The Case of the Thinking Machine
AI Detectives
Investigate how intelligent systems make decisions, where bias appears, and how to ask better questions.
- Ages 10–15
- 8 Weeks · Teams of 3–4 Detectives
- 60–90 Minute Missions
The Big Idea
AI Detectives is an 8-week investigation into what AI is, the different kinds of AI, how to use it wisely, why it matters, and how to build it — powered by the Inspiration Vault, the real stories of innovators of color who built the intelligent world. Every E+ Kid learns how AI sees, predicts, recommends, creates, and decides, then builds a system that does it themselves. They are not users of the future. They are builders of it.
What learners build
Every detective trains a real no-code AI classifier — image, sound, or pose — in Google Teachable Machine, aimed at a school or community problem. Innovator-tier detectives extend the build into a beginner Python prediction model with train/test evidence.
The Showcase
The course closes with the AI Innovation Showcase: a live demo, a pitch defended with evidence, a Model Card, and a permanent deposit into the learner's own Innovation Vault.
The 8-Week Roadmap
#1 The Thinking Machine
Interrogate five mystery machines, define AI with evidence, and map where AI shows up in your own day.
AI Evidence Board + class definition of AI
#2 The Five Faces of AI
Meet the AI family — seeing, talking, predicting, moving, creating — and sort real systems into types.
Illustrated “Field Guide to AI Species”
#3 The Art of the Ask
Learn prompt anatomy, perform Prompt Surgery, and verify AI claims in Fact or Fabrication.
Before/after prompt + verification log
#4 The Case That Matters
Investigate AI changing medicine, farming, energy, and cities, then put a biased AI on trial.
Community AI impact map + trial verdict
#5 Teach the Machine
Build Sprint 1 — train a first classifier with Teachable Machine, learning features, labels, and training vs. test data.
Working classifier v1 + Dataset Evidence Card
#6 Break It to Make It
Build Sprint 2 — red-team the model, hunt for errors and blind spots, then retrain and improve.
Failure log + improved classifier v2
#7 AI for My Block
Aim the build at a real school or neighborhood problem, completing a privacy check, fairness check, and Model Card.
Community AI prototype + Model Card Lite
#8 The Final Investigation
AI Innovation Showcase — live demo, evidence defense, judge Q&A, and an Innovation Vault deposit.
Showcase pitch + Vault portfolio artifact
Three Tiers, one Showcase
Explorer
Ages 10–11 · I Discover
A guided first case: picture-based evidence boards, teacher-assisted builds, and a two-class image classifier.
Builder
Ages 12–13 · I Build
More independence: written input-process-output maps, a balanced 3+ class dataset, and a full Model Card Lite.
Innovator
Ages 14–15 · I Deploy
The deepest case: hybrid-system research, adversarial testing, and a parallel Python path training a real decision tree.
How learning works
See It
Learners examine a real example of the problem or technology before touching it themselves.
Simulate It
A guided, low-stakes walkthrough lets learners try the idea out before they build for real.
Build It
Learners create something real: code, a prototype, a design, a plan.
Explain It
Learners put what they built into words, defending their choices and their reasoning.
Own It
What they built goes into their Innovation Vault: their own intellectual property, kept and carried forward.
The Innovation Vault
Every mission ends with a Trailblazer Card from the Inspiration Vault — the true stories of 16 innovators of color, from Katherine Johnson to Rediet Abebe — matched to whatever a detective is stuck on. The final Showcase deposits each learner's model, dataset card, failure log, Model Card, and pitch into their own Innovation Vault: permanent intellectual property they carry forward.
The Inspiration Vault
A few of the Trailblazers detectives meet along the way — real innovators of color who built the intelligent world.
Katherine Johnson
NASA mathematician; John Glenn asked her to hand-verify the computer's orbit calculations before his 1962 flight.
Dorothy Vaughan
Taught herself and her team FORTRAN when electronic computers arrived at NASA, and became its first Black supervisor.
Gladys West
Her satellite-data models of Earth's exact shape became the mathematical foundation of GPS.
Fei-Fei Li
Created ImageNet, the labeled-image library that launched modern computer vision.
Safety, privacy & ethics
Every build stays in simulation: no real personal data, ever. Each detective signs the six-part Detective Code — protect people, investigate outputs, test beyond easy cases, credit AI help, never disguise a real person, and keep a human responsible for the final decision. Every build also completes a privacy check, a fairness check, and a human checkpoint before it ships.