Stage 3: AI literacy for ages 11 to 13

1. Intelligence in Context

I can read an AI capability claim as evidence about one task and design a team where people stay accountable.

What Counts as Intelligence? · Design the Right Human-AI Team

2. Evidence and Uncertainty

I can separate observation, inference and assumption, state honest confidence, and check AI claims against sources.

From Observation to Claim · The Uncertainty Investigation

3. Research and Source Evaluation

I can write answerable, ethical research questions and judge sources by reading laterally and corroborating.

Questions Worth Investigating · Build a Defensible Answer

4. Problem Framing and Stakeholders

I can map stakeholders, turn needs into testable requirements, and document trade-offs.

Whose Problem Are We Solving? · Requirements, Constraints and Trade-offs

5. Systems Thinking and Consequences

I can map a system with its feedback loops, predict side effects, and redesign with monitoring.

Map the Whole System · Predict Consequences and Redesign

6. Data Purpose and Governance

I can test each data field against its purpose and write a data plan with consent, access and deletion rules.

Why Are We Collecting This? · Design a Responsible Data Plan

7. Sampling and Data Quality

I can find sampling, measurement and missing-data problems and repair a biased sample, checking results by group.

Is This Dataset Good Enough? · Repair the Biased Sample

8. Features, Labels and Representation

I can choose features and labels for a real mini model and explain how they shape what it learns.

How Data Becomes a Model’s View · Engineer Better Features and Labels

9. Training, Testing and Generalization

I can train, validate and test a real mini model on separate data and judge how it handles new cases.

Train, Validate, Test · Can the Model Handle New Cases?

10. Metrics and Model Documentation

I can read a confusion matrix, choose the metric that fits the harm, and write a model card.

Accuracy Is Not the Whole Story · Create and Defend a Model Card

11. Algorithm Design and Efficiency

I can compare algorithms by counting steps and worst cases and choose one by its trade-offs.

One Problem, Several Algorithms · Measure, Compare and Improve

12. Programming with Abstraction

I can use conditions, loops and functions with parameters, and debug a program with test cases.

Conditions, Loops and Functions · Build and Debug a Modular Program

13. Rules, Search and Machine Learning

I can choose between rules, search, machine learning and no AI, and design a hybrid system with a fallback.

Which AI Approach Fits the Task? · Build a Hybrid Decision System

14. Language Models and Generative Systems

I can explain next-token prediction and show, with a real tiny model, why fluent text can be false.

How Machines Work With Language · Investigate a Small Language Model

15. Enterprise AI Architecture and Oversight

I can map the parts around a model and design human review, logging and escalation.

Inside an Enterprise AI System · Design a Safe Human-in-the-Loop Workflow

16. Prompting and Generative Evaluation

I can write a prompt as a specification and build a test suite to evaluate generated outputs.

Prompting Is Specification · Create an Evaluation Test Suite

17. Verification and Synthetic Media

I can verify generated claims and citations and trace the provenance of a media file.

Fluent Is Not the Same as True · Verify a Claim and a Synthetic Media Case

18. Privacy, Cybersecurity and Adversarial Thinking

I can threat-model an AI system and defend it against prompt injection and data leakage.

Threat Model the AI System · Defend Against Manipulation and Data Leakage

19. Fairness, Explainability and Governance

I can measure who benefits and who bears errors, explain a decision, and run an appeal review.

Who Benefits, Who Bears the Risk? · Run an AI Impact and Appeal Review

20. Responsible AI Capstone

I can build, test, document and defend a responsible AI idea with evidence.

Build, Test and Document · Independent Defence and Public Exhibition