Double-Check the Bot
An AI assistant just gave you an answer — fast, clear, and totally sure of itself. But is it right? Crack four locks to learn how to check before you trust.
The premise
A chatbot can sound confident and still be wrong. It predicts words that fit — it does not truly know if they are true. Read the clues, then open all four locks. Each lock reveals the reasoning behind the answer.
How this breakout builds CLEAR thinking
Every Critical Thinking Online Breakout runs on the CLEAR process — Claim, Lens, Evidence, Alternatives, Response (full framework). Here is where students practice each step in this activity:
- Claim — for each lock, decide what answer it is really asking for, and make it specific.
- Lens — notice what you already assume before you choose.
- Evidence — weigh the six clues; “Lunch Special” is true but off-topic, so it is not evidence for any lock.
- Alternatives — the multiple-choice and multi-select locks force a comparison of competing options.
- Response — the reason revealed after each lock models the justified answer to act on.
Standards alignment
Aligned to Texas TEKS Technology Applications, Grade 6, §126.17 — the AI-adjacent strands (Computational Thinking; Creativity & Innovation; Data Literacy; and Practical Technology Concepts), subsections (c)(7). Skills are paraphrased; read the official standard at tea.texas.gov. This page shows the reasoning focus of each lock — not the answers.
| Lock | Reasoning focus | TEKS substrand |
|---|---|---|
| Lock 1 Confident but Wrong |
Confident but Wrong | (c)(7) Data Literacy — Communicate Results |
| Lock 2 Why It Can Be Wrong |
Why It Can Be Wrong | (c)(7) Data Literacy — Communicate Results |
| Lock 3 Name the Habit |
Name the Habit | (c)(7) Data Literacy — Communicate Results |
| Lock 4 The Verify Habit |
The Verify Habit | (c)(7) Data Literacy — Communicate Results |
In the classroom
- Warm up (5 min). Ask: “Be an Answer Detective” — what does this mean when working with AI? Surface prior experience before the activity.
- Model one clue. Open a single clue together and think aloud: what does this tell us, and which lock might it help?
- Reason, don’t guess. Require students to say why before checking a lock. After each solve, read the revealed reason and compare it to their thinking.
- Spot the decoy. One clue is true but useless. Debrief which one and how they knew — a core evaluation skill that mirrors judging an AI’s output.
- Language support. The 🌐 picker offers 7 languages; pair newcomers to discuss clues in their home language, then answer.
- Extend. Have students find (or safely describe) a real example of the focus skill — a biased result, a hallucination, a good prompt — without using any real product’s screenshots.