AI UNIT 2 • FOUNDATIONS • FREE

How Machines Learn

A machine doesn't study, and it doesn't remember the way you do. It learns by being shown examples and copying the patterns it finds. In this unit you'll watch that happen, and see why whoever chooses the examples holds a lot of power.

🪶 5 Stages
🌱 Foundations · No coding
🎯 30-45 min per stage
💬 Guided AI Sandbox

The Big Idea

In Unit 1 you learned that an AI predicts patterns instead of understanding. This unit goes one level deeper: where do those patterns come from? The answer is examples. Show a machine thousands of examples of something, and it learns to spot and copy what they have in common. That pile of examples is called its training data.

Here's the part that matters most. A machine can only learn from the examples it's given. If the examples are rich and fair, it learns well. If they're thin, one-sided, or full of stereotypes, it learns those too, and repeats them with total confidence. The machine never questions its examples. That job belongs to people.

So the real question behind every AI isn't just "how smart is it?" It's "whose examples did it learn from?" For your nation and community, that question is everything, because so much of what's online was written about Native peoples by outsiders. In this unit you'll teach a machine yourself, in the sandbox, and feel exactly how much the examples decide.

💬 How this unit works. Same guided sandbox as Unit 1: an Ask the AI panel and a Field Notes panel, no coding. This time you'll give the AI examples and watch how it copies the patterns you feed it.

By the end of this unit, you'll be able to say "I can..."

  • Explain that a machine learns by finding patterns in examples, not by understanding
  • Show, with my own example, how the data you give shapes what the AI produces
  • Explain what "generalizing" means, and where it goes wrong
  • Connect whose data is included to what a machine can fairly represent
  • Explain why choosing the training examples is a decision that carries responsibility

What You'll Make

A set of Field Notes from your own experiment in teaching a machine.

You'll feed the AI examples and watch what it copies, then feed it thin or one-sided examples and watch its answers skew. You'll finish by writing about the question at the heart of it all: when the examples are about your nation, who should get to choose them?

By the end, "training data" won't be a buzzword. It'll be something you've felt with your own hands.

Your learning path

The 5 Stages

STAGE 1 Learning by Example Teach the AI a pattern and watch it copy STAGE 2 Garbage In, Garbage Out Thin examples, skewed answers STAGE 3 What's in the Pile? Training data at internet scale STAGE 4 Who Chooses the Data? The human decisions behind the machine STAGE 5 Your Field Notes Whose examples about your nation?
Let's Begin →

← Back to All AI Units

Sources

The way we describe machine learning follows the leading K-12 AI education frameworks, and the data-and-community framing follows Indigenous data governance scholarship. We prioritize the national bodies that write AI standards, Native-led organizations, and peer-reviewed research.

AI Literacy & Education

Indigenous Data & Knowledge

Digital Citizenship

Educational Standards

This unit aligns with the national AI education framework (AI4K12), Indigenous data governance principles (CARE), and national and state computer science, media literacy, and social studies standards. Open any section to see how the unit meets it.

  • Big Idea 3, Learning (Grades 6-12): Computers can learn from data. (This is the core of the unit. Students directly experience a machine learning a pattern from examples, generalizing it, and repeating the biases in the examples they choose.)
  • Big Idea 2, Representation & Reasoning (Grades 6-12): Agents maintain representations of the world and use them for reasoning. (Students see that the machine's "knowledge" is really a representation built from data, no more complete or fair than the data it was built from.)
  • Big Idea 5, Societal Impact (Grades 5-12): AI can impact society in both positive and negative ways. (Students connect data choices to real impact: what a machine can and cannot fairly say about their community.)
  • Collective Benefit (C): Data ecosystems should be designed to benefit Indigenous peoples. (Students ask who benefits when data about their community is used to train a machine, and what benefit would look like for the community itself.)
  • Authority to Control (A): Indigenous peoples' rights in their data must be recognized. (The unit's closing question is exactly this: when the training examples are about your nation, who has the authority to choose them?)
  • Responsibility (R): Those working with Indigenous data are responsible to the community. (Students learn that choosing training data is a decision with responsibility attached, not a neutral technical step.)
  • OSEU 6, Indigenous Ways of Knowing: Knowledge is passed through relationship and oral tradition across generations. (Students contrast careful, accountable knowledge-passing with a machine that copies patterns from whatever data it was handed, no matter the source.)
  • OSEU 2, Sovereignty: Tribal nations have the right to self-determination and self-representation. (Deciding whose examples represent a nation is framed as part of sovereignty, previewing the data-sovereignty units later in the track.)
  • OSEU 7, Learning & Identity: Language and understanding shape identity. (Students examine how the data behind a machine can shape, and distort, how their community is understood by others.)
  • CSTA 2-DA-08 (Grades 6-8): Collect data using computational tools and transform it to make it more useful. (Students work hands-on with examples as data, seeing how the shape of the data changes what a machine produces.)
  • CSTA 2-IC-21 (Grades 6-8): Discuss issues of bias and accessibility in the design of existing technologies. (Students trace bias in AI output back to biased or missing training data.)
  • ISTE 1.5, Computational Thinker: Students understand how automation works and use algorithmic thinking to develop solutions. (Students build an intuition for how "learning from data" actually works before ever writing code.)
  • ISTE 1.3, Knowledge Constructor: Students evaluate the accuracy and relevance of information. (Students judge machine output by questioning the data it came from.)
  • MN ELA, Evaluating Sources & Evidence (Grades 6-8): Assess whether reasoning is sound and evidence is relevant and sufficient. (Students treat a machine's answer as only as trustworthy as the evidence, the data, behind it.)
  • MN Media Arts 2.6.5.10.1 / 2.7.5.10.1 / 2.8.5.10.1 (Grades 6-8): Media artworks are influenced by cultural and historical contexts, including MN American Indian Tribes and communities. (Students analyze how the data behind AI media reflects, or erases, MN Tribal nations.)
  • Note for RF: verify exact MN codes against the current published standards before publishing.
  • ND CS 6.S.1 / 7.S.1 / 8.S.1 (Grades 6-8): Examine positive and negative impacts of technology, including equitable access. (Students weigh how the data behind AI affects how Native communities are represented.)
  • ND Indigenous Language Standard 2.1 (All Grades): Investigate and reflect on the relationship of practices to the perspectives of the cultures studied. (Students reflect on accountable community knowledge versus machine pattern-copying.)
  • Note for RF: verify exact ND CS codes against the current published frameworks before publishing.
  • OSEU Standard 6 (All Grades): Understand Oceti Sakowin ways of knowing, including oral tradition. (The unit contrasts relational knowledge-keeping with learning from a data pile.)
  • OSEU Standard 2 (All Grades): Understand sovereignty. (Choosing whose data represents a nation is framed as part of sovereignty.)
  • SD CS 6-8.IC.01 (Grades 6-8): Compare tradeoffs of computing technologies affecting urban, rural, and reservation communities. (Students consider how data gaps shape AI's usefulness and fairness for Native communities.)
  • Note for RF: verify exact SD CS codes and OSEU numbering against the current published frameworks before publishing.