Lesson · Investigation & evidence

The detective’s notebook

Students crack a mystery using clues, hypotheses and tests. Then they interview AI as a witness, and learn that a witness who sounds sure isn’t the same as a detective who has checked the evidence.

Level 1 AI after thinkingGrades 4–1050 minutesScience · Any subjectPaper + 1 teacher screen

The big idea

AI can be a witness. You are the detective.

Students will be able to, without AI:

Generate several hypotheses from clues
Match evidence for and against each hypothesis
Design a simple test
Sort a witness’s statements by how reliable they are

Lesson flow · the 5 P’s

Think first. Then the tool.

Most of the lesson happens on paper. AI appears for one short, structured phase, and the teacher can run it from a single projector screen. How the 5 P’s work →

Ponder

15 min○ Screens off · paper only

Hand out (or read) the case file. Teams list every possible explanation before choosing one.

The case of the wilting plant: The class plant is drooping. Its lowest leaves are yellow. It is watered every day. The soil is always damp and smells sour. The pot has no hole in the bottom. It sits near a sunny window.
List at least three hypotheses (too much water, too little light, too hot, pests, not enough food…).
For each, write the clues that support it and the clues against it.
Rank them from most to least likely.

Say to students“Detectives don’t start with a suspect. They start with the evidence.”

Plan

8 min○ Screens off · paper only

Teams plan how they would prove it.

Pick your top hypothesis.
Design a test that could prove it wrong. (What would you change? What would you watch?)
Write two questions for the “witness.”

Partner

10 min● AI on · teacher screen or lab

On the teacher’s screen, teams interview AI as a witness. Type exactly what students ask.

Start with the opening question below. Then each team asks one of its two prepared questions.
Students record each statement the AI makes. Notice: did it ask for more clues, or jump to an answer?

My plant is drooping and its lowest leaves are turning yellow. What could cause that? Answer in under 80 words.

[Team question, e.g.] Could it be too much water if the soil smells sour?

Probe

10 min○ Screens off · paper only

Screens off. Teams sort the witness statements.

Sort each AI statement: matches our clues, general information, a guess, or contradicts our evidence.
Which statements could the AI not have known without seeing the plant?
Did the AI change your ranking? Why or why not?

Prove

7 min + optional test○ Screens off · paper only

Teams write a case report in their own words.

Conclusion, the three strongest pieces of evidence, and what would change your mind.
Optional: test it for real. Gently slide the plant out of its pot and look at the roots. Healthy roots are white and firm; rotting roots are soft and brown or black. If the roots are white and firm, the root-rot hypothesis is wrong. Then test the watering idea fairly: change only one thing (for example, water only when the top inch of soil feels dry) and record the leaves for two weeks.
Teacher key: The clues point to overwatering with no drainage, leading to root rot (sour smell, constantly damp soil, yellowing lower leaves).

Printable

Student worksheet

Print one per student. Everything students hand in is handwritten, so their thinking is visible.

Detective’s notebook

Case file: the wilting plant

Name:Date:Class:

Hypotheses (at least three), then rank them (1 = most likely)

Clues for and against each hypothesis

My test: what will I change and what will I watch?

Two questions for the witness

Witness statements (AI) and how I sorted themM = matches clues · G = general info · ? = guess · X = contradicts evidence

Case report: conclusion, strongest evidence, what would change my mind

Assessment

Grade the thinking, not the output

Every row scores something students did themselves. An AI can produce a product; it can’t produce a student’s reasons.

Criteria	4 · Excellent	3 · Solid	2 · Developing	1 · Beginning
Hypotheses	Several plausible hypotheses tied to clues	Two or more plausible hypotheses	One hypothesis	None
Use of evidence	Weighs evidence for and against each	Uses evidence for the top hypothesis	Mentions clues loosely	No evidence
Test design	Test could clearly prove the hypothesis wrong	Reasonable test	Vague test	No test
Witness sorting	Accurately sorts and explains reliability	Sorts most statements	Sorts a few	Accepts all statements

Why this is hard to cheat

The reasoning happens in teams on paper, and the AI is just one witness whose statements must be judged against the class’s evidence. For a case nobody can look up, write one from your own classroom or school.

Adjust it

Younger students: Use picture clue cards and only two hypotheses. The teacher acts as the AI witness’s typist and reader.

Older students: Write your own case file from a real local problem (a school energy bill spike, a drop in lunch participation) and investigate it.

No devices at all? Print or read aloud these three sample AI witness statements (one helpful, one general, one confidently wrong) for teams to sort.

1. “Your plant is definitely thirsty. Drooping leaves always mean underwatering, so water it twice a day.” 2. “Yellow lower leaves on a plant whose soil stays wet and smells sour usually point to overwatering. Without a drainage hole, water collects at the bottom and the roots can rot.” 3. “Most houseplants need light, water and nutrients to stay healthy, and different plants need different amounts.”

Key: 1 = X, contradicts the evidence (the plant is watered every day and the soil is always damp). 2 = M, matches the clues. 3 = G, general information.

Discussion questions

Why is a confident witness not always a reliable one?
What questions make a witness more useful?
Where do real investigators (doctors, mechanics, scientists) use this same process?
When should you trust an AI’s answer?

Extensions & connections

Run the real experiment and graph the results.
Connect to how scientists check what AI finds: AI in space.
Pair with Fact-Check the Bot.

Put curiosity to work

Read it. Try it. Question it.

Explore new AI research, or take a paper-first investigation into your classroom.

Classroom starter pack →Research explained →