What you will learn
- Turn an interesting topic into a claim that can be tested.
- State what evidence would weaken or falsify the working hypothesis.
- Connect the problem, objectives, methods, milestones, and deliverables.
- Use AI to support organization and drafting without treating it as evidence.
Begin with a research problem, not only a topic
A topic identifies the area you want to study; a research problem explains what is unresolved, consequential, or insufficiently understood. Before choosing methods, write the problem in concrete terms: what is happening, why it matters, what is already known, and what remains uncertain. This framing gives the literature review and the eventual research question a clear purpose.
Make the hypothesis vulnerable to evidence
A useful hypothesis is not protected from failure. It identifies a relationship, mechanism, or expected outcome that observations could contradict. Ask, “What result would make me change my mind?” If no plausible result could do that, the claim needs to be narrowed or reformulated before data collection begins.
Design the project backward from the evidence
Once the research question is clear, identify the evidence needed to answer it. Then choose variables, comparisons, data sources, and analytical methods that can produce that evidence. This backward sequence prevents a common failure: collecting substantial data that never resolves the original question.
Translate the design into milestones
Milestones should mark meaningful states of readiness, not merely dates on a calendar. Examples include an approved question set, validated instruments, a complete pilot, a defensible analysis plan, or evidence sufficient for a committee review. Each milestone should have a clear output and a criterion for deciding whether the project can advance.
Keep AI in a supporting role
AI can help generate alternatives, organize notes, test the clarity of an explanation, and improve a draft. It cannot establish that a scientific claim is true. Verify generated material, protect confidential information, document consequential use, and keep experimental evidence and researcher judgment at the center of the project.
Frequently asked questions
What should a graduate research plan include?
At minimum, define the problem, research questions or hypotheses, relevant literature, evidence requirements, methods, risks, milestones, deliverables, and the criteria for deciding when each stage is ready to advance.
What makes a hypothesis falsifiable?
A hypothesis is falsifiable when you can identify an observation or result that would count against it. The goal is not to guarantee failure, but to make the claim genuinely testable.
Should I choose a methodology before writing the research question?
Usually no. The research question and evidence requirements should guide the methodology. Choosing a familiar method first can force the project into a design that does not answer the question.
How can AI be used responsibly in a research plan?
Use it for brainstorming, organization, and communication support. Verify outputs, do not treat generated text as evidence, protect sensitive information, and follow institutional disclosure and integrity requirements.
Episode transcript
Transcript supplied from the published episode script and lightly formatted for readability.
Read the full transcript
Every research project begins with an idea. Some ideas become important discoveries. Others quietly disappear because they cannot survive careful testing.
So how do you tell the difference?
A useful answer comes from the physicist Richard Feynman, who famously observed that it doesn't matter how elegant a theory is or how intelligent its author may be. If the theory does not agree with experiment, it is wrong.
That simple statement captures the foundation of scientific research.
Research is not about proving that your idea is correct. It is about giving your idea every reasonable opportunity to fail. If it survives those tests, your confidence in it grows. If it fails, you have still learned something valuable because you have eliminated one possible explanation.
This way of thinking is very different from everyday argument. Outside of research, people often defend their opinions. Researchers do the opposite. They actively look for evidence that challenges their own assumptions.
One of the most important concepts supporting this mindset is called falsifiability.
A hypothesis is falsifiable when it can be tested in a way that allows it to be proven wrong. If there is no possible observation or experiment that could contradict your claim, then the claim cannot be evaluated scientifically.
Imagine someone tells you there is an invisible dragon living in their garage.
You look inside and see nothing.
They explain that the dragon is invisible.
You suggest spreading flour on the floor to reveal footprints.
They respond that the dragon floats.
You propose using a thermal camera.
They reply that the dragon breathes fire without producing heat.
Every possible test is met with another exception.
Eventually you have to ask an important question.
What practical difference exists between an undetectable dragon and no dragon at all?
This famous example, popularized by Carl Sagan, reminds us that explanations which cannot be tested cannot be strengthened by evidence.
The same principle applies to engineering, computing, and scientific research.
Before investing weeks or months into a project, ask yourself one question.
What observation or experiment would convince me that my hypothesis is wrong?
If you cannot answer that question, your research idea probably needs more work before you begin collecting data.
Artificial intelligence makes this even more important.
Large language models are excellent assistants for brainstorming, organizing information, and improving communication. They are far less reliable when asked to judge whether your scientific reasoning is correct. They often produce persuasive explanations that sound convincing but have not been experimentally verified.
Treat AI as a research assistant, not as experimental evidence.
Your strongest defense against weak research is still the same today as it has always been: a clearly defined hypothesis, a well-designed experiment, and the willingness to change your mind when the evidence demands it.
That mindset is the real foundation of good research.
Continue the SciResMethods series
Continue to Episode 2: How to Find a Research Gap for a Thesis