Ask a struggling research scholar what they are working on and you often hear a field, not a problem: "AI in healthcare." "Employee motivation." "Sustainable construction materials." These are perfectly good places to start and impossible places to finish. A committee cannot approve a direction; an examiner cannot evaluate one. The single most valuable thing you can do in your first year is convert your interest into a researchable problem — and there is a method to it.
What makes a problem researchable
A researchable problem has four properties. It is specific (one phenomenon, one population, one context — not a field). It is answerable with evidence you can realistically collect (data, experiments, documents, participants you can actually access). It is unanswered (a genuine gap, verified in the literature, not assumed). And it is worth answering (someone — a discipline, an industry, a policy-maker — should care about the result).
Most stuck proposals fail the first test; most rejected ones fail the third.
The narrowing ladder
Work down these five rungs, writing a sentence at each level. Using "AI in healthcare" as the running example:
- Field — Artificial intelligence in healthcare.
- Topic — Machine-learning models for early disease detection.
- Aspect — Why clinically validated detection models fail when deployed in new hospitals.
- Problem — Existing diabetic-retinopathy detection models lose accuracy on Indian patient populations because training data under-represents them, and no systematic method exists for assessing this degradation before deployment.
- Question — To what extent does population-specific fine-tuning restore the accuracy of pre-trained diabetic-retinopathy models on Indian screening data, compared with full retraining?
Notice what happened between rungs three and five: a complaint about the world became a comparison you can actually run. That transformation — from "X is a problem" to "what happens to A when B, compared with C" — is the heart of the exercise.
Three tests before you commit
The data test. Write down exactly what evidence answering your question requires, then ask honestly: can you get it within your registration period, with your resources and permissions? A brilliant question you cannot collect data for is a trap with a deadline.
The gap test. Search the literature for your rung-four sentence — not your rung-one field. You are looking for the specific claim. Finding fifty papers on your field is expected; finding three on your exact problem is a good sign; finding one that already answers your question means you climb back one rung and pick a different aspect.
The so-what test. Complete the sentence: "If this question is answered, then ___ can ___ differently." If nothing fills the blanks, the problem is researchable but not worth researching — a distinction committees feel even when they do not name it.
A note on how long this should take
Scholars often feel guilty spending two months on problem formulation. The arithmetic says otherwise: a vague problem silently taxes every later stage — the literature review sprawls, the methodology cannot be pinned down, the objectives multiply. Two focused months here routinely save a year later. The proposal is written when you know the least; this process is how you know enough.
Problem refinement and proposal development are where most ResearchGiri engagements begin. If you are circling a field and cannot land on the problem, that is exactly the conversation to bring us.