Final Research-Integrity and Accountability Reminder
Table of Contents
- 1. Lesson Overview
- 2. Learning Outcomes
- 3. Before You Begin
- 4. Retrieval Practice
- 5. Opening Scenario
- 6. Core Concepts
- 7. RIGOUR Walkthrough (Complete Framework)
- 8. Research-Process Walkthrough
- 9. Worked Example
- 10. Guided Learning Activities
- 11. Exercises
- 12. Homework Problems
- 13. Mini-Project
- 14. Self-Check Quiz
- 15. Assessment Connection
- 16. Assignment 1 Journal Connection
- 17. Key Takeaways
- 18. Key Terms
- 19. Looking Ahead
- 20. Sources and Further Reading
1. Lesson Overview
This culminating lesson brings together the entire Unit 13 sequence. You will apply the RIGOUR framework—the author-approved model for responsible AI-supported research—to govern every stage of your research process. You will learn to document all AI interactions and human decisions, preserve provenance to support reproducibility and auditability, and disclose material AI use accurately and proportionately.
You will also learn to distinguish reproducibility from validity and correctness—a reproducible process can still be wrong. You will identify when to correct, reanalyse, escalate, withdraw, or stop research, and you will integrate all the frameworks from Units 10–13 (TRACE, CLAIM, LEARN, CREATE, PROTECT, FAIRER, PCAFE-V, SCOPE, PARTNER) with RIGOUR. Finally, you will complete Module 4 and prepare to move to Module 5: AI for Professional Practice and Everyday Life.
Research that is not documented, reproducible, or transparent is not trustworthy—regardless of how "good" the results look. The ultimate responsibility for scholarly integrity rests with you, the researcher. This lesson equips you to exercise that responsibility with confidence and rigor.
2. Lesson Learning Outcomes
By the end of this lesson, you will be able to:
- Apply the author-approved RIGOUR framework to an AI-supported research process.
- Create a research provenance record documenting questions, prompts, sources, data, code, outputs, decisions, and revisions.
- Assess reproducibility and its limitations in AI-supported research.
- Distinguish reproducibility from validity and correctness.
- Prepare an accurate and proportionate AI-use disclosure.
- Integrate TRACE, CLAIM, LEARN, CREATE, PROTECT, FAIRER, PCAFE-V, SCOPE, and PARTNER with RIGOUR.
- Identify correction, reanalysis, escalation, withdrawal, and stop conditions.
- Evaluate whether AI improved the research process or merely accelerated output production.
- Preserve human scholarly judgment and accountability.
- Connect Unit 13 learning to Assignment 1, Assignment 5, and later applied work.
3. Before You Begin
- Assigned reading: Read the sections of Chapter 14 on reproducibility, documentation, and disclosure. Review Chapter 11 (Copyright, Authorship, Attribution, and Academic Integrity) and Appendix H (Sample AI Use Statements).
- Prerequisite knowledge: You should have completed Lessons 13.1–13.3 and be familiar with research planning, literature work, and analysis verification.
- RIGOUR framework: The author-approved RIGOUR expansion has not been supplied; we will use the operational definition provided in the course materials. Replace placeholders when the exact expansion becomes available.
- Documentation tools: You may use any method for documenting your research (e.g., spreadsheets, lab notebooks, version control). The lesson provides templates.
- Optional AI tool: You may use an AI system to help draft documentation, but remember that you are responsible for verifying and finalising all disclosures.
- Non-platform alternative: All activities can be completed without AI or paid tools.
- Academic integrity: Disclosures must be truthful and complete. Do not omit or understate AI use.
- Preparation time: Approximately 15–20 minutes for reading and retrieval practice.
4. Retrieval Practice
These questions activate knowledge from Lessons 13.1–13.3. Attempt each one before revealing the feedback.
Question 1: In Lesson 13.1, you learned to formulate a research question. Why is the research question the foundation for reproducibility?
Show answer
Answer: The research question defines what the study aims to answer. Without a clear question, you cannot assess whether the methods, data, and analysis are appropriate. Reproducibility depends on knowing exactly what was being tested or explored.
Explanation: A well-documented question is the anchor for all subsequent decisions. If the question changes, reproducibility is compromised.
Question 2: From Lesson 13.2, what is the difference between a literature map and a scholarly synthesis? How does this distinction affect reproducibility?
Show answer
Answer: A literature map organises sources thematically; a synthesis integrates them into a coherent narrative. Reproducibility requires that the map and synthesis are traceable to the original sources. If the synthesis is based on unverified sources, it cannot be reproduced.
Explanation: Documentation of the search, verification, and mapping steps is essential for others to reconstruct the review.
Question 3: In Lesson 13.3, you learned to verify AI-generated code and calculations. How does this verification relate to reproducibility?
Show answer
Answer: Verification ensures that the code and calculations are correct and can be run by another researcher. If the code contains errors or is not documented, the analysis cannot be reproduced.
Explanation: Reproducibility requires that the code, data, and environment are available and that the results can be independently obtained.
A process can be reproducible (another person can follow the same steps and get the same results) but still be incorrect (e.g., due to a flawed method or wrong assumptions). Reproducibility is about transparency and consistency, not about truth.
5. Opening Scenario
Scenario: Dr. J. Kim is a social scientist who conducted a mixed-methods study on AI adoption in non-profit organisations. She used AI for transcription, brainstorming codes, drafting statistical code, and generating a first draft of her report. However, she:
- Cannot identify the model version of the AI system she used.
- Did not retain the prompts she used.
- Included a source list that contains several fabricated citations (discovered after submission).
- Used a retrieval-augmented generation tool whose retrieved results changed between sessions.
- Drafted code with AI but did not test it on synthetic data.
- Modified her analysis plan after seeing results (e.g., added a subgroup analysis).
- Did not disclose any AI use in her final report.
Task: Can Dr. Kim's research process be reconstructed? What must she correct? What are the implications for reproducibility, disclosure, and scholarly responsibility?
Show guided analysis
Reconstruction assessment: No, the process cannot be reliably reconstructed because:
- Missing model version and prompts prevent replication of AI-assisted steps.
- Retrieved results vary, so others cannot reproduce the same literature list.
- Untested code may contain errors.
- Post-hoc analysis changes (methodological drift) are undocumented.
Corrections required:
- Verify all sources and remove fabricated citations.
- Document the AI system and, if possible, retrieve prompts from logs.
- Test code on synthetic data and document its version.
- Disclose all material AI use.
- If the post-hoc changes are substantial, reanalyse with the original plan or justify the changes.
Implications: Dr. Kim's research is not reproducible, and her failure to disclose AI use violates academic integrity. She must issue a correction or, in severe cases, withdraw the work. Her scholarly reputation is at risk. This scenario underscores the importance of documentation, provenance, and disclosure.
6. Core Concepts
6.1 The RIGOUR Framework (Operational)
RIGOUR is the author-approved model for responsible AI-supported research and scholarly practice. Until the exact expansion is supplied, we use the operational elements defined by the course materials. RIGOUR addresses six governance domains:
- R – Research purpose and question (legitimate, focused, significant).
- I – Information, data, source, and ethics governance (authority, consent, community rights).
- G – Method selection and methodological fit (coherence of question, data, method).
- O – Verification of analysis and interpretation (independent checks, human judgment).
- U – Documentation, reproducibility, and provenance (records for reconstruction).
- R – Uncertainty, disclosure, human responsibility, and review (limitations, transparency, accountability).
RIGOUR is not a guarantee of validity, ethical approval, reproducibility, or publication quality. It is a governance framework to guide responsible practice.
6.2 Research Provenance
Provenance documents the origin and transformation of every research element:
- Research questions – initial versions and revisions.
- Sources – search strategy, verification records, inclusion/exclusion decisions.
- Data – collection methods, cleaning steps, transformations.
- Prompts – exact prompts used, date/time, model version.
- AI outputs – what was generated and retained.
- Code and calculations – scripts, outputs, test results.
- Categories and themes – codebooks, theme definitions, evolution.
- Figures and tables – how they were produced.
- Interpretations – alternative interpretations considered.
- Decisions – why certain choices were made.
- Revisions – changes to the plan or analysis.
Provenance supports reproducibility, auditability, and accountability.
6.3 Version Control
Record versions of:
- Research question and subquestions.
- Protocol or analysis plan.
- Search strategy (including dates).
- Source lists and literature maps.
- Datasets (raw and processed).
- Code and scripts.
- Prompts and AI outputs.
- Manuscript drafts.
- Disclosure statements.
Use a systematic naming convention and track changes over time.
6.4 Reproducibility, Validity, and Correctness
| Concept | Definition | Example |
|---|---|---|
| Reproducibility | Another researcher can follow your documented process and obtain consistent results. | Someone runs your code on your data and gets the same mean. |
| Validity | The study measures what it claims to measure. | A survey actually measures anxiety, not general distress. |
| Correctness | The conclusions correspond to reality. | The treatment actually reduces symptoms. |
Key insight: A study can be reproducible but invalid (e.g., a flawed survey that gives consistent but wrong results). Reproducibility is necessary for science but not sufficient. Researchers must also ensure validity and correctness through careful design and interpretation.
6.5 AI-Use Disclosure
A scholarly disclosure typically identifies:
- AI system – name, developer, version (where available).
- Dates of use – when the AI was used.
- Research stages affected – planning, discovery, analysis, writing, etc.
- Purpose – what the AI was used for.
- Prompts or prompt categories – representative examples.
- Source and data boundaries – what data were input.
- Outputs retained – which outputs were used.
- Human verification – how outputs were checked.
- Corrections – any errors found and corrected.
- Limitations – what AI could not do.
- Final human responsibility – statement that the researcher takes full responsibility.
Follow applicable institutional, disciplinary, funder, and publisher requirements. If none exist, err on the side of transparency.
6.6 Reproducibility Limits
AI-supported research may be difficult to reproduce because:
- Models change over time (version instability).
- Outputs vary due to stochasticity (temperature settings).
- System instructions (system prompts) are proprietary.
- Proprietary tools are used (closed-source code).
- Prompts are incomplete or not recorded.
- Retrieval results change (dynamic search).
- Context is not preserved (e.g., chat history).
- Model settings (temperature, top-p) are unknown.
- Source access changes (paywalled articles).
Document these limitations explicitly in your research report.
6.7 Correction, Reanalysis, Escalation, Withdrawal, and Stop
- Correction: Fix minor errors (typos, misplaced citations) without changing conclusions.
- Reanalysis: Re-run analyses when errors are found (e.g., incorrect code, missing data) and update conclusions if needed.
- Escalation: Seek guidance from supervisors, ethics boards, statisticians, or community authorities when issues exceed your expertise.
- Withdrawal: Remove a published or submitted work if errors are so severe that the findings are invalid, or if there are ethical violations.
- Stop: Pause or discontinue research when authority is lacking, risks are unacceptable, or evidence cannot be verified.
These actions are signs of scholarly integrity, not failure.
6.8 Integrating TRACE, CLAIM, LEARN, CREATE, PROTECT, FAIRER, PCAFE-V, SCOPE, and PARTNER with RIGOUR
| Framework | Role in RIGOUR |
|---|---|
| TRACE | Identify hallucinations, errors, omissions, and uncertainty in AI outputs. |
| CLAIM | Verify claims, sources, citations, and evidence. |
| LEARN | Ensure AI supports learning without substituting for it. |
| CREATE | Govern authorship, attribution, and originality. |
| PROTECT | Safeguard privacy, confidentiality, and data security. |
| FAIRER | Address bias, fairness, inclusion, and equity. |
| PCAFE-V | Evaluate performance, cost, authority, fairness, ethics, and verification. |
| SCOPE | Define the scope, context, and boundaries of AI use. |
| PARTNER | Allocate responsibilities between human and AI. |
Together, these frameworks provide a comprehensive governance toolkit. RIGOUR integrates them into a coherent research process.
7. RIGOUR Walkthrough (Complete Framework)
This walkthrough applies every RIGOUR element to a synthetic research project. Since the exact author-approved expansion is not supplied, we use the operational elements. Placeholders are used where specific terminology is required.
Definition: A legitimate, clearly defined, appropriately scoped, methodologically answerable, ethically permissible, and significant research question.
Guiding questions: What is the specific problem? What is the intended contribution? Is the question focused? What ethical considerations are embedded?
Example: "How do community health workers perceive the usability of AI translation tools in rural clinics?"
Common misuse: Accepting an AI-generated question without evaluation.
Verification: The researcher must justify the question with evidence and ethics approval.
Escalation/stop: If the question is not ethically defensible, stop or escalate.
Residual limitation: A good question does not guarantee good research.
Definition: All information, data, and sources must be governed by principles of authority, consent, community rights, and intellectual integrity.
Guiding questions: Have I verified every source? Does my coverage reflect diversity? Have I respected community-governed knowledge?
Example: A researcher obtains permission from an Indigenous community before using their language data.
Common misuse: Including unverified AI-generated references.
Verification: Independent verification of sources and data authority.
Escalation/stop: If data authority is unclear, escalate to the community or ethics board.
Residual limitation: Even verified sources may be misinterpreted.
Definition: The research question, data, method, analysis, and interpretation must be coherent.
Guiding questions: Does the method answer the question? Are assumptions met? Is the method justified?
Example: A qualitative study on lived experiences uses interviews, not a survey.
Common misuse: Using a method because an AI recommended it without evaluation.
Verification: Confirm that the method is appropriate for the question and data.
Escalation/stop: If the method is unsuitable, redesign or consult a methodologist.
Residual limitation: A well-fit method may still be poorly implemented.
Definition: All analytical outputs must be independently verified; interpretation must be grounded in evidence.
Guiding questions: Have I checked AI-generated codes/themes? Have I verified calculations? Have I considered alternatives?
Example: A researcher compares AI-generated themes with manual coding and corrects errors.
Common misuse: Accepting AI outputs without checking.
Verification: Independent recomputation, reading transcripts, testing code.
Escalation/stop: If verification fails, correct or stop.
Residual limitation: Verification is time-consuming but essential.
Definition: All research steps must be documented to support reconstruction and audit.
Guiding questions: Can another researcher follow my process? Are prompts, code, and decisions recorded? Are versions tracked?
Example: A researcher saves all prompts, outputs, and code in a shared repository.
Common misuse: Not documenting AI interactions.
Verification: Check that documentation is complete and understandable.
Escalation/stop: If documentation is incomplete, pause and fill gaps.
Residual limitation: Documentation does not guarantee reproducibility if tools are proprietary.
Definition: Researchers must state limitations, disclose AI use, take final responsibility, and submit to appropriate review.
Guiding questions: What are the limitations? Is AI use disclosed? Who is responsible?
Example: A researcher includes a comprehensive AI-use disclosure and states "I take full responsibility."
Common misuse: Omitting AI use or downplaying limitations.
Verification: Review by peers, ethics board, or community.
Escalation/stop: If disclosure is incomplete, correct it; if serious, withdraw.
Residual limitation: Even with disclosure, trust depends on the researcher's integrity.
Show RIGOUR application exercise
Prompt: Using the RIGOUR framework, evaluate the following research scenario: A researcher used AI to generate a literature review, did not verify the sources, and now wants to publish. What RIGOUR elements are violated?
Analysis: All six elements are violated:
- R (Research question): May be based on AI-generated assumptions.
- I (Info governance): Sources unverified, likely fabricated.
- G (Method fit): Method may not match question.
- O (Verification): None performed.
- U (Documentation): AI use not documented.
- R (Disclosure/Responsibility): No disclosure, no accountability.
Corrective action: Withdraw the manuscript, verify all sources, re-analyse if needed, document AI use, and disclose fully.
8. Research-Process Walkthrough: Complete AI-Assisted Research Process with RIGOUR
This walkthrough follows a complete research project from question to disclosure, applying RIGOUR at every stage.
Stage 1: Define research purpose and question (RIGOUR R)
- Human activity: Researcher identifies a problem, articulates purpose, formulates a focused question.
- Bounded AI activity: AI may generate candidate questions for evaluation.
- Documentation: Record the problem, purpose, question, and how AI was used.
Stage 2: Establish governance (RIGOUR I)
- Human activity: Obtain ethics approval, consent, and community permission.
- Bounded AI activity: None.
- Documentation: Ethics approval number, consent forms, data management plan.
Stage 3: Select method (RIGOUR G)
- Human activity: Choose and justify a method that fits the question.
- Bounded AI activity: AI may suggest methods, but human selects and justifies.
- Documentation: Method description, justification, and any AI suggestions considered.
Stage 4: Literature discovery and mapping (RIGOUR I)
- Human activity: Develop search strategy, verify sources, build a map, write synthesis.
- Bounded AI activity: AI may generate search terms, help organise sources, suggest themes.
- Documentation: Search strategy, verification records, map, synthesis, and AI prompts.
Stage 5: Data preparation and analysis (RIGOUR O and G)
- Human activity: Prepare data, conduct analysis (coding, statistics), verify AI outputs.
- Bounded AI activity: AI may assist with transcription, code drafting, provisional coding.
- Documentation: Data cleaning steps, code, outputs, verification logs.
Stage 6: Interpretation and conclusion (RIGOUR O and R)
- Human activity: Interpret results, draw conclusions, identify limitations.
- Bounded AI activity: AI may suggest alternative interpretations, but human decides.
- Documentation: Interpretive notes, alternative explanations, limitations.
Stage 7: Document provenance and assess reproducibility (RIGOUR U)
- Human activity: Create a provenance record (questions, prompts, sources, data, code, outputs, decisions).
- Bounded AI activity: AI may help draft documentation.
- Documentation: Complete provenance file, version control, reproducibility assessment.
Stage 8: Prepare disclosure and take responsibility (RIGOUR R)
- Human activity: Write an AI-use disclosure, state final responsibility.
- Bounded AI activity: AI may help draft disclosure, but human verifies and signs.
- Documentation: Disclosure statement, final sign-off.
Stage 9: Review, correct, escalate, or stop (RIGOUR all)
- Human activity: Conduct a final review. If errors are found, correct or reanalyse. If issues are serious, escalate to supervisor or ethics board. If irreparable, withdraw or stop.
- Documentation: Review notes, correction logs, escalation records.
At every stage, human judgment is paramount. AI is a tool, not a decision-maker. Documentation, reproducibility, and disclosure are not optional—they are integral to scholarly responsibility.
9. Worked Example
This worked example shows a complete AI-supported research project with full RIGOUR application.
Research question: "How do older adults (65+) in urban Canada perceive the usefulness of AI-powered digital literacy apps?"
Step 1: RIGOUR – Research purpose and question
The researcher identifies a gap: little research on older adults' perceptions of AI literacy tools. The question is focused, researchable, and ethically defensible. The researcher documents the rationale.
Step 2: RIGOUR – Information governance
The researcher obtains ethics approval from the university, with consent that covers AI-assisted transcription (using a secure, approved service). Participants are informed that transcripts will be de-identified before any AI processing.
Step 3: RIGOUR – Method selection
Qualitative semi-structured interviews are chosen to capture depth of experience. The researcher justifies this choice in the protocol.
Step 4: RIGOUR – Verification of analysis
The researcher uses AI for transcription and to suggest initial codes. She reads all transcripts, verifies transcripts against audio, and codes manually. She compares AI-suggested codes with her own and adjusts the codebook. All discrepancies are documented.
Step 5: RIGOUR – Documentation, reproducibility, provenance
The researcher keeps a research log with:
- Research question and revisions.
- Ethics approval number.
- Interview protocol and consent forms.
- Search strategy for literature.
- AI system name and version, prompts, and outputs.
- Transcription verification notes.
- Codebook versions.
- Final themes and synthesis.
Step 6: RIGOUR – Uncertainty, disclosure, responsibility
The researcher writes a disclosure: "We used [system name] for transcription and initial theme suggestion. All transcripts were verified, and coding was done manually. The AI did not interpret data or draw conclusions. We take full responsibility for the analysis and findings." Limitations include small sample size and urban context.
Step 7: Review and correction
After a peer review, a reviewer suggests re-examining one theme. The researcher reanalyses the data, confirms the theme, and updates the synthesis. No withdrawal or stop is needed.
Every AI use was documented, verified, and disclosed. The researcher maintained full control and accountability. The project is reproducible and ethically sound.
10. Guided Learning Activities
Instructions: For the following research scenario, identify how you would apply each RIGOUR element.
Scenario: A researcher wants to study the impact of AI-generated feedback on student writing. She plans to use AI to generate feedback drafts and to analyse student revisions.
Show RIGOUR application
- R (Question): Formulate a clear question: "How do students in a first-year writing course revise their drafts after receiving AI-generated feedback, and what revisions do they report?"
- I (Governance): Obtain ethics approval; ensure students consent to having their drafts and feedback analysed; verify all literature sources.
- G (Method): Use a mixed-methods design: qualitative analysis of revisions and survey of student perceptions.
- O (Verification): Verify AI-generated feedback against the original drafts; check that revisions are accurately coded.
- U (Documentation): Record all AI prompts, feedback outputs, coding decisions, and survey data.
- R (Disclosure/Responsibility): Disclose AI use; state that the researcher takes full responsibility for the analysis and conclusions.
Instructions: Create a provenance record for a hypothetical AI-assisted study. Include: research question, AI system, prompts, data sources, analysis steps, and decisions.
Show sample provenance record
- Research question: "How do librarians perceive the usefulness of AI for reference services?"
- AI system: ChatGPT (version 4.0, date: 2025-01-15).
- Prompts: "List interview questions for librarians on AI use."; "Suggest themes from these transcripts."
- Data sources: 12 interview transcripts (de-identified).
- Analysis steps: Transcribed with AI, coded manually, themes verified against transcripts.
- Decisions: Chose qualitative analysis; used thematic analysis; included contradictory examples.
- Versions: Codebook v1 (initial), v2 (after verification).
Instructions: Given the following research description, identify what is missing for reproducibility.
Description: "I used an AI system to find articles, generated themes, and wrote my report. I saved my final report and references."
Show missing elements
Missing for reproducibility:
- AI system name and version.
- Prompts used.
- Dates of use.
- Search strategy (databases, terms).
- Verification records for sources.
- Code or analysis steps.
- Version control.
- Disclosure.
11. Exercises
Complete these eight exercises. Each includes a solution or solution design inside a details element.
Task: For a research question of your choice, outline how you would apply each RIGOUR element. Use the operational definitions.
Show sample outline
Question: "How do social workers in rural areas use AI for case documentation?"
- R: Define the problem (lack of research on rural social workers).
- I: Obtain ethics approval, protect client data.
- G: Use qualitative interviews.
- O: Verify transcripts, code manually.
- U: Document all steps, prompts.
- R: Disclose AI use (transcription only), take responsibility.
Task: Create a provenance record for a hypothetical study on AI and employee productivity. Include at least 5 elements.
Show sample record
- Question: "How do remote employees perceive AI productivity tools?"
- AI: ChatGPT 4, 2025-02-10.
- Prompts: "Generate interview questions."; "Suggest themes."
- Data: 15 interview transcripts.
- Analysis: Thematic analysis; codebook v1, v2.
- Decisions: Included both positive and negative perceptions.
Task: A researcher provides only the final dataset and code. Is this reproducible? What is missing?
Show analysis
Missing: The researcher did not provide the raw data, the analysis environment (package versions), or the documentation of preprocessing steps. Also missing is the research question, protocol, and any AI-assisted steps. Reproducibility requires full transparency.
Task: A study finds a correlation between AI use and productivity, and the analysis is reproducible. Is it necessarily valid? Why or why not?
Show explanation
No. Reproducibility means the analysis can be repeated, but validity means it measures what it claims. The correlation could be due to confounding variables, measurement error, or a flawed instrument. Reproducibility does not guarantee correctness.
Task: Describe a simple version-control plan for a research project that uses AI. What files would you version?
Show plan
Files to version: Research question, protocol, search strategy, source list, dataset (raw and processed), code, prompts, outputs, manuscript drafts, disclosure.
Method: Use a naming convention (e.g., v1, v2) and a log of changes. For code, use Git or a similar tool.
Task: Write a disclosure statement for a study that used AI for transcription, brainstorming codes, and drafting the introduction. The researcher verified all AI outputs.
Show sample disclosure
Disclosure:
"In this study, we used [AI system name, version] for the following tasks: transcription of audio interviews (using a secure, approved service), brainstorming initial code suggestions, and drafting the introduction. All transcripts were verified against the audio by the lead researcher. Code suggestions were reviewed and modified; the final coding was done manually. The introduction was drafted with AI assistance but was substantially revised and verified for accuracy. The AI system did not interpret data, draw conclusions, or make any analytical decisions. All outputs were reviewed and verified by the human researchers. The lead author takes full responsibility for the final content."
Task: Choose three frameworks (e.g., TRACE, CLAIM, PROTECT) and explain how they support RIGOUR in a research project.
Show integration
TRACE supports the "Verification of analysis" (O) by identifying AI hallucinations and errors.
CLAIM supports "Information governance" (I) by verifying sources and claims.
PROTECT supports "Information governance" (I) by ensuring data privacy and security.
Together, they reinforce the RIGOUR elements and ensure comprehensive governance.
Task: For each of the following situations, decide whether to correct, reanalyse, escalate, withdraw, or stop.
- a. A minor typo in a citation.
- b. A fabricated source was included in the literature review.
- c. The analysis code produced an error that changes a key result.
- d. Participants' data were uploaded to an unapproved AI system.
- e. The research question was changed after seeing results, but the change is not documented.
Show decisions
- a. Correction: Fix the typo.
- b. Reanalysis and correction: Remove the source, re-analyse if needed, and update the review.
- c. Reanalysis: Fix the code and re-run the analysis; update conclusions.
- d. Escalate and possibly withdraw: Report to ethics board; consider withdrawing if data cannot be retrieved.
- e. Reanalyse or stop: If the change is substantial, reanalyse with the original question or justify the change; if unjustified, stop.
12. Homework Problems
- a. Research purpose and question
- b. Information, data, source, ethics governance
- c. Method selection and fit
- d. Verification of analysis
- e. Documentation and reproducibility
- f. Uncertainty, disclosure, responsibility
Descriptions: (i) verifying outputs; (ii) defining the question; (iii) documenting all steps; (iv) selecting a method that fits; (v) protecting data and sources; (vi) disclosing AI use.
Show matches
a-ii, b-v, c-iv, d-i, e-iii, f-vi.
Show analysis
Element violated: O – Verification of analysis. The researcher must read transcripts and verify themes. If errors are found, correct and re-analyse.
Show analysis
Elements violated: I (governance) and R (responsibility). The researcher must seek retroactive consent or, if not possible, consider withdrawing the data. They should also report to the ethics board.
Show explanation
Provenance: The record of origins and transformations of all research elements. Four elements: research question, AI prompts, data cleaning steps, code versions.
Show explanation
Provenance allows another researcher to understand exactly what was done, trace decisions, and reconstruct the process. Without provenance, reproducibility is impossible.
Show explanation
Not sufficient. Also needed: environment/package versions, preprocessing steps, raw data, research question, protocol, and any AI-assisted steps documentation.
Show explanation
Reproducible but invalid: A survey that consistently measures something other than what it claims (e.g., measuring social desirability instead of anxiety). Another researcher can repeat the survey and get the same results, but the results are invalid because the instrument is flawed.
Show explanation
Version control tracks changes to files (e.g., research questions, code, prompts) over time. It is important because AI outputs and research decisions evolve; version control allows you to trace changes, revert if needed, and demonstrate a transparent process.
Show sample disclosure
"We used [AI system] to generate an initial list of sources for the literature review and to check grammar and style. All sources were independently verified and manually checked for relevance and accuracy. The AI did not make any decisions about inclusion or interpretation. We take full responsibility for the final content."
Show explanation
CREATE addresses authorship, attribution, and originality. It supports RIGOUR's "Disclosure and responsibility" element by ensuring that AI use is properly credited and that human authorship is not misrepresented. It also helps prevent plagiarism and fake authorship.
Show recommended action
You should issue a correction to the paper, re-analyse the relevant data, and if the interpretation changes, update the conclusions. Inform the journal editors and follow their correction policy.
Show recommended action
Re-run the analysis with corrected code, update the affected results, and clearly state the correction. If the change is minor, you may include it in an erratum; if substantial, reanalyse and possibly revise the paper.
Show sample response
Example (engineering): In my discipline, I would apply RIGOUR to ensure that AI-generated simulation code is verified, that data from sensors is properly governed, and that reproducibility is ensured by sharing code and data. I would disclose AI use in simulations and take responsibility for the design and interpretation.
13. Mini-Project: Complete RIGOUR Research and Scholarly Practice Review
Purpose: To integrate all Unit 13 learning into a comprehensive review of a synthetic research case, applying RIGOUR and all supporting frameworks.
Learning outcomes: Apply RIGOUR, create provenance, assess reproducibility, draft disclosure, and identify stop conditions.
Estimated effort: 3–4 hours.
Context: You will analyse a supplied synthetic research case and produce a structured report.
Supplied synthetic research case
A researcher conducts a study on the effectiveness of AI-powered tutoring for mathematics. They:
- Used AI to generate the research question.
- Generated a literature review with AI, but did not verify sources.
- Collected data from 100 students (scores and survey).
- Used AI to write code for a t-test and regression.
- Accepted the AI's interpretation and conclusion without verification.
- Did not document AI prompts or code.
- Prepared a report with no disclosure of AI use.
Required sections
- Research topic – state it.
- Research problem – identify the gap.
- Research purpose – state it.
- Research question – evaluate and revise if needed.
- Hypothesis – if appropriate.
- Method – describe the planned method.
- Methodological fit – assess fit.
- Significance – justify.
- Feasibility – assess.
- Stakeholders – identify.
- Research authority – what approvals are needed?
- Ethics requirements – identify.
- Community or Indigenous governance – if applicable.
- Proposed AI roles – what AI tasks were proposed?
- Prohibited AI roles – what should be prohibited?
- Search strategy – develop.
- AI-assisted literature discovery – how would you use AI?
- Source verification – plan.
- Literature map – construct (synthetic).
- Scholarly synthesis – write (synthetic).
- Coverage limitations – identify.
- Data inventory – list data types.
- Data authority – who has authority?
- Consent and confidentiality – plan.
- Data minimization – what data are needed?
- Data preparation – plan.
- Qualitative or quantitative analysis – describe.
- Code or calculation verification – plan.
- Interpretation – who interprets?
- Methodological-drift analysis – identify risks.
- Reflexivity – statement.
- TRACE analysis – apply.
- CLAIM verification – apply.
- FAIRER analysis – apply.
- PROTECT analysis – apply.
- CREATE analysis – apply.
- PCAFE-V prompt record – document.
- SCOPE workflow – define.
- PARTNER responsibility allocation – assign.
- LEARN connection – if relevant.
- Provenance record – create.
- Version-control record – plan.
- Reproducibility assessment – assess.
- Reproducibility limitations – identify.
- AI-use disclosure – draft.
- Human scholarly responsibility – statement.
- Correction and reanalysis process – define.
- Escalation – conditions.
- Withdrawal question – when?
- Stop conditions – list.
- Complete RIGOUR analysis – apply all elements.
- Residual uncertainty – what remains?
- Final recommendation – proceed, proceed with verification, redesign, etc.
- Assignment 1 reflection connection – note.
- Preparation for Module 5 – what will you carry forward?
Show staged hints
Hint 1: Start with the research question. Revise the AI-generated question to be clear, focused, and researchable. Then use that as the anchor for all subsequent sections.
Hint 2: Use RIGOUR as a checklist. Go through each RIGOUR element and ask: "Have I addressed this? What evidence do I have?"
Hint 3: Be transparent about limitations. Identify what is not known, what could go wrong, and what you would do differently. This demonstrates scholarly maturity.
- I have applied all RIGOUR elements.
- I have created a provenance record.
- I have assessed reproducibility and its limitations.
- I have drafted a disclosure statement.
- I have integrated TRACE, CLAIM, LEARN, CREATE, PROTECT, FAIRER, PCAFE-V, SCOPE, and PARTNER.
- I have identified correction, reanalysis, escalation, withdrawal, and stop conditions.
- I have preserved human scholarly responsibility.
Show partial example: Education context
Research question: "How do students in a first-year writing course revise their drafts after receiving AI-generated feedback?"
RIGOUR analysis (excerpt):
- R: Question is clear and focused.
- I: Ethics approval obtained; sources verified.
- G: Qualitative method fits.
- O: Themes verified manually.
- U: Provenance documented.
- R: Disclosure: AI used for feedback generation; human final responsibility.
Recommendation: Proceed with verification.
14. Self-Check Quiz
Question 1. What does RIGOUR stand for (operational)?
Show answer
Answer: B. Research purpose, Information governance, Method fit, Verification, Documentation/Reproducibility, Uncertainty/Disclosure/Responsibility (operational).
Explanation: The exact expansion may vary when the author-approved version is supplied, but this is the operational definition.
Question 2. What is the difference between reproducibility and validity?
Show answer
Answer: A. Reproducibility is about consistency; validity is about truth.
Explanation: A study can be reproducible (results can be repeated) but invalid (measuring the wrong thing).
Question 3. Which of the following should be included in an AI-use disclosure?
Show answer
Answer: D. All of the above.
Explanation: A complete disclosure includes all these elements.
Question 4. What is the primary purpose of provenance?
Show answer
Answer: B. To enable reproducibility and auditability.
Explanation: Provenance records allow others to understand and reconstruct the research process.
Question 5. When should a researcher escalate an issue?
Show answer
Answer: D. All of the above.
Explanation: Escalation is appropriate whenever the researcher lacks authority, expertise, or clarity on ethical or governance matters.
Question 6. What is methodological drift?
Show answer
Answer: B. Changing the analysis plan after seeing results without documentation.
Explanation: Methodological drift is a threat to integrity; all changes should be justified and documented.
Question 7. Which framework helps identify hallucinations in AI outputs?
Show answer
Answer: B. TRACE.
Explanation: TRACE is designed to detect hallucinations, errors, and omissions in AI-generated content.
Question 8. What is the final responsibility of a researcher using AI?
Show answer
Answer: B. To take full responsibility for the final work.
Explanation: The researcher, not the AI, is accountable for the integrity and quality of the research.
15. Assessment Connection
- Assignment 1 (AI Learning and Reflection Journal): Preserve evidence of your RIGOUR application, provenance records, reproducibility assessments, and disclosure drafts. Use the mini-project as a key source.
- Assignment 5 (Verification and Fact-Checking): The verification skills you practised are integral to RIGOUR's "O" element.
- Assignment 4 (Applied AI Workflow Project): The documentation and reproducibility principles will inform your workflow design.
- Capstone (Assignment 6): RIGOUR is a comprehensive framework for managing AI use in any research project.
Action: Save your mini-project and homework answers as evidence for Assignment 1.
16. Assignment 1 Journal Connection
- Prompt 1: How has the RIGOUR framework changed your view of AI-assisted research?
- Prompt 2: What does research provenance mean to you, and why is it important?
- Prompt 3: How did you assess reproducibility in your research? What limitations did you identify?
- Prompt 4: What will you include in your AI-use disclosure? Why?
- Prompt 5: How have your responsible-use boundaries changed after completing Unit 13?
Show reflection planning checklist
- I can apply RIGOUR to a research project.
- I can create a provenance record.
- I can assess reproducibility and its limits.
- I can write an AI-use disclosure.
- I can integrate multiple frameworks.
- I can identify correction, reanalysis, escalation, withdrawal, and stop conditions.
17. Key Takeaways
- RIGOUR is the author-approved framework for responsible AI-supported research. It covers research purpose, information governance, method fit, verification, documentation/reproducibility, and disclosure/responsibility.
- Provenance is the record of all research decisions and transformations. It is essential for reproducibility and auditability.
- Version control tracks changes to research artefacts over time.
- Reproducibility is about consistency; it does not guarantee validity or correctness.
- AI-use disclosure must be accurate, complete, and proportionate.
- Correction, reanalysis, escalation, withdrawal, and stop are legitimate scholarly actions to maintain integrity.
- TRACE, CLAIM, LEARN, CREATE, PROTECT, FAIRER, PCAFE-V, SCOPE, and PARTNER all support RIGOUR and should be integrated.
- Human scholarly responsibility is non-negotiable. You are accountable for the final work.
18. Key Terms
- RIGOUR framework
- The author-approved model for responsible AI-supported research, covering Research purpose, Information governance, Method fit, Verification, Documentation/Reproducibility, and Uncertainty/Disclosure/Responsibility.
- Provenance
- A documented record of the origin, transformation, and decisions affecting all research elements.
- Version control
- The practice of tracking changes to files and documents over time.
- Reproducibility
- The ability of another researcher to obtain consistent results by following the documented process.
- Validity
- The extent to which a study measures what it claims to measure.
- Correctness
- The correspondence of conclusions to reality.
- AI-use disclosure
- A statement that transparently reports how AI was used in a research project.
- Correction
- Fixing minor errors without changing conclusions.
- Reanalysis
- Re-running analyses after errors are found and updating conclusions if needed.
- Escalation
- Seeking guidance from supervisors, ethics boards, or specialists when issues exceed expertise.
- Withdrawal
- Removing a published or submitted work due to serious errors or ethical violations.
- Stop condition
- A situation requiring pausing or discontinuing research due to risk or lack of authority.
- Methodological drift
- Unplanned deviations from the approved research method without justification.
- Reflexivity
- Examining how the researcher's assumptions and position influence the research.
19. Looking Ahead
Unit 13 is complete! You have now completed Module 4: AI for Learning, Research, and Knowledge Creation.
Next: You will move to Module 5: AI for Professional Practice and Everyday Life, beginning with Unit 14: Writing, Communication, and Creative Work. Unit 14 will cover how to use AI responsibly for drafting, editing, design, and creative projects.
Before you move on: Complete the Unit 13 consolidation exercise and the cumulative project, and make your Assignment 1 journal entries for Module 4. These will solidify your learning and provide evidence for your final portfolio.
Congratulations on completing Unit 13!
20. Sources and Further Reading
- Primary: COMP 3270 Textbook, Chapter 14: Research and Scholarly Practice (sections on reproducibility, documentation, and disclosure).
- Supporting: Chapter 7 (TRACE), Chapter 8 (CLAIM), Chapter 11 (Copyright, Authorship, Attribution, and Academic Integrity), Appendix D (Responsible-Use Toolkit), Appendix E (AI Workflow Design Toolkit), Appendix H (Sample AI Use Statements).
- Further reading: [INSERT VERIFIED RESEARCH-METHODS REFERENCE].
- Institutional policies: Consult your institution's research ethics, data management, and academic integrity policies.
- Indigenous research: Refer to TCPS 2 Chapter 9 and community-specific governance protocols.