Evidence and sources · Lesson 2 of 3

Searching so that what you find is not an accident

Your findings are downstream of your search terms. Design them, write them down, and use more than one database.

9 min read · Free — no account needed to read

Typing a sentence into a search box returns what the ranking thinks you want. That is fine for orientation and disastrous as a method, because your evidence base ends up selected by an algorithm optimising for something other than truth.

Build search terms deliberately. Take the key concepts from your question and list synonyms for each — the ones your field uses, the ones neighbouring fields use, and the ones used ten years ago. Missing a synonym is how whole literatures stay invisible.

Boolean operators. `AND` narrows, `OR` widens, `NOT` excludes. The usual shape is `(concept1 OR synonym1) AND (concept2 OR synonym2)`. Quotation marks fix a phrase; an asterisk truncates a stem so `adolescen*` catches adolescent and adolescence.

Use several databases. Each indexes a different slice. Crossref is broad, PubMed is biomedical, arXiv is preprints in physics, maths and computing, and a general web search reaches grey literature none of them hold. This product searches several at once and tells you which answered — which is the point, because a search that silently returned fewer sources looks identical to a search that found less.

Citation chaining is where the real finds usually are. Backwards: read the reference list of a good paper. Forwards: find everything that has cited it since, which surfaces work published after it and the critiques. Doing both on three well-chosen papers often beats another hour of keyword searching.

Record what you searched. Terms, databases, dates, counts. Not bureaucracy: without it you cannot tell what you have covered, cannot repeat it in six weeks, and cannot honestly say your search was systematic.

See it happen

Searching "does remote work improve productivity" returns mostly opinion pieces and consultancy reports.

The read that arrives on its own

The academic literature on this must be thin, so you work with what came back.

The same situation, with this idea

You rebuild the search. Concepts: remote work (OR telework OR telecommuting OR work from home OR distributed work) AND productivity (OR performance OR output OR efficiency). "Telecommuting" alone opens two decades of work you had not seen, because that is the term the field used before 2015. The literature was not thin — your terms were.

After this, you can

treat an empty-looking field as a hypothesis about your search terms first.

The takeaway

Your findings are downstream of your search terms. Design them, write them down, and use more than one database.

Easy to take the wrong way

That more results means a better search. A good search is one you can describe, repeat and defend — which often returns fewer, better things.

Try this

Write a Boolean string for your own question with at least two synonyms per concept. Run it, then run your original plain-language search, and compare what each surfaced.