Skip to content

2 · The scientific method (your map)

Every good science project follows the same path. Scientists call it the scientific method — it's just a careful, step-by-step way to ask a question and find a trustworthy answer instead of guessing. Science Buddies, a site teachers and science-fair students use all over the world, lays the steps out like this:

StepWhat you doIn plain words
1. QuestionAsk something you can test"Can an AI tell my dog from my cat by photo?"
2. Background researchFind out what's already knownRead trusted sources — and cite them
3. HypothesisWrite a testable guess"I think it will be right at least 8 out of 10 times."
4. ExperimentTest the hypothesis fairlyTrain the model, then test it on new photos
5. Analyze + concludeLook honestly at the data"It was right 7/10 — close, but not quite my guess."
6. CommunicateShare results so others can checkBuild a poster, present, and cite your sources

Three things make this the scientific method and not just "doing a project":

  • Your guess is testable. A good hypothesis can turn out wrong — and that's allowed. "AI is cool" can't be tested. "The model will be right at least 8 of 10 times on new photos" can.
  • You change one thing at a time. That one thing you change is called the variable. If you change ten things at once, you won't know which one mattered.
  • You report what really happened. Not what you wished happened. A surprising result is still a real result — sometimes the best ones.

Notice that AI is a tool inside the method, not a replacement for it. AI might help you research, train a model, or organize your data — but you ask the question, you run the test, and you decide what the results mean. That's the difference between a scientist who uses AI and someone who just lets AI do the thinking.

Think about it. Turn this into a testable hypothesis: "I think AI is good at sorting recycling." (Hint: how good? right how many times out of how many?)

Sources