Research question

Research Question Formulation

Bridge the gap between what exists (SOTA) and what you'll contribute.

Prerequisite

Run sota-survey first. You need a gap table before formulating RQ.

Workflow

Step 1 — Gap → Problem Statement

For each gap from SOTA survey, apply this template:

Despite [what exists], [limitation/gap] remains unsolved because [root cause].
This causes [concrete consequence] in [application domain].

Example:

Despite federated learning achieving competitive accuracy, communication efficiency in heterogeneous IoT networks remains unsolved because existing methods assume uniform client capabilities. This causes excessive energy drain in resource-constrained healthcare devices.

Step 2 — Formulate Research Questions

Main RQ (1 sentence):

How can [method/approach] be designed to [achieve goal] in [context/constraint]?

Sub-questions (2-4, each answerable in ~1 paper/chapter):

• RQ1: [measurable, specific]
• RQ2: [builds on RQ1]
• RQ3: [validates or extends]

RQ Quality Check:

• Specific — not "improve performance" but "reduce communication rounds by X%"
• Feasible — answerable in your timeline with your resources
• Novel — not already answered in literature
• Significant — matters to the field
• Ethical — no harm, IRB if needed

Step 3 — Hypothesis (if experimental)

H0 (null): [Method A] performs no better than [baseline] on [metric] in [setting]
H1 (alternative): [Method A] achieves [≥X% improvement] over [baseline] on [metric]

Step 4 — Contribution Statement

Structure for proposal and paper Introduction:

## Contributions

This work makes the following contributions:

1. **[Type: Algorithm/Framework/Dataset/Analysis]**: We propose [name], which [what it does] to address [gap].
2. **[Type]**: We demonstrate [empirical result] on [benchmark], showing [magnitude] improvement over [baseline].
3. **[Type]** (optional): We release [artifact] for reproducibility.

Contribution types: Algorithm, Framework, Dataset, Benchmark, Analysis, Survey, System, Theorem.

Step 5 — Scope Boundary

Explicitly state what you will NOT do (saves you from committee scope creep):

Out of scope:
- [Related problem you won't solve and why]
- [Assumption you're making]
- [Limitation you accept]

Output Format

## Research Problem Definition

**Problem:** [1-2 sentence problem statement]

**Main RQ:** [Single question]

**Sub-questions:**
- RQ1: ...
- RQ2: ...

**Hypothesis:**
- H0: ...
- H1: ...

**Contributions:**
1. ...
2. ...

**Out of scope:** ...

Links to Other Skills

• Requires → sota-survey (gap table as input)
• Feeds into → phd-proposal (problem statement section)
• Feeds into → paper-writing (Introduction + Contribution sections)