Module 5: Research and Scholarship¶
Module 5 of 7
What You Will Be Able to Do¶
- Use AI for literature work with a workflow that catches its failure modes.
- Apply current disclosure norms for AI assistance in scholarly writing.
- Know how to handle material that belongs to someone else or identifies research participants.
The Core Idea¶
The Four-Minute Version
Research offers real gains from artificial intelligence (AI) assistance (literature screening, summarization, drafting, code for analysis), and AI-assisted work is held to the same integrity standards as any other method: what you report has to be accurate and honestly described.
Literature Work¶
Models are genuinely useful for screening titles and abstracts against inclusion criteria and for first-pass synthesis; recent studies (linked as further reading under the research prompts) report strong sensitivity when prompts state criteria explicitly and a human verifies. Two habits make it work: check every reference an AI suggests at the original source before you cite it, because fabricated citations remain a signature failure; and use a screening assistant for recall rather than judgment, keeping the borderline calls for yourself. The Literature Screening Assistant prompt in the library is built around exactly this division of labor.
Writing and Disclosure¶
Two norms are now widely shared across journals: an AI tool cannot be an author, because authorship requires accountability no tool can hold; and AI assistance must be disclosed. What must be disclosed, and where, still varies by venue. The International Committee of Medical Journal Editors (ICMJE) recommendations now carry a dedicated section on AI use by authors, reviewers, and editors; check it and your target journal's instructions before submitting. Presenting AI-generated data, images, or references as real empirical material is fabrication, which research integrity standards treat as misconduct.
Confidential Material and Participant Data¶
Manuscripts and grant applications you receive for peer review are shared with you in confidence: their authors have not agreed to anyone else seeing them, and many journals and funders bar reviewers from putting them into AI tools. That reason holds for a paid plan as much as a free one, and the reviewer instructions say whether any AI use is permitted. Funder rules for AI in proposal preparation vary and change; check the current policy of your funder before drafting with assistance. Data that can identify research participants is governed by your institutional review board (IRB) protocol, which sets where participant data may go, and the AI Responsible Use Policy is the reference for it too. For analysis planning, a description of the variables, the design and the data structure, or a few invented example rows, gives a model a useful starting point without sharing participants' data.
The three zones of research AI use
Assisted, verified
- literature screening
- first-pass synthesis
- analysis code drafting
Every citation checked at the original source.
Disclosed
- AI cannot be an author
- assistance disclosed per venue
- check ICMJE and the journal
Fabricated data, images, or references are misconduct.
Protected
- manuscripts under review
- grant confidentiality, funder rules
- participant data and unpublished findings
Your IRB protocol sets where participant data may go.
Accountability never moves: it stays with you and your coauthors in every zone.
Self-Check¶
A model produced a beautifully formatted reference list for your introduction. Five of the twelve citations check out so far. What do you do with the other seven?
Verify every one at the original source before any of them enter the manuscript, and expect some to be fabrications: plausible authors, real journals, nonexistent papers. A partially verified AI reference list is not a reference list; it is a list of leads.
You are reviewing a manuscript for a journal and want an AI summary to speed things up. What is the problem?
Manuscripts under review are shared with you in confidence; uploading one to an AI service passes the authors' unpublished work to a third party without their consent, and many journals and funders bar it in their reviewer instructions. Check those instructions for any tool the journal provides; otherwise the summary is yours to write.
Where does your accountability sit when AI helped with the analysis code and the drafting?
Exactly where it sat before AI: with you and your coauthors. Disclose the assistance per the venue's rules, verify what the tool produced, and stand behind every number and sentence. No disclosure transfers responsibility to the tool.
Going Deeper¶
Color key: PromptGuide
- AI for Research The tool landscape mapped to each stage of a project, starting with Scopus AI, which the American University of Antigua licenses.
- Reviewing the Literature The full workflow from question to appraised evidence.
- Research Prompts The question coach, pre-submission reviewer, and reporting guideline auditor.
Content last reviewed September 1, 2026. Review dates are maintained as described on the About page.