Paper Type Playbooks

This page gives standard example routes for the four canonical paper types:

• systematic-review
• empirical
• methods
• theory

These are not the only valid routes. They are the recommended defaults when you want a defensible baseline workflow.

How To Read These Examples

Each playbook includes:

• a recommended route
• a narrower route for lighter work
• key skills usually involved
• typical outputs
• a starter command

Use them as operating defaults, then narrow or deepen based on your actual constraints.

1. Systematic Review

Use this when

You are building a PRISMA-style review, evidence synthesis, or structured related-work base with transparent search and screening logic.

Recommended route

1. A1: clarify question and scope
2. B1: run reproducible search
3. B1_5: refine concepts and Boolean logic
4. B2: extract papers
5. B3: map the literature
6. E1: synthesize evidence
7. E2: assess quality / risk of bias
8. G1: run PRISMA check
9. F3: write the review manuscript

Narrower route

Use this when you already have a stable corpus:

1. B2
2. B3
3. E1
4. F3

Typical skills

• academic-searcher
• concept-extractor
• paper-screener
• paper-extractor
• literature-mapper
• evidence-synthesizer
• quality-assessor
• prisma-checker
• manuscript-architect

Typical outputs

• search log
• screening log
• extraction table
• literature map
• synthesis memo or meta-analytic result
• quality assessment
• PRISMA compliance report
• manuscript draft

Starter command

python3 -m bridges.orchestrator task-run \
  --task-id B1 \
  --paper-type systematic-review \
  --topic ai-in-education \
  --cwd . \
  --research-depth deep

Common narrowing rule

If the review starts producing too many auxiliary artifacts, stay with:

• B2
• B3
• E1
• F2 or F3

and use --focus-output plus --output-budget.

2. Empirical Paper

Use this when

You are writing a standard empirical paper with a design, dataset, analysis, interpretation, and submission path.

Recommended route

1. A1: define question
2. A1_5: generate hypotheses
3. C1: build the design
4. C2 / C3: operationalize variables and validate data path
5. C4: specify robustness logic
6. I1 / I2 / I3 or Stage-I code lane if implementation is substantial
7. F1: manuscript structure
8. F3: full draft
9. F4: tables/figures/results support
10. G2: reporting check
11. H1: submission package

Narrower route

Use this when the study is already run and you mainly need writing plus checks:

1. F1
2. F3
3. F4
4. G2
5. H1

Typical skills

• question-refiner
• hypothesis-generator
• study-designer
• variable-constructor
• dataset-finder
• robustness-planner
• analysis-interpreter
• table-generator
• figure-specifier
• reporting-checker

Typical outputs

• question and hypothesis set
• design spec
• variable / dataset plan
• robustness plan
• manuscript draft
• tables and figure specs
• reporting compliance memo
• submission bundle

Starter command

python3 -m bridges.orchestrator task-run \
  --task-id C1 \
  --paper-type empirical \
  --topic policy-effects \
  --cwd .

Common decision rule

If code is light, stay in the writing/design route. If code becomes central to the paper, switch into the full Stage-I code lane instead of using generic drafting alone.

3. Methods Paper

Use this when

The core contribution is a method, algorithm, pipeline, or code-supported procedure, and the code is first-class evidence.

Recommended route

1. A1: define problem and contribution claim
2. A3: frame theory or methodological positioning
3. C1: state evaluation design
4. I5: code specification
5. I6: zero-decision plan
6. I7: implementation and profiling
7. I8: review logic and statistical validity
8. I4: reproducibility audit
9. F1: manuscript structure
10. F3: methods paper draft
11. H3: peer-review simulation for harsh stress test

Narrower route

Use this when you are still locking implementation before building:

1. A1
2. C1
3. I5
4. I6

Typical skills

• theory-mapper
• study-designer
• code-specification
• code-planning
• code-execution
• code-review
• reproducibility-auditor
• stats-engine
• manuscript-architect

Typical outputs

• method positioning memo
• evaluation design
• code specification
• execution plan
• performance profile
• code review
• reproducibility audit
• methods manuscript draft

Starter command

python3 -m bridges.orchestrator code-build \
  --method "Staggered DID" \
  --topic policy-effects \
  --domain economics \
  --focus full \
  --paper-type methods \
  --cwd .

Common decision rule

If you are unsure whether the code lane is necessary, ask:

• Is code a core contribution?
• Will reviewers evaluate reproducibility and implementation quality directly?
• Does the paper need strict audit artifacts such as code_review.md and reproducibility_audit.md?

If yes, use the Stage-I route.

4. Theory Paper

Use this when

The paper’s main contribution is conceptual, theoretical, or mechanism-building rather than data-heavy execution.

Recommended route

1. A1: refine the core question
2. A1_5: turn it into propositions
3. A2: map the theory base
4. A4: identify unresolved theoretical gap
5. B2: targeted literature extraction
6. E1: synthesize conceptual evidence
7. F1: design the manuscript logic
8. F3: full theory draft
9. G4: tone tightening
10. H4: fatal-flaw stress test

Narrower route

Use this when the theory base is already stable:

1. A2
2. A4
3. F1
4. F3

Typical skills

• question-refiner
• hypothesis-generator
• theory-mapper
• gap-analyzer
• paper-extractor
• evidence-synthesizer
• manuscript-architect
• tone-normalizer
• fatal-flaw-detector

Typical outputs

• refined conceptual question
• propositions
• theory map
• theoretical gap memo
• theory manuscript draft
• style normalization log
• fatal-flaw memo

Starter command

python3 -m bridges.orchestrator task-run \
  --task-id A2 \
  --paper-type theory \
  --topic organizational-ai-governance \
  --cwd .

Common decision rule

Do not over-import the code lane into a theory paper unless the method or simulation is itself part of the contribution.

Cross-Playbook Advice

When to go deeper

Go deeper when:

• reviewers will expect reproducibility or checklist evidence
• the paper type itself has strong reporting standards
• the evidence base is contested or heterogeneous
• you need stronger adversarial review

When to stay narrow

Stay narrow when:

• you already have stable inputs
• the task is a revision rather than a greenfield build
• artifact sprawl is becoming expensive
• you only need one core deliverable

Useful controls:

• --focus-output
• --output-budget
• --research-depth deep
• --only-target

Related Pages

• Need scenario-based task guidance: Task Recipes
• Need the global skill map: Skills Guide
• Need exact command flags: CLI Reference