AI-R02 — AI for Literature Review and Evidence Synthesis

Course Code: AI-R02  |  School: School of Artificial Intelligence  |  Cluster: Level 2 — Research & Academic Practice

Level: Intermediate  |  Duration: 6 weeks · 20–26 learning hours  |  Language: English  |  Certificate: Professional Certificate (non-degree)  |  Format: Self-paced with AI support under human supervision

Overview

Literature review is the task generative tools appear to solve and in fact make more dangerous. A model will produce a fluent survey containing citations that do not exist, attributions to the wrong author, and confident summaries of papers it has not read. The professional consequence of publishing such a review is severe and irreversible.

This course therefore teaches review as a documented, auditable process. Learners build a search strategy that another person could re-run, apply explicit inclusion criteria, extract evidence into a structured matrix, appraise quality, and synthesise. AI is used at specific, bounded points: query expansion, screening support with human adjudication, and first-pass extraction that is then checked against the source. Every citation in the final review must be verified against the publisher record, and the verification is part of the submission.

The course follows the reporting conventions of systematic review even for narrative and scoping reviews, on the grounds that transparency about method is valuable regardless of the review type.

Learning outcomes

On completion, a successful learner will be able to:

1. Formulate a review question with explicit inclusion and exclusion criteria.
2. Build and document a reproducible search strategy across appropriate databases, including the exact strings and dates.
3. Screen records at title, abstract and full-text stages with recorded reasons for exclusion.
4. Use AI assistance for query expansion, screening support and first-pass extraction while retaining adjudication.
5. Verify every citation against the publisher record and document the verification.
6. Appraise study quality with an appropriate instrument and integrate the appraisal into the synthesis.
7. Produce a synthesis that reports the state of evidence, including its gaps and inconsistencies, and a flow diagram accounting for every record.

Who this course is for

Doctoral researchers, academics preparing review articles, policy analysts producing evidence reviews, and librarians supporting systematic search.

Prerequisites

Postgraduate familiarity with academic literature and database searching. AI-F06 recommended, since the verification discipline taught there is assumed here. Learners need access to at least two bibliographic databases through their institution.

Syllabus

Module 1 — Review types and the question

Focus. Systematic, scoping, rapid, narrative and integrative reviews, and matching type to purpose. Framing the question and writing inclusion criteria that a second reviewer could apply identically.

Lessons. 1.1 Review types and their claims. 1.2 Question frameworks. 1.3 Inclusion and exclusion criteria. 1.4 Registering the protocol.

Core reading. Sandy Oliver, David Gough & James Thomas, An Introduction to Systematic Reviews, 2nd edition (London: SAGE, 2017), chapters 1–3.

Deliverable. Review protocol: question, type, inclusion and exclusion criteria.

Module 2 — Search strategy as a reproducible artefact

Focus. Database selection, controlled vocabulary and free text, Boolean structure, and documentation sufficient for another researcher to reproduce the search exactly. Grey literature and language bias.

Lessons. 2.1 Database selection and coverage. 2.2 Controlled vocabulary and free-text combination. 2.3 Documenting strings, dates and limits. 2.4 Grey literature and language bias.

Core reading. Julian P. T. Higgins et al. (eds), Cochrane Handbook for Systematic Reviews of Interventions, version 6 (Chichester: Wiley & Cochrane), chapter 4 on searching.

Deliverable. Full search documentation with exact strings, databases, dates and result counts.

Module 3 — Screening, and where AI may assist

Focus. Two-stage screening with recorded exclusion reasons, inter-rater agreement, and the bounded use of machine assistance for prioritisation and duplicate detection under human adjudication.

Lessons. 3.1 Title and abstract screening. 3.2 Full-text screening and exclusion reasons. 3.3 Agreement and adjudication. 3.4 Machine assistance: prioritisation, not decision.

Core reading. Gerit Wagner, Roman Lukyanenko & Guy Paré, “Artificial Intelligence and the Conduct of Literature Reviews”, Journal of Information Technology 37, no. 2 (2022): 209–226.

Deliverable. Screening log with counts, exclusion reasons and adjudication record.

Module 4 — Extraction and citation verification

Focus. Structured extraction into a matrix, and the non-negotiable verification step. Fabricated and misattributed citations, how they arise, and the procedure that catches them before submission.

Lessons. 4.1 Designing the extraction matrix. 4.2 First-pass machine extraction and its error profile. 4.3 Verifying a citation against the publisher record. 4.4 Handling retracted and predatory sources.

Core reading. Ziwei Ji et al., “Survey of Hallucination in Natural Language Generation”, ACM Computing Surveys 55, no. 12 (2023).

Deliverable. Extraction matrix plus a citation verification table covering every included source.

Module 5 — Quality appraisal

Focus. Selecting an appraisal instrument appropriate to the study designs included, applying it consistently, and letting the appraisal shape the weight given to each study in the synthesis.

Lessons. 5.1 Choosing an appraisal instrument. 5.2 Applying it consistently. 5.3 Risk of bias and its reporting. 5.4 From appraisal to weighting.

Core reading. Higgins et al., Cochrane Handbook, chapters 7–8 on bias assessment. Oliver, Gough & Thomas, An Introduction to Systematic Reviews, chapter 6.

Deliverable. Completed appraisal table with justification of the instrument chosen.

Module 6 — Synthesis and transparent reporting

Focus. Narrative and thematic synthesis, reporting inconsistency rather than smoothing it, identifying genuine gaps, and producing a flow diagram that accounts for every record from identification to inclusion.

Lessons. 6.1 Synthesis approaches and their limits. 6.2 Reporting inconsistency honestly. 6.3 Identifying gaps that are gaps. 6.4 Flow diagram and reporting checklist.

Core reading. Matthew J. Page et al., “The PRISMA 2020 Statement: An Updated Guideline for Reporting Systematic Reviews”, BMJ 372 (2021): n71.

Deliverable. Final submission: complete review with synthesis, flow diagram, reporting checklist and AI-use disclosure.

Assessment

Pass mark 70 per cent. All assessed components must be attempted. Every mark in this course is issued by a human assessor; no assessment outcome is generated automatically.

Rubric criteria

Each assessed artefact is marked against four criteria at four levels (distinction, pass with merit, pass, fail).

1. Reproducibility: could a competent stranger re-run this search and arrive at the same record set?
2. Verification completeness: is every included citation verified against a publisher record, with the check evidenced?
3. Appraisal consistency: is the quality instrument applied uniformly, and does the appraisal actually inform the synthesis?
4. Interpretive restraint: does the synthesis report what the evidence supports, including inconsistency and absence?

Reading list

Core. Sandy Oliver, David Gough & James Thomas, An Introduction to Systematic Reviews, 2nd edition (London: SAGE, 2017). Julian P. T. Higgins et al. (eds), Cochrane Handbook for Systematic Reviews of Interventions, version 6 (Chichester: Wiley & Cochrane).

Reporting standard. Matthew J. Page et al., “The PRISMA 2020 Statement”, BMJ 372 (2021): n71.

Peer-reviewed. Gerit Wagner, Roman Lukyanenko & Guy Paré, “Artificial Intelligence and the Conduct of Literature Reviews”, Journal of Information Technology 37, no. 2 (2022). Ziwei Ji et al., “Survey of Hallucination in Natural Language Generation”, ACM Computing Surveys 55, no. 12 (2023).

All items are published works identifiable by author, title and publisher. Learners obtain them through an institutional library or the publisher. The Academy does not distribute copyrighted texts.

Academic integrity and use of AI

Generative tools may be used in producing assessed work under three conditions. Use must be disclosed in a short statement appended to each submission, naming the tool and the task it performed. Any factual or technical claim originating from a generative tool must be verified against a citable source before it enters assessed work, and the verification must be evidenced. The analytical judgement in each artefact must be the learner’s own and must be defensible in a short follow-up. A single unverifiable or fabricated citation in the final submission is a fail, irrespective of the quality of the remaining work. This mirrors the consequence in publication.

Instructor: pending owner confirmation. Pricing: pending owner approval. Reference list verified against publisher records; any later addition is marked for verification before publication.