation
5
Implementing AI data protection advice without legal consultation
Phase 3
Data Collection and Generation
The nature of this phase varies enormously by discipline but involves gathering the empirical material
that will form the basis of the contribution to knowledge.
Preparation
Procuring equipment, materials, or software
+
1
Research options independently and consult with experts
2
Using AI to compare options or explain specifications
3
Asking AI to recommend equipment for specific purposes
4
Having AI produce procurement specifications and supplier comparisons
5
Purchasing based entirely on AI recommendations without verification
Recruiting participants or securing access to sites and archives
+
1
Conduct all recruitment and access negotiation personally
2
Using AI to draft recruitment materials
3
Having AI suggest recruitment strategies or access approaches
4
Asking AI to produce recruitment campaign content
Level 5 not applicable: recruitment and access negotiation require human interaction and relationship-building
Training in specific techniques or equipment operation
+
1
Learn entirely through formal training and expert instruction
2
Using AI to explain techniques or answer questions during learning
3
Asking AI to create practice exercises or knowledge checks
4
Having AI serve as primary training resource
Level 5 not applicable for hands-on skills; partial offload possible only for conceptual learning
Active Data Collection
Conducting laboratory experiments or computational simulations
+
1
Conduct all experiments manually with full researcher control
2
Using AI to monitor for anomalies or suggest parameter adjustments
3
Having AI automate routine aspects of experimental procedures
4
Asking AI to control experimental variables and make real-time decisions
5
AI conducting experiments autonomously without human oversight
Carrying out interviews, focus groups, or ethnographic engagement
+
1
Conduct all qualitative data collection personally
2
Using AI to refine questions between interviews
3
Having AI suggest follow-up probes based on participant responses
4
Using AI to conduct structured interviews
5
AI conducting qualitative interviews without human interviewer
Gathering archival, textual, or visual materials
+
1
Locate and select all materials through independent research
2
Using AI to help locate relevant materials in large archives
3
Having AI filter and prioritise materials based on relevance criteria
4
Asking AI to systematically extract materials meeting specified criteria
5
AI assembling the complete corpus without human selection
Engaging in practice-based or creative research activities
+
1
Conduct all creative work independently
2
Using AI as one tool among many in creative process
3
Collaborating with AI as creative partner with shared authorship
4
Having AI generate creative outputs that researcher curates or refines
5
AI producing creative work with researcher claiming authorship
Data Management
Transcribing audio or video recordings
+
1
Transcribe all recordings manually
2
Using AI transcription with human verification of accuracy
3
Using AI transcription with spot-checking
4
Using AI transcription with minimal review
5
Using AI transcription without any verification
Anonymising or pseudonymising data as required
+
1
Anonymise all data manually with careful checking
2
Using AI to check anonymisation completeness
3
Having AI identify potentially identifying information
4
Asking AI to perform anonymisation automatically
5
AI anonymising data without human verification of adequacy
Phase 4
Analysis and Interpretation
This phase transforms raw data into meaningful findings and situates them within broader scholarly conversations.
Data Processing
Cleaning and preparing datasets for analysis
+
1
Clean all data manually with full documentation
2
Using AI to identify data quality issues
3
Having AI suggest cleaning procedures
4
Asking AI to clean data according to specified rules
5
AI cleaning data autonomously without researcher verification
Coding qualitative data using predetermined or emergent approaches
+
1
Code all data manually through close reading
2
Using AI to check coding consistency
3
Having AI suggest codes for difficult passages
4
Asking AI to code data with researcher reviewing a sample
5
AI coding all data without systematic human verification
Analytical Work
Applying statistical techniques appropriate to the data and questions
+
1
Conduct all statistical analysis independently
2
Using AI to explain statistical procedures
3
Having AI check statistical code or suggest appropriate tests
4
Asking AI to conduct statistical analysis
5
AI performing all statistical analysis without researcher understanding
Conducting thematic, discourse, narrative, or other qualitative analyses
+
1
Conduct all qualitative analysis through close engagement with data
2
Using AI to help articulate emerging themes
3
Having AI identify patterns across coded data
4
Asking AI to conduct systematic analysis and produce findings
5
AI performing qualitative analysis without researcher interpretive engagement
Iterating between data and emerging interpretations
+
1
Develop all interpretations through researcher reflection
2
Using AI to test interpretations against evidence
3
Having AI suggest alternative interpretations
4
Asking AI to develop interpretations from data
5
AI generating interpretations without researcher intellectual engagement
Interpretation and Synthesis
Interpreting findings in relation to the original research questions
+
1
Develop interpretation entirely through researcher analysis
2
Using AI to check whether findings address questions
3
Having AI suggest interpretive connections
4
Asking AI to draft interpretive account
5
AI producing interpretation without researcher intellectual authorship
Developing explanatory accounts or theoretical contributions
+
1
Develop theoretical contributions through independent intellectual work
2
Using AI as sounding board for emerging explanations
3
Having AI articulate or refine explanatory accounts
4
Asking AI to develop theoretical contribution
5
AI producing theoretical contribution without researcher intellectual ownership
Acknowledging limitations and alternative interpretations
+
1
Identify limitations through critical self-reflection
2
Using AI to identify potential limitations
3
Having AI generate alternative interpretations
4
Asking AI to produce comprehensive limitations discussion
5
AI identifying limitations without researcher critical reflection
Phase 5
Writing and Synthesis
The researcher transforms analysis into a coherent, scholarly document that makes an original contribution to knowledge.
Drafting
Writing initial drafts of individual chapters or sections
+
1
Write all drafts independently
2
Using AI to overcome writer's block or clarify expression
3
Having AI expand bullet points into prose
4
Asking AI to draft sections based on notes and findings
5
AI writing chapters without researcher drafting
Drafting the literature review and situating the study in context
+
1
Write literature review entirely based on own reading
2
Using AI to improve flow and clarity
3
Having AI help synthesise across sources
4
Asking AI to draft literature review from annotated bibliography
5
AI producing literature review without researcher reading of sources
Developing discussion chapters that interpret and synthesise
+
1
Write discussion entirely through own intellectual engagement
2
Using AI to strengthen argumentation
3
Having AI suggest interpretive connections
4
Asking AI to draft discussion chapters
5
AI producing discussion without researcher intellectual contribution
Revision and Refinement
Responding to supervisory feedback on drafts
+
1
Address all feedback through own revision
2
Using AI to clarify confusing feedback
3
Having AI suggest how to address feedback
4
Asking AI to revise text based on feedback
5
AI implementing feedback without researcher engagement with critique
Revising arguments for clarity, coherence, and strength
+
1
Revise all arguments through own critical analysis
2
Using AI to identify unclear passages
3
Having AI suggest revisions to strengthen argument
4
Asking AI to revise argumentation throughout
5
AI revising arguments without researcher authorial control
Finalisation
Proofreading and copy-editing the complete document
+
1
Proofread entirely manually
2
Using AI to catch errors and typos
3
Having AI copy-edit for style and grammar
4
Asking AI to proofread and correct entire document
5
AI finalising text without human review
Compiling bibliography and checking citation accuracy
+
1
Compile and check bibliography entirely manually
2
Using AI to identify formatting inconsistencies
3
Having AI check citations against sources
4
Asking AI to compile and verify bibliography
5
AI producing bibliography without human verification of accuracy
Phase 6
Dissemination and Defence
The research is shared with scholarly communities and subjected to formal examination.
Conference and Publication Activity
Writing and submitting conference abstracts
+
1
Write all abstracts independently
2
Using AI to sharpen abstract language
3
Having AI suggest abstract structure
4
Asking AI to draft abstract from paper
5
AI writing and submitting abstracts without researcher authorship
Writing journal articles or book chapters derived from thesis
+
1
Write all publications independently
2
Using AI to adapt thesis prose for publication
3
Having AI restructure material for journal format
4
Asking AI to draft article from thesis content
5
AI producing publications without researcher authorship
Navigating peer review and responding to reviewer comments
+
1
Respond to all reviews independently
2
Using AI to understand reviewer concerns
3
Having AI suggest response strategies
4
Asking AI to draft response letter and revisions
5
AI managing peer review without researcher engagement with critique
Thesis Examination
Preparing for the viva voce examination
+
1
Prepare entirely through own review and human mock vivas
2
Using AI to identify likely areas of questioning
3
Having AI generate practice questions
4
Asking AI to conduct mock viva
Level 5 not applicable: preparation must develop researcher's own understanding for the actual examination
Attending and participating in the formal examination
+
All levels above 1 not applicable: the viva examination requires unassisted human participation. The researcher must be able to discuss and defend their work without AI assistance.
Completing required corrections or revisions post-viva
+
1
Complete all corrections independently
2
Using AI to clarify what corrections require
3
Having AI suggest approaches to corrections
4
Asking AI to draft corrections
5
AI making corrections without researcher engagement
Throughout
Cross-cutting Activities
These activities occur throughout the PhD and support or intersect with all phases.
Completing progress reports and annual reviews
+
1
Write all reports based on own reflection
2
Using AI to improve report clarity
3
Having AI structure reports
4
Asking AI to draft progress reports
5
AI writing reports without researcher reflection
Undertaking required and elective training and skills development
+
1
Learn entirely through formal courses and human instruction
2
Using AI to supplement formal training
3
Using AI as tutor for specific skills
4
Using AI as primary learning resource
Level 5 not applicable: skills require human development and cannot be fully offloaded
Teaching, demonstrating, or other departmental duties
+
1
Prepare and deliver teaching entirely independently
2
Using AI to prepare teaching materials
3
Having AI help design sessions
4
Asking AI to produce teaching content
Level 5 not applicable: teaching requires human presence and engagement with students
Attending seminars, workshops, and reading groups
+
1
Prepare and participate based on own reading
2
Using AI to prepare for discussions
3
Having AI summarise readings
4
Relying on AI summaries rather than reading
Level 5 not applicable: attendance and participation require human presence
Networking and building scholarly relationships
+
1
Build all relationships through personal engagement
2
Using AI to draft initial communications
3
Having AI research potential contacts
Levels 4-5 not applicable: relationships require human cultivation and cannot be delegated
Managing funding, budgets, and administrative requirements
+
1
Manage all administration independently
2
Using AI to track expenses or explain requirements
3
Having AI prepare budget reports
4
Asking AI to manage administrative tasks
5
AI managing administration without researcher oversight
Factors Affecting Appropriate AI Usage
What constitutes appropriate AI usage varies significantly based on context. Research teams should consider these factors when establishing boundaries for their specific situation.
Disciplinary Norms
Different fields have different expectations about tools, authorship, and intellectual contribution.
- STEM fields may accept more computational assistance
- Humanities may emphasise individual interpretation
- Social sciences vary by methodological tradition
- Creative disciplines have unique questions about AI co-creation
Methodological Paradigm
The epistemological foundations of the research affect what AI involvement means.
- Interpretive research may require human meaning-making
- Positivist research may accommodate more automation
- Critical approaches may require researcher reflexivity
- Mixed methods need consideration at each stage
Type of Doctoral Programme
The purpose and structure of the degree affects expectations.
- Traditional PhDs emphasise independent scholarship
- Professional doctorates may allow more practical AI use
- Practice-based doctorates raise unique questions
- Collaborative programmes may have different norms
Career Development
The PhD is a training programme, not just a research project.
- Which skills must the researcher develop?
- What competencies will future employers expect?
- How will AI proficiency itself be valued?
- What forms of expertise remain distinctly human?
Ethical Considerations
AI use raises specific ethical questions in research contexts.
- Transparency and disclosure requirements
- Data privacy when using cloud AI services
- Bias and fairness in AI-assisted analysis
- Environmental impact of AI computation
Institutional Requirements
Universities and funders may have specific policies.
- Existing academic integrity policies
- Funder requirements for transparency
- Journal and publisher policies
- Professional body guidelines
Questions for Research Teams
Use these questions to guide discussions about AI usage in your specific research context.
About the Task
- Is this task central to the doctoral contribution?
- Does it require skills the researcher must develop?
- Would AI assistance affect the validity of the research?
- Are there accuracy risks if AI makes errors?
About Transparency
- How will AI use be documented and disclosed?
- What would examiners, reviewers, or employers expect?
- Does the researcher understand what the AI produced?
- Can they explain and defend all aspects of the work?
About Learning
- What is the researcher missing by not doing this themselves?
- Will they need this skill in future roles?
- Is struggling with this task part of intellectual development?
- Could AI assistance prevent deeper understanding?