wen-chen research-gap-finder
1. Overview
research-gap-finder identifies research gaps by systematically analyzing a literature corpus from three angles: topical coverage gaps (what has not been studied), methodological gaps (what has been studied inadequately), and controversy gaps (what remains unresolved or contradictory). For each gap identified, the module generates a specific gap statement, an actionable research question, and — when requested — a testable hypothesis. This module is Step 2 of the Grant pipeline: it receives context_output from literature-reviewer and produces focused research directions for grant-writer. When running in the pipeline, parameters already present in upstream context are not re-collected from the user.
2. Parameters
Required
| Parameter | Values | Description |
|---|---|---|
domain |
tcm | bioinformatics | clinical | pharmacology |
Research domain. Controls the variables used in the Literature Matrix Gap Scan. |
Optional
| Parameter | Values | Description |
|---|---|---|
scope |
narrow | broad (default: narrow) |
narrow = focused gap within one sub-area; broad = map all major gaps across the field |
output_style |
questions | hypotheses | both (default: both) |
Whether to produce questions only, hypotheses only, or both |
n_gaps |
integer 3–10 (default: 5) |
Number of gaps to identify and present |
Parameter collection rule
If domain is already in an upstream context_output, inherit it silently — do NOT ask again. If missing and cannot be inferred, ask once. Do not ask for optional parameters unless the user raises them.
3. Workflow
Step 1 — Check upstream context.
If context_output from literature-reviewer is present, extract: domain (inherit, skip collection), key_findings (primary evidence base), references (verify gap claims), summary (narrative field context). If no upstream context, ask the user for a literature summary or key findings before proceeding. Do not generate placeholder gaps.
Step 2 — Collect missing parameters.
If domain was not inherited, ask for it first. Default scope to narrow without asking on standalone runs. Never ask for multiple parameters in one message.
Step 3 — Load gap-analysis framework.
Read references/gap-analysis-frameworks.md. Apply all three methods in Steps 4–6.
Step 4 — Apply Method 1: Literature Matrix Gap Scan.
Construct a mental 2D matrix using the two or three most important domain variables (e.g., for tcm: herb × disease; for bioinformatics: gene × cancer type × regulatory mechanism; for clinical: intervention × population subgroup). Identify unstudied cells (no evidence) and sparse cells (one or two small studies). These are topical coverage gaps.
Step 5 — Apply Method 2: Controversy Mapping.
From key_findings, identify contradictory finding pairs. For each: state both findings; diagnose the likely cause (cell line, population, assay, dose); formulate a resolution question specifying the factor that would reconcile the contradiction.
Step 6 — Apply Method 3: Methodological Gap Scan. For each topic, flag evidence that is: small-N (<30/group for RCTs, <5 studies for systematic reviews); in vitro only; animal only; retrospective only; or correlational only. For each limitation, specify the methodological upgrade (study design + inferential gap it closes).
Step 7 — Synthesize and rank.
Rank candidate gaps on feasibility, novelty, and impact (each 1–5). Select top n_gaps. For each: write a gap statement (1–2 sentences), a research question (specific, named variables), and — if output_style is hypotheses or both — a testable hypothesis.
Step 8 — Format and pipeline handoff. Present output per Section 4. In the Grant pipeline, append: "context_output is ready for grant-writer. Confirm to proceed." Standalone: offer to continue to grant-writer but wait for user confirmation.
4. Output Format
## Research Gaps Identified: [Topic]
### Gap 1: [Gap name]
**Statement:** [1–2 sentences describing what is missing and why it matters]
**Research Question:** [Specific, answerable question naming the key variables]
**Hypothesis:** [Testable, falsifiable hypothesis — if output_style includes hypotheses]
**Feasibility:** [1–5] | **Novelty:** [1–5] | **Impact:** [1–5]
### Gap 2: [Gap name]
...
If output_style is questions, omit the Hypothesis line. If output_style is hypotheses, include both lines — the question provides context for the hypothesis.
5. context_output
Reads from upstream (literature-reviewer)
| Field | Use in this module |
|---|---|
domain |
Inherited as required parameter — not re-collected |
language |
Passed through unchanged to downstream modules |
key_findings |
Primary input for Methods 1–3 in Steps 4–6 |
references |
Verify gap claims; identify underrepresented study designs |
summary |
Narrative context for field coverage |
Writes to output
{
"module": "research-gap-finder",
"domain": "<inherited>",
"language": "<inherited>",
"summary": "<paragraph describing the 3 most important gaps, 100–200 words>",
"key_findings": ["<Gap 1 research question>", "<Gap 2 research question>", "..."],
"parameters": { "scope": "<collected>", "n_gaps": 5, "output_style": "both" },
"status": "success",
"error_message": null
}
Note on key_findings repurposing: This module's key_findings carries research questions, not literature findings. grant-writer reads them as the basis for the proposal's Innovation and Approach sections — each question maps to a specific aim. Modules needing original literature findings (paper-writer, citation-checker) should read from literature-reviewer's output directly.
Status: "success" when all three methods applied and ≥ n_gaps gaps produced. "partial" when upstream findings are sparse (<3) or upstream status was "partial". error_message is null unless no usable output can be generated.
6. References
See references/ for:
gap-analysis-frameworks.md— Three systematic methods (Literature Matrix Gap Scan, Controversy Mapping, Methodological Gap Scan); gap quality criteria; common pitfalls