research-architect-design

Research Architect Study Design

Use this branch after confirmed_research_spine.md exists.

Blocking rule

When the confirmed research spine is missing, return to research-architect-literature for user confirmation. When a target paper exists, also require a usable exemplar_adaptation_plan.md.

Inputs

Read:

paper_output/confirmed_research_spine.md
paper_output/sota_gap_map.md
paper_output/source_inventory.md
paper_output/project_config.json
paper_output/exemplar_logic_profile.md       # when target references exist
paper_output/exemplar_adaptation_plan.md     # when target references exist

Read references/design-family-grammars.md and select the grammar that matches the target paper, question, and available evidence.

Required outputs

Create:

paper_output/study_design_brief.md
paper_output/analysis_plan.md
paper_output/evaluation_plan.md
paper_output/study_component_registry.md
paper_output/threats_to_validity.md

For experiment-centered or computational projects, also create:

paper_output/experiment_registry.md

Read an existing experiment_registry.md as a compatible specialized view of the study component registry.

Study design table

| Component | Decision | Rationale | Reference Principle Adapted | Claim Supported | Risk | Mitigation | |---|---|---|---|---|---|---| | Research question | | | | | | | | Theoretical/conceptual orientation | | | | | | | | Unit, case, population, or corpus | | | | | | | | Evidence source or research material | | | | | | | | Sampling or selection logic | | | | | | | | Data-generation or source-construction procedure | | | | | | | | Analysis or interpretation approach | | | | | | | | Comparison, counterfactual, rival explanation, or warrant | | | | | | | | Evaluation or credibility criterion | | | | | | | | Sensitivity, triangulation, or scope check | | | | | | | | Negative case, counterexample, or failure condition | | | | | | | | Ethics, access, and researcher-position considerations | | | | | | | | Inference and transfer boundary | | | | | | |

Use only rows that serve the selected design family and claims.

Study component registry

Use:

| Component ID | Research Function | User Action | Evidence or Material | Analysis or Interpretation | Comparison or Warrant | Output | Claim Supported | Completion Criterion | Status | |---|---|---|---|---|---|---|---|---|---|

This registry is the user's action plan. Every row should explain what to do, why it is needed, what artifact it produces, and which claim or boundary it serves.

Design families

Supported families include:

• quantitative descriptive or associational;
• quantitative causal or explanatory;
• qualitative interpretive;
• case study, comparative, historical, archival, or process-tracing;
• mixed methods;
• theoretical, conceptual, critical, or normative;
• systematic, scoping, or integrative review;
• computational, experimental, or design/build/evaluation.

Hybrid studies may combine grammars when the integration point is explicit.

Adaptive terminology

Translate the target paper's function rather than copying its vocabulary:

• baseline/control -> comparison, counterfactual, contrast, negative case, rival explanation, or counterargument;
• metric -> evaluation, credibility, or warrant criterion;
• robustness -> sensitivity, triangulation, rival-explanation, dependability, or scope check;
• figure -> evidence display, including tables, quotations, matrices, process maps, conceptual models, and joint displays;
• generalization -> statistical generalization, analytical generalization, theoretical transfer, or contextual transferability.

Rules

• Every study component must support a claim, rule out an alternative, establish credibility, or define scope.
• Every recommended action should point to the reference-paper function it adapts.
• Match validity criteria to the design family.
• Quantitative association does not support causal language without a causal design.
• Qualitative interpretation requires context, selection logic, researcher role, analytic transparency, and negative-case or alternative-interpretation checks.
• Case and historical work requires case/source selection, source criticism, rival explanations, and bounded inference.
• Conceptual and normative work requires explicit premises, argument steps, counterarguments, and applicability boundaries.
• Mixed-method conclusions require a declared integration point.
• Return to literature or brainstorming when no feasible design can support the selected spine.