A vague question gets a vague answer. The Three-Why Game turns
what you kind of want to know into something you can actually search.
Learning⏱ 20 min💻 Device optional
The big idea
The quality of your answer depends almost entirely on the quality of your question.
Asking better questions is a skill — and it's one you can practise in five minutes.
The Three-Why Game — from vague to searchable
Most searches start too broad. Instead of rewording the same vague question
three times, try asking yourself "why" three times — each answer brings you
closer to what you actually want to know.
How to play: Start with the broad question. Ask "why do I want to know this?"
Use the answer to form a better question. Ask "why" again. By the third round,
you have a question specific enough to get a useful answer.
💬Start
Original question (vague)
"Why is the sky different colours at different times of day?"
❓Why 1
Why do I want to know this? → I want to understand what changes — is it the light, the atmosphere, something else?
Better question
"What causes the colour of sunlight to change — is it the sun, the atmosphere, or both?"
❓Why 2
Why does that matter? → I want to know if the answer is different for sunrise vs sunset, or if it's the same process.
Even better
"Is the physics of a red sunset the same as a red sunrise — does the same scattering happen in both directions?"
🎯Why 3
Why do I care about that specifically? → I'm trying to write a scene accurately — I need to know if the colours are literally identical or just similar.
Searchable question
"Rayleigh scattering sunrise vs sunset — are the colours physically identical?"
Try it yourself: Pick something you've been wondering about lately.
Run it through three whys. The question you end up with is the one worth searching —
and it will return a much more useful answer than the one you started with.
Why Google and AI return different answers
Both tools answer questions — but they work differently, and those differences
matter for deciding which one to use and how much to trust the answer.
Search engines (Google)
Index existing web pages. Return links to documents that humans wrote, edited, and published. The algorithm decides which pages are most relevant.
When the result is from a reliable source (university, government, well-known science publication), it's citable and stable. You can check who wrote it.
Best for: facts, citations, current events
AI chat tools (Claude, ChatGPT)
Generate text based on patterns in training data. Synthesise information from many sources — but don't always know which sources they're combining, and may have a training data cutoff date.
Sound confident even when uncertain. Better for explaining concepts and exploring ideas than for verifying specific facts.
Best for: explanations, brainstorming, synthesis
Neither is always right. Search results can surface unreliable pages.
AI tools can generate confident-sounding errors. The skill is knowing which kind of
answer each tool is good at giving — and when to check a second source.
The comparison activity — same question, two tools
Read both answers below to the same question. Then work through the comparison
questions with your caregiver. You don't need a device — both answers are
printed here.
Activity · With caregiver · No device needed
The question
How do migratory birds navigate?
🔍 Search engine result (summary from a science source)
Migratory birds use several independent navigation systems simultaneously.
They detect Earth's magnetic field using magnetite crystals in the beak
and a light-sensitive protein (cryptochrome) in the eye that may allow
them to "see" magnetic field lines. They also use a sun compass —
calibrating direction against the sun's position and the time of day —
and, at night, navigate by star patterns, particularly the rotation point
of the night sky. Studies removing individual cues show birds can compensate
using the remaining systems.
Source: journal article, peer-reviewed, 2022
🤖 AI tool response
Migratory birds navigate using a combination of the Earth's magnetic field,
the position of the sun, star patterns, and landmarks like coastlines and
rivers. Research suggests they may have specialised cells that help them
detect magnetic fields, and young birds often learn routes partly by following
experienced birds on their first migration. Different species rely on these
cues in different proportions — some are more dependent on magnetism,
others more on visual landmarks.
What did both answers agree on?
What did the search result include that the AI didn't — or vice versa?
The search result named a specific protein (cryptochrome) and cited a year. The AI answer didn't. Does that make one more trustworthy than the other?
Which answer would you cite in a school project? Why?
Which answer helped you understand the topic better?
The pattern you're looking for: Use the search result to get the
citable, specific fact. Use the AI explanation to help you understand what the fact
means. Then go back and verify the understanding against the source — not the other
way around.
The one question that makes any source more useful
The metacognitive question
"What would have to be true about where this source comes from for this answer to be correct?"
This sounds abstract, but it's one of the most practical questions you can ask.
It stops you from just accepting an answer and makes you think about
what kind of thing is capable of giving a reliable answer to this particular question.
Try it on the bird navigation answers:
For the search result to be correct: the journal would need to be peer-reviewed,
the methodology would need to be sound, the finding would need to have been
replicated. Those are things you can check — not in five minutes, but in principle.
For the AI answer to be correct: the training data would need to have included
accurate scientific sources, and the model would need to have combined them correctly
without introducing errors or outdated information. Those are things you generally
cannot check — which is why AI answers need verification against citable sources
for anything that matters.
When answers disagree — what to do
🔍
Check the source behind the search result
Who published it? When? For what audience? A government health agency and
a personal blog can appear in the same search results — they don't carry the
same weight.
📅
Check whether the AI answer might be outdated
AI tools have a training data cutoff — they don't know about events or
research published after that date. For anything in a fast-moving field
(medicine, technology, current events), prefer a dated search result.
📰
Find a third source
Two disagreeing sources tell you there's a question to resolve.
A third independent source that agrees with one of them tells you which
direction the evidence leans. This is what "do your research" actually means.
❓
Ask a better question
Sometimes disagreement signals that your original question was too vague —
the sources are answering different versions of it. Run the Three-Why Game
again and see if a more precise question eliminates the conflict.
✅ Caregiver check-in
Can your child run the Three-Why Game on a question right now — and get to a specific, searchable version in three steps?
In the comparison activity: which answer would they cite in a school project, and why?
Can they state the difference between what a search engine does and what an AI tool does in one sentence each?
Can they apply the metacognitive question to one of the two bird navigation answers? (What would have to be true for this to be correct?)
When they get two disagreeing answers: what's their next step?
🎉 10–12 cohort complete
This is the final activity in the Ages 10–12 set. All 9 modules are now live.
Visit the 10–12 activity page to see the full set,
or explore the Ages 4–6 and
Ages 7–9 cohorts.