Add research workflow guide

README.md

@@ -1,3 +1,74 @@
-# interdisciplinary-search-vocabulary
+# Mapping Vocabulary Across Disciplines for Interdisciplinary Literature Searches
 
-Mapping vocabulary across disciplines for interdisciplinary literature searches
+Interdisciplinary questions are some of the most interesting in research, and some of the hardest to search. The same idea can carry completely different names in psychology, computer science, economics and medicine. A concept that education researchers call "retrieval practice", cognitive psychologists call the "testing effect", and software designers building learning apps call "active recall". If your search only uses the words of your home discipline, it will quietly miss most of the relevant work from neighbouring fields. This guide describes a systematic way to map vocabulary across disciplines before you search, and shows where an AI research agent such as [Yila](https://yila.ai) can make the mapping faster and more complete.
+
+## Why vocabulary drift happens
+
+Disciplines develop their own terminology for good reasons: precision, historical accidents, different theoretical framings and different methods. Over time, the same phenomenon gets studied in parallel under several names. Common patterns include:
+
+- **Synonyms with different connotations.** "Misinformation", "fake news", "information disorder" and "rumour" overlap heavily but carry different assumptions about intent and scale.
+- **Method-driven names.** Economists may speak of "treatment effects" where clinicians speak of "efficacy" and computer scientists of "uplift".
+- **Level-of-analysis differences.** Sociologists may study "social capital" at the community level while psychologists study "perceived social support" in individuals.
+- **Field-specific acronyms.** The same acronym can mean different things in different fields, and the same idea may be abbreviated differently.
+
+Bibliographic databases also differ. MEDLINE uses MeSH headings, PsycINFO uses its own thesaurus, ERIC has descriptors for education, and engineering databases often have no controlled vocabulary at all. A good interdisciplinary search must therefore combine several vocabularies.
+
+## Step 1: Define the concept independently of any discipline
+
+Start by writing a plain-language definition of each core concept in your question, without using any discipline's preferred term. For example: "giving learners questions to answer from memory, rather than asking them to re-read material, in order to improve long-term retention". This neutral definition becomes the anchor against which you judge whether a term from another field really means the same thing.
+
+## Step 2: Collect seed papers from each relevant field
+
+Identify at least two or three clearly relevant papers from each discipline you expect to contribute. Sources include known review articles, reference lists of key papers, and quick exploratory searches in discipline-specific databases. Read the titles, abstracts, keywords and introductions closely and note every term used for your concept.
+
+## Step 3: Harvest terms systematically
+
+From each seed paper and each database, record:
+
+- **Free-text terms** used in titles and abstracts.
+- **Author keywords.**
+- **Controlled vocabulary** assigned by each database, for example MeSH terms, PsycINFO thesaurus terms or ERIC descriptors.
+- **Broader and narrower terms** in each thesaurus.
+
+Keep them in a table with one row per term and columns for the discipline, the source and an example paper. This table is your vocabulary map.
+
+An AI research agent can accelerate this step considerably. Given a neutral definition and a few seed papers, Yila can extract candidate terms from abstracts and keywords across fields, suggest the corresponding controlled-vocabulary headings in different databases and group near-synonyms together, so you start from a broad, organised list instead of a blank spreadsheet.
+
+## Step 4: Check that each term really matches the concept
+
+Not every related term belongs in the search. For each candidate, ask:
+
+- Does it mean the same thing as the neutral definition, or something broader or narrower?
+- Would including it bring in a large volume of irrelevant material?
+- Is it used consistently within its field, or does it have several meanings?
+
+Classify each term as "core" (clearly the same concept), "broader" (includes the concept but also much else) or "related but distinct". Core terms go into the search; broader terms may be combined with other concepts to keep precision acceptable; distinct terms are left out but recorded, so you can explain the choice.
+
+## Step 5: Build discipline-aware search blocks
+
+For each concept, build a search block that combines core terms from all disciplines with OR, plus each database's controlled vocabulary. Then adapt the block to each database:
+
+- In MEDLINE, use MeSH headings plus free text.
+- In PsycINFO, use thesaurus terms plus free text.
+- In databases without controlled vocabulary, rely on free text and use proximity operators where available.
+
+Test each block by checking whether it retrieves all of your seed papers from every discipline. A seed paper that is missed signals a missing term.
+
+## Step 6: Test recall with a validation set
+
+Collect a small validation set of 15 to 30 papers you know should be included, drawn from all relevant fields and not used to build the search. Run the full search and check how many of them it retrieves. If papers from one field are consistently missed, return to the vocabulary map for that field. Report the validation process in your methods section; it is strong evidence that the search is comprehensive.
+
+## Step 7: Document the map
+
+Include the vocabulary map, or a summary of it, in your supplementary material. It shows reviewers and readers exactly how you handled interdisciplinary terminology, and it is valuable for anyone updating the review later.
+
+## Common pitfalls
+
+- **Searching only in your home database.** Even a perfect term list fails if the relevant literature is indexed elsewhere.
+- **Over-broad terms.** Adding a broad term from another field can multiply the result set without adding relevant papers. Combine broad terms with other concepts.
+- **Ignoring older terminology.** Fields rename concepts over time. Older studies may use terms that are no longer fashionable.
+- **Not testing.** Without a validation set it is impossible to know whether the map works.
+
+## Summary
+
+Interdisciplinary searches succeed or fail on vocabulary. Define each concept neutrally, collect seed papers from every relevant field, harvest and classify terms, build discipline-aware search blocks and validate them against known relevant papers. An AI research agent like Yila can do much of the harvesting, grouping and testing quickly, which leaves researchers free to make the conceptual judgements about which terms truly belong — the part of the work that determines whether the review really captures what every field has learned.