AI Investigators
Middle schoolers run mini ML pipelines, stress-test bias, analyze recommenders — and master the power unit AI as a Work Tool: specs before prompts, verification, privacy, and agency.
Unit 1
The Machine Learning Pipeline
From question to data to model to evaluation — and back again.
Essential question: What steps turn a real-world question into a tested model?
Questions AI Can (and Cannot) Answer
45 min · 1 quiz items · 0 labs
Distinguish prediction/classification tasks from questions needing human judgment alone
Train, Test, and Confusion
50 min · 1 quiz items · 1 labs
Use train/test split vocabulary correctly
Bias in the Pipeline
50 min · 1 quiz items · 1 labs
Locate at least three places bias can enter an ML pipeline
Unit 2
Recommenders & Attention
How feeds rank content and what that means for agency.
Essential question: Who shapes what you see next — and how do you stay in charge?
Unit 3
Build With Ethics
Create a small AI-assisted project with a model card and evaluation plan.
Essential question: How do we ship something useful without shipping harm?
Unit 4
AI as a Work Tool
Use generative AI for learning and projects without outsourcing your brain — specs, verification, privacy, and agency.
Essential question: How do I get real help from AI tools without becoming dependent, exposed, or wrong?
Spec Before Prompt
50 min · 3 quiz items · 1 labs
Write a task specification (goal, audience, constraints) before opening an AI tool
Verify or It Didn’t Happen
50 min · 3 quiz items · 2 labs
Apply a claim → evidence → confidence workflow to AI outputs
Privacy Is Part of the Skill
45 min · 3 quiz items · 1 labs
Classify common AI-use scenarios by privacy risk
Attention, Feeds & Your Agency
50 min · 2 quiz items · 1 labs
Explain how recommenders create feedback loops