Your global study desk

In one group, someone pastes interview data into an AI assistant, someone bans it outright, and someone has no idea what is allowed. How to write a one-page group policy covering data, permitted uses and disclosure.

Delegating what you have mastered saves time; delegating what you are still learning removes the chance to learn it. Which skills to practise yourself, signs of over-reliance and AI-free zones worth keeping.

A polished illustration in three minutes, a mechanism diagram without drawing skills: image generators are tempting. In journals, though, the lines are sharp. A breakdown by image type, common rules and how to disclose.

Having AI label 20,000 student comments in an afternoon is entirely feasible. But once those labels become variables, the model is a measurement instrument and needs validating like any other.

An AI assistant cannot replace an expert reviewer, but it makes a tireless tough reader: it catches mismatches between question and conclusion, missing limitations and abstracts that contradict the text.

Machine translation is now fluent enough to be trusted too easily, yet it can quietly turn “is associated with” into “causes” or drop a vital caveat. A section-by-section workflow, a glossary and the checks that matter.

Screening thousands of titles and abstracts is the most laborious stage of a systematic review. AI can shorten it, but only safely if you measure recall, set a stopping rule and keep humans at the point of decision.

Paste 30 interview transcripts into an AI assistant and get eight tidy themes in five minutes. Tempting, but that is not analysis. Where AI genuinely helps in qualitative coding, how to check it and how to report it.

Generic answers usually come from generic questions. Five components of a good prompt for academic work, six templates for everyday situations, and the kinds of request worth avoiding from the start.