Use Inciteful to build a traceable literature-search shortlist
A citation map is useful when you already know one relevant paper and want to explore connected work. This exercise turns that exploration into a shortlist with reasons you can inspect. Allow about 30 minutes. You need a browser, access to article abstracts and a spreadsheet or reference manager.
Updated 6 November 2026. The two paper decisions below are editor-selected teaching examples. They are not presented as results returned by a live Inciteful search. Your graph, counts and ordering may differ.
Download the search-run log and worked decision log. The pack also includes a separate CSV-analysis exercise.
Set a question narrow enough to guide decisions
Use this exercise question: “How does access to generative AI affect completion time and quality on discrete professional writing tasks?”
For the practice shortlist, seek empirical comparisons involving professional writing tasks and a measured time or quality outcome. Keep broader workplace-productivity studies as background. This is a teaching scope for exploratory discovery; a systematic review requires its own protocol and additional search sources.
Start with Noy and Zhang's 2023 Science article, Experimental evidence on the productivity effects of generative artificial intelligence. Its abstract describes an experiment involving professional writing tasks. The DOI is 10.1126/science.adh2586.
Run a small, documented discovery search
Open Inciteful Paper Discovery and search for the seed title or DOI. Check the title, authors and publication year of the selected record.
Record the date, seed DOI and initial settings in search-run-log.csv. Record any filter you apply, including year or keyword restrictions.
Inspect a manageable set of candidates from the important and similar-paper results. For this exercise, aim to assess five candidates and record where each appeared.
Open each source record. Check the abstract and publication identity, then save the DOI and your reason for retaining it, setting it aside or seeking more information.
Add relevant papers to refine the graph if useful. Save that as a new search run with its changed seeds and settings. Export the available records and keep the raw file unchanged.
Inciteful's official quick start describes seed-based discovery, graph refinement and export. Its importance and similarity rankings use citation relationships. Treat those rankings as discovery aids; assess relevance and evidence quality by reading the sources.
Compare two worked decisions
Candidate for full-text review: Noy and Zhang, 2023. The professional-writing setting matches the practice question. Record the abstract as the initial evidence locator. Before including it in a substantive synthesis, inspect the task design, outcome measures and limitations in the full article.
Background only: Brynjolfsson, Li and Raymond, 2025, Generative AI at Work. Section III describes customer-support work. Under this exercise's focus on discrete professional writing tasks, retain it as contextual reading. A review asking about workplace productivity more broadly could make a different decision.
These decisions show how the same search topic can produce different relevance judgments once the task definition is explicit. Neither decision is a judgment that a paper is good or bad.
Ask AI to organize your notes
After reading the source abstracts, paste your notes into an assistant with this prompt:
“Organize these source notes into a candidate log. Preserve each DOI exactly. Use the columns title, task context, measured outcomes, possible relevance, evidence locator, and unresolved question. Use only the supplied text. Write ‘not reported in supplied notes’ for missing information. Keep my provisional decisions separate from your suggestions. Do not invent papers, quotations or study results.”
Check every row against the original note and source. If the assistant says a paper reports an outcome you cannot locate, mark it unresolved and return to the article. Keep the source evidence locator alongside the summary.
Finish with an auditable output
Your output should contain a dated search-run log, an unchanged export where available, and a reviewed shortlist. Each decision should identify a source and a concrete reason. Record versions separately when a working paper and published article differ.
The example log includes an earlier author-hosted working paper by Noy and Zhang as a source-access option; its title page identifies it as a March 2023 working paper. Use the published DOI for the published version and check which version supports each claim.
Extend the search deliberately
Try a second seed from a different research tradition and compare which candidates appear. Combine citation discovery with a keyword search in a suitable scholarly database. Record what each route contributes; a citation graph's coverage depends on its records, links and your starting points.
See our Inciteful tool entry for the provider link, and check an AI-generated CSV analysis to practise validating numerical outputs.
Comments