Chapter 6. AI for Learning, Research, Writing, and Knowledge Work

6.1 Learning and teaching

AI can explain a difficult concept in simpler language, generate practice questions, compare theories, provide feedback on a draft, create revision schedules, and simulate a discussion. Students should use it to deepen understanding rather than bypass learning. A strong method is to attempt a problem first, ask for feedback, revise independently, and explain the reasoning in their own words.

Teachers can use AI to develop examples, differentiated learning materials, formative quizzes, and administrative drafts. They should verify subject accuracy, protect student information, check accessibility, and ensure that assessment rules are clear. Institutions should define acceptable assistance, disclosure expectations, and the process for reviewing suspected misuse.

6.2 Research and literature reviews

AI may help refine research questions, generate search terms, organise notes, compare supplied papers, identify possible themes, and explain methods. It should not replace systematic searching, critical appraisal, or reading the original evidence. References generated by a model must be checked against publisher, journal, library, or DOI records.

Researchers should preserve a clear audit trail: databases searched, inclusion criteria, documents reviewed, extraction decisions, analytic steps, and any AI assistance. Do not invent quotations or citations. When the task is based on a supplied document, distinguish what the document says from external background knowledge.

6.3 Writing and editing

AI can assist with structure, plain-language editing, translation, tone adjustment, summaries, and consistency checks. Authors remain responsible for originality, accuracy, attribution, argument, and compliance with institutional or publisher rules. A polished paragraph may still contain unsupported claims.

Good practice is to keep source notes, verify quotations, review every reference, and disclose AI assistance when required. Do not submit material you cannot explain or defend. For sensitive or confidential documents, use only an approved environment.

6.4 Data analysis

AI can help explain statistical concepts, draft code, identify possible data-quality problems, and suggest visualisations. It may also misunderstand variable definitions, select inappropriate tests, or generate code that silently produces incorrect results. The analyst should validate code, inspect missingness and outliers, check assumptions, compare results with an independent method, and document decisions.

Never upload identifiable research participants' data to an unapproved service. De-identification may be insufficient if combinations of attributes can reveal identities. Follow ethics approval, consent conditions, data-management plans, and local rules.

6.5 Meetings and knowledge management

AI can summarise meeting notes, extract action items, draft minutes, and search approved internal documents. Before recording or transcribing, ensure that participants are informed and that consent and organisational policies are respected. Check names, decisions, deadlines, and assigned owners because automated summaries can omit qualifications or attribute statements incorrectly.

Organisations should establish which version of a document is authoritative, who can access it, how updates are reflected, and how AI-generated summaries are corrected. Searchable knowledge is only as dependable as the source content and its governance.