Keep the AI Conversation Going.

Give teams three polished AI answers about a current class topic, including one answer with a planted error. Students underline every claim that needs evidence and rank the answers from easiest to hardest to trust.

Debrief: What made an unsupported answer feel believable?

Students choose one factual sentence from an AI response and build a three-step trail back to a reliable original source. They record where the claim began, who published the evidence, and whether the evidence actually supports it.

Debrief: When does a link count as evidence, and when is it only decoration?

Present a technically accurate but incomplete AI summary. Pairs list the people, dates, viewpoints, definitions, or conditions a reader would need before making a decision from it, then write two follow-up prompts.

Debrief: Can a statement be accurate and still mislead someone?

Students sort twelve AI-generated statements into fact, forecast, opinion, and unknown. For each fact, they name a way to verify it; for each forecast, they identify the assumption behind it.

Debrief: Which category was easiest for AI to blur?

Teams compare an AI explanation with a textbook, primary document, lab result, or trusted database. They create a short referee report naming agreements, conflicts, and the strongest evidence.

Debrief: What should you do when two credible-looking sources disagree?

Pairs receive five vague prompts such as “help with science.” They repair each one by adding a learning goal, useful context, boundaries, and a request for questions or hints instead of a finished answer.

Debrief: Which added detail improved the response most?

Students write a four-rung prompt sequence for a difficult problem: ask for a reminder, then a clue, then a similar example, and only then feedback on their own attempt. They stop at the first rung that gets them unstuck.

Debrief: How can asking for less help produce more learning?

Groups improve a prompt by adding three meaningful constraints, such as approved sources, vocabulary level, response format, or topics to avoid. They predict how every constraint will change the output before testing it.

Debrief: Which constraints improved quality, and which only made the prompt longer?

Students rewrite one assignment prompt three ways so AI acts as a coach, a critic, and a creator. They decide which role preserves the most student thinking and mark any version that would violate the teacher’s rules.

Debrief: When does helpful feedback cross into doing the work?

Students ask for explanations of one course concept for a younger child, a classmate, and a knowledgeable adult. They compare vocabulary, examples, missing details, and accuracy across the three versions.

Debrief: Why should a simpler explanation still be checked?

Using a real upcoming assignment, students make a green-yellow-red map of possible AI uses. Green is clearly allowed, yellow requires teacher permission, and red replaces the learning the assignment is meant to show.

Debrief: Which yellow-zone use would you ask about first?

Students draft accurate one-sentence notes explaining how AI did or did not shape brainstorming, feedback, translation, coding, or revision. Partners check whether each note is specific enough to be honest.

Debrief: What information would a teacher need to understand your process?

Learners design a simple process record for an assignment with checkpoints for original notes, prompts used, useful output, rejected output, revisions, and sources checked. Then they complete one checkpoint with sample work.

Debrief: Which receipt best proves that the final thinking is yours?

Read short situations that gradually move from studying with AI to submitting AI work. Students stand on an allowed, ask-first, or not-allowed line and defend their position using the assignment’s learning goal.

Debrief: Did the class disagree about the tool, the task, or the teacher’s rule?

Teams take an assignment that AI could complete too easily and redesign it to reward drafts, local evidence, personal reasoning, oral explanation, or reflection. They keep the original academic standard intact.

Debrief: How can an assignment welcome AI while still measuring learning?

Students imagine every AI prompt goes into a backpack they cannot empty. They sort sample details—first name, full schedule, medical information, draft poem, password, and school name—into safe, ask-first, and never-share pockets.

Debrief: Which harmless-looking details become risky when combined?

Display a fictional screenshot containing tabs, notifications, names, location clues, and another student’s work. Teams circle everything that should be removed or protected before any upload.

Debrief: What did you notice on the second look that you missed on the first?

Groups build a decision tree for receiving a shocking voice clip or video: pause, avoid sharing, inspect the source, compare reliable reporting, save evidence if harm is occurring, and tell a trusted adult.

Debrief: Why is slowing down a safety skill?

Students examine scenarios involving a friend’s photo, a teacher’s voice, team chat messages, and collaborative artwork. They identify whose permission is needed before using the material with AI.

Debrief: Who can be affected even when they never used the AI tool?

Teams transform a scary or vague AI warning into a calm action message for students: name the risk, give two concrete steps, and identify an adult or school channel that can help.

Debrief: Which wording makes someone more likely to ask for help?

Students review an AI-generated list of influential people, community needs, or historical milestones. They identify missing voices and write a follow-up prompt that asks for broader representation without assuming a single correct list.

Debrief: How does omission shape the story an answer tells?

Pairs inspect a generated story, image description, career list, or event plan for assumptions about names, families, abilities, money, culture, or location. They replace assumptions with purposeful choices.

Debrief: When can a ‘normal’ default exclude real people?

Teams ask how one proposed school or community change might affect four different groups. They verify the AI-suggested viewpoints with course evidence or human input rather than treating generated perspectives as testimony.

Debrief: Why can AI suggest a perspective but not speak for a community?

Students highlight words in two AI summaries that create sympathy, blame, certainty, or fear. They rewrite both in neutral language, then discuss what important meaning may be lost by making everything neutral.

Debrief: How can word choice influence a reader before evidence does?

Groups test an imaginary AI system used for club selection, tutoring recommendations, or event scheduling. They name who could benefit, who could be overlooked, what data is missing, and where a person must review the decision.

Debrief: What makes a process efficient but not necessarily fair?

Students brainstorm five ideas alone, request five deliberately different possibilities from AI, and then create a final sixth idea that combines or rejects elements from both lists. They label the origin of each ingredient.

Debrief: Did AI expand your options or pull you toward familiar ideas?

Students identify three features of their own writing or speaking voice, compare them with a generic AI paragraph, and revise the paragraph without copying personal work into a tool. The goal is intentional voice, not imitation.

Debrief: What makes a piece sound genuinely like you?

Students critique an AI-generated plan, explanation, or design using a teacher-provided rubric. They must reject at least one major choice, keep one useful element, and build a stronger version in their own words.

Debrief: Why is disagreeing with AI an advanced use of AI?

For a poem, presentation, design, or solution, students add one ingredient AI cannot supply on its own: a firsthand observation, local detail, measured result, personal value, or conversation with a real person.

Debrief: How did the human-only ingredient change the final work?

Students review a hypothetical month of AI use and identify one skill it strengthened, one skill it may have weakened, one privacy choice, and one integrity choice. They write a two-week adjustment plan with a measurable goal.

Debrief: What would responsible progress look like, not just responsible rules?