GSOE9011 Chap.1 Searching and Evaluating Research
Searching and Evaluating Research
Literature search is the first research design decision in GSOE9011. A defensible search translates a problem into concept blocks, combines synonyms deliberately, records databases and filters, screens with explicit criteria, and follows citations to test whether the evidence set is complete enough for the question.
Source evaluation then separates relevance from credibility and identifies the conditions that restrict transfer. The goal is not the largest pile of papers. It is an auditable evidence boundary that can support a literature review and reveal a meaningful research gap.
What this chapter covers
- 01
Problem framing before database searching
- 02
Concept blocks and discipline synonyms
- 03
Boolean logic and field selection
- 04
Inclusion and exclusion criteria
- 05
Backward and forward citation searching
- 06
Relevance, credibility and transfer
- 07
Search logs and reproducibility
- 08
Saturation and explicit boundaries
Worked example · free
Repairing a broad search for an engineering comparison
- +1Separate the material, failure mode and engineering system into concept blocks, adding accepted synonyms inside each block.
- +1Join synonyms with OR and blocks with AND, then select title, abstract and keyword fields before adding justified filters.
- +1Screen for measured early-age cracking outcomes and an interpretable material comparison; exclude records that address only environmental performance.
- +1Record the query and boundary, then use reference and citation chains to find foundational and later testing studies.
Key terms
- Concept block
- A group of synonyms or related search terms representing one part of a research question.
- Search string
- The complete combination of terms, operators and fields entered into a database.
- Inclusion criterion
- A pre-stated condition a record must satisfy to enter the evidence set.
- Citation chaining
- Following references backward and citations forward to locate related research.
- Search saturation
- The point where repeated searches and citation routes add little new relevant information.
- Evidence boundary
- The declared scope within which the collected sources can support conclusions.
Searching and Evaluating Research FAQ
How do I make a literature search reproducible?
Record the database, date, full query, fields, filters and screening criteria. Keep enough detail for another researcher to repeat the route and understand why records entered or left the evidence set. A screenshot alone rarely captures the decisions that shaped the result.
Should I search my full research question as one sentence?
Usually no. Break the question into concept blocks, list discipline synonyms and spelling variants inside each block, join alternatives with OR, and join the blocks with AND. This makes missing terms and overly restrictive combinations visible.
How do I decide whether a paper is credible?
Inspect the question, method, comparator, measurement, analysis, uncertainty and reporting rather than relying only on journal prestige or citation count. A credible paper may still be unsuitable when its system or conditions do not transfer to your claim.
When can I stop searching?
Stop when repeated database variations and citation chains add little new conceptual or methodological information for the defined question. Report the boundary and any known gaps instead of claiming that every possible source was found.
Assessment move
Build a living search log rather than repeating unrecorded experiments in the database. Begin with a one-sentence problem and convert it into concept blocks. Test each block alone, inspect the language used by useful records, and revise synonyms deliberately. Screen a small sample before committing to criteria, then apply those criteria consistently.
For every retained source, label its likely role in the proposal: context, mechanism, method precedent, conflicting evidence or limitation. Use citation chaining after the main query so influential older work and later corrections are not missed. In forum contributions, explain why a source matters and where it stops being transferable.
Before writing the review, reopen the saved query and verify that the search record reproduces the expected core results. Practise query diagnosis with three deliberately different failures. For an overbroad result set, identify which concept block is absent from many records and add a field-specific term rather than piling on unrelated exclusions.
For an empty result set, remove one block at a time and check whether a phrase, spelling variant or database field caused the collapse. For a biased result set, inspect whether the chosen outcome language already assumes a benefit. Maintain a small benchmark set of clearly relevant papers and test whether revisions still retrieve them.
When a useful record appears, capture its controlled vocabulary, references and later citations, but keep those discovery routes distinct in the log so another researcher can reproduce them. Evaluate transferability with a fixed comparison: system or population, intervention or exposure, comparator, outcome definition, operating conditions and study design.
Record not only why a paper is credible but what proposition it can support in this proposal. A rigorous experiment in a different material or loading regime may be excellent evidence for a method and poor evidence for the expected effect. End each search session with a decision note: which block changed, what evidence motivated the change, how many records were screened, and what uncertainty remains.
Before finalising the review, rerun the principal query, document the date and database, inspect newly published records, and explain any limits created by language, indexing or access. This converts searching from a hidden prewriting activity into an auditable part of the research design.
