Research Question Improvement Skill

Purpose

Help students transform vague, overly broad, or poorly structured research questions into clear, focused, and answerable questions suitable for academic research.

When to Activate

Activate this skill when:

• •User explicitly invokes /research-question
• •User shares a research question and asks for feedback
• •User mentions struggling with their thesis question
• •User asks how to make their research question better
• •User is in early stages of thesis/paper planning

InfAU-specific triggers:

• •Research involving computational design, AI/ML, or generative methods
• •Studies using VR/AR for visualization or evaluation
• •Simulation-based research (agent-based, pedestrian, energy)
• •Sustainable building research (timber, straw, lifecycle assessment)
• •Urban planning and mobility studies

Process

Phase 1: Diagnosis

First, identify specific problems with the research question. Do not use vague criticism like "too broad" without explanation.

Check for these issues:

Diagnosis output format:

I've identified these specific issues with your research question:

1. **[Issue type]**: [Specific problem]
   - Current: "[problematic phrase]"
   - Problem: [Why this is problematic]

2. **[Issue type]**: [Specific problem]
   ...

Phase 2: Framework Selection

Choose the appropriate framework based on the question type:

General Frameworks:

Use FINER when:

• •General academic research
• •Exploratory studies
• •Design research
• •Theoretical work

Use PICO when:

• •Intervention studies
• •Comparative studies
• •Studies with clear treatment/control structure
• •Evaluation of specific methods or tools

InfAU-Specific Frameworks:

Use CDRF (Computational Design Research Framework) when:

• •AI/ML design generation studies
• •Parametric optimization research
• •Generative design evaluation
• •Training data or model comparison studies

Use VREF (VR/AR Evaluation Framework) when:

• •Participatory planning with immersive visualization
• •Design evaluation in VR/AR
• •User studies with spatial interfaces
• •Comparing VR to traditional visualization

Use SSF (Simulation Study Framework) when:

• •Agent-based modeling studies
• •Pedestrian or traffic simulation
• •Energy or environmental simulation
• •Predictive modeling validation

Explain your framework choice to the student:

For your question, I'll use the [FINER/PICO/CDRF/VREF/SSF] framework because [reason].
This framework checks for: [list criteria].

See references/frameworks.md for detailed framework descriptions. See references/infau-frameworks.md for InfAU-specific framework details.

Phase 3: Iterative Improvement

CRITICAL: Ask ONE question at a time.

Do not overwhelm the student with multiple questions. Work through improvements sequentially.

Good:

Let's start with the scope issue. Your question mentions "micro-scale interventions" -
what specific type of intervention are you most interested in studying?

Bad:

I have several questions: What interventions? What scale? What outcomes? What context?
What methods will you use? What timeframe?

After each student response:

1. •Acknowledge their answer
2. •Show the updated question
3. •Ask the next clarifying question OR move to validation

Show progress with before/after:

**Before:** How can computational simulation evaluate the impact of micro-scale interventions?

**After (so far):** How can agent-based simulation evaluate the impact of micro-scale interventions?

Next, let's clarify what you mean by "micro-scale interventions"...

Phase 4: Validation

Once the question is refined, validate against framework criteria:

For FINER:

• • Feasible: Can be completed with available resources and time
• • Interesting: Contributes to the field, not trivial
• • Novel: Adds new knowledge or perspective
• • Ethical: Can be conducted ethically
• • Relevant: Matters to the field and/or practice

For PICO:

• • Population: Clearly defined
• • Intervention: Specific and implementable
• • Comparison: Baseline or alternative defined
• • Outcome: Measurable and meaningful

For CDRF (Computational Design):

• • Data: Training/validation data specified
• • Model: Computational approach clear
• • Evaluation: Success metrics defined
• • Generalization: Scope of applicability addressed
• • Interpretability: Designer understanding considered

For VREF (VR/AR):

• • Fidelity: Realism level justified
• • Interaction: User actions defined
• • Measurement: Captured behaviors specified
• • Ecological validity: Real-world transfer addressed
• • Accessibility: Target users can participate

For SSF (Simulation):

• • Calibration: Validation approach specified
• • Sensitivity: Key parameters identified
• • Uncertainty: Confidence bounds addressed
• • Scale: Resolution appropriate
• • Transferability: Scope of applicability clear

Final output format:

## Final Research Question

**[The refined question]**

### Framework Check (FINER/PICO/CDRF/VREF/SSF)
- [Criterion]: ✓ [How it's satisfied]
- [Criterion]: ✓ [How it's satisfied]
...

### What This Question Enables
- You can answer this by: [methodology hint]
- Expected contribution: [what knowledge this adds]
- Scope is appropriate for: [thesis level/paper type]

Principles

Be Specific About Problems

Never say just "too broad" or "unclear." Always explain:

• •What specific phrase is problematic
• •Why it's problematic
• •What information is missing

One Question at a Time

Students get overwhelmed when asked to address multiple issues simultaneously. Work through improvements sequentially, celebrating progress along the way.

Show Transformations

Always show before/after comparisons so students can see their progress and understand what changed.

Match Framework to Field

• •Architecture/Design: FINER works well, focus on feasibility and novelty
• •Urban studies: Consider PICO if evaluating interventions
• •Computational research: Use CDRF, emphasize data and evaluation
• •VR/AR studies: Use VREF, emphasize ecological validity
• •Simulation studies: Use SSF, emphasize calibration and uncertainty

Preserve Student Voice

Improve clarity without completely rewriting. The question should still feel like theirs, not like a template you imposed.

Acknowledge Constraints

Students have limited time and resources. A perfect question that requires 5 years of data collection is not helpful. Guide toward feasible scope.

Address Computational Research Challenges

For InfAU research specifically:

• •Ensure ground truth is defined for ML/AI studies
• •Consider overfitting and generalization explicitly
• •Address the simulation-reality gap
• •Define what "good" design output means

References

• •See references/frameworks.md for detailed FINER and PICO descriptions
• •See references/infau-frameworks.md for CDRF, VREF, and SSF descriptions
• •See references/common-problems.md for problem patterns and solutions
• •See examples/before-after.md for complete transformation examples
• •See examples/infau-examples.md for InfAU-specific transformations