Research Synthesis Protocol¶
Synthesis protocol — instantiates Compositional Assembly
Selects and reconciles findings, models, and cases into one internally consistent knowledge product that answers a defined question and stands up to scrutiny.
A Research Synthesis Protocol is the disciplined procedure for turning a scattered body of evidence — trials, field studies, mechanistic models, cases, and their uncertainties — into one coherent knowledge product that answers a stated question. Here the components being assembled are claims, and the defining, hard part is reconciliation: resolving contradictions, weighing conflicting evidence, and keeping the resulting whole internally consistent, rather than merely summarizing sources side by side. It begins from the question the synthesis must answer, enforces that its conclusions do not contradict each other or the retained evidence, and tests whether the synthesized conclusion actually holds up under scrutiny. It does not catalog parts or wire modules; it composes a defensible answer out of evidence that disagrees.
Example¶
A panel must answer a sharp question: for adults with chronic lower-back pain, does a particular class of physiotherapy meaningfully reduce disability over six months? The evidence is a mess of disagreement — several randomized trials show a modest benefit, a large observational cohort shows none, and a mechanistic model predicts benefit only for a specific subgroup. A weak synthesis would average these into an uninformative shrug.
The protocol instead starts by fixing the question precisely — which population, which outcome, over what horizon — because that is the whole the assembly must serve. It then selects and grades the evidence for inclusion, and reconciles the disagreement rather than pooling it flat: the observational cohort was confounded by baseline severity, and the trial benefit concentrates in exactly the subgroup the model flags. The reconciled conclusion — "a modest benefit concentrated in subgroup X, low certainty elsewhere" — is checked for internal consistency: it must not assert both "works" and "doesn't work." Finally the panel runs a sensitivity analysis: does the conclusion flip if the two largest trials are excluded? It does not, so the whole holds. The output is a defensible, reasoned recommendation, not a stack of summarized abstracts.
How it works¶
What separates a synthesis from a literature dump is that its labor is reconciliation-to-consistency and validation of a claim-whole:
- Fix the question first. The population, outcome, and decision the synthesis must serve are stated up front, so evidence is judged by its contribution to that answer rather than collected for its own sake.
- Select and grade evidence. Sources are included and weighted by relevance and quality, so the whole is built from what actually bears on the question.
- Reconcile the disagreements. Where sources conflict, the protocol explains why — confounding, population differences, method — and builds one consistent account instead of averaging contradictions into mush.
- Validate the whole. Sensitivity and robustness checks ask whether the conclusion survives dropping the weakest, or the most influential, inputs.
Tuning parameters¶
- Inclusion strictness — how high the bar for evidence to enter. A strict bar raises quality but narrows the base and risks missing signal; a loose bar is comprehensive but noisier.
- Reconciliation stance — mechanical pooling versus explanatory reconciliation of disagreement. Pooling is reproducible but can bury real heterogeneity; explanation is richer but leans hard on judgment.
- Uncertainty representation — a single point conclusion versus a graded-certainty statement. Graded certainty is more honest but harder for a decision-maker to act on directly.
- Validation depth — from one sensitivity check to a full robustness and subgroup sweep. Deeper validation buys confidence at real analytic cost.
When it helps, and when it misleads¶
Its strength is that it produces a defensible, internally consistent answer from evidence that conflicts, with the disagreements explained rather than hidden, and with the conclusion's robustness made explicit. That is exactly what a raw evidence pile cannot give a decision.
Its failure modes are sharp. Coherence is not correctness: a biased evidence base yields a synthesis that is perfectly consistent and confidently wrong — garbage in, coherent garbage out. And false reconciliation forces consistency by suppressing genuine heterogeneity, papering over a disagreement that was the real finding. The classic misuse is running the protocol backwards — assembling and citing evidence to support a conclusion already chosen, a seek-and-cite review dressed as a synthesis. The guarding discipline is to pre-specify the question and inclusion criteria before seeing results, to grade the certainty of the whole rather than asserting it,[n1] and to carry uncertainty through to the recommendation rather than collapsing it into false confidence.
How it implements the components¶
A Research Synthesis Protocol fills the archetype's whole-function-and-coherence slots, over claims rather than parts:
intended_whole_function— the exact question and decision the synthesis must answer, which anchors what evidence counts and prevents the assembly from becoming a summary.coherence_invariant— the reconciled conclusions must stay internally consistent with each other and the retained evidence; contradictions are resolved, not averaged away.system_level_validation— sensitivity and robustness analysis test whether the synthesized conclusion holds as a whole, not merely whether each source is individually sound.
This protocol composes claims, not artifacts: it does not draw the interaction_map or fix the interface_contract between components — that is Architecture Blueprint, its nearest structural counterpart, which renders how parts connect where this reconciles what evidence concludes. Nor does it maintain a component_catalog or run a compatibility_check on versioned parts (Bill of Materials and Configuration), or assign human role_definition and assembly_owner (Team Composition Matrix).
Related¶
- Instantiates: Compositional Assembly — it composes disagreeing evidence into one coherent, validated knowledge whole.
- Sibling mechanisms: Architecture Blueprint · Bill of Materials and Configuration · Team Composition Matrix · System Integration Workflow · Integration Test Plan · Curriculum Map · Policy Package Design
Editorial Notes¶
Form Classification¶
Form family: Analysis, Modeling & Optimization
Rationale: Research Synthesis Protocol operates as an analytical, modeling, inference, comparison, or optimization procedure that derives insight or a solution because it selects and reconciles findings, models, and cases into one internally consistent knowledge product that answers a defined question and stands up to scrutiny.
Independent corroboration: The frozen evidence defines Research Synthesis Protocol as 'Selects and reconciles findings, models, and cases into one internally consistent knowledge product that answers a defined question and stands up to scrutiny', so its operative form is Analysis, Modeling & Optimization.
Nearest alternative: Representation, Specification & Plan — Research Synthesis Protocol includes features of a static representation, map, specification, schema, or prospective plan that externalizes information, but its defining operation is an analytical, modeling, inference, comparison, or optimization procedure that derives insight or a solution.
Review outcome: Independent reviewer agreement; medium confidence.
Origin Attribution¶
Primary origin: Statistics & Experimental Design
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Universal
Rationale: Systematic review methods define questions, eligibility, search and selection, bias assessment, and quantitative or qualitative synthesis before reconciling findings into an answer.
Related originating lineages:
- Ethnography & Qualitative Methods — ethnography_qualitative_methods contributes participatory inquiry and contrasting-case interpretation to the mechanism’s formative or independently convergent form; that contribution does not displace the primary statistics_experimental_design lineage.
- Library & Information Science — library_information_science contributes classification, provenance, access, and persistent-reference stewardship to the mechanism’s formative or independently convergent form; that contribution does not displace the primary statistics_experimental_design lineage.
- Medicine & Healthcare — medicine_healthcare contributes clinical research, consent, response, and recovery practice to the mechanism’s formative or independently convergent form; that contribution does not displace the primary statistics_experimental_design lineage.
Review resolution: The blind reviewers disagreed on primary lineage; authoritative research supports statistics_experimental_design over the competing primary. Systematic review methods define questions, eligibility, search and selection, bias assessment, and quantitative or qualitative synthesis before reconciling findings into an answer. The cited Cochrane Handbook for Systematic Reviews provides direct evidence for that defining form. Alternates are retained only where they contributed an independent formative tradition, while domain_reach=universal records later transfer separately from historical origin.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
Notes¶
[n1] GRADE (Grading of Recommendations Assessment, Development and Evaluation) is a widely used framework for rating the certainty of a body of evidence and the strength of the recommendations drawn from it. It formalizes the move from many studies of varying quality to one graded conclusion — the same reconcile-into-a-defensible-whole discipline this mechanism enacts. ↩