AI & Digital Ethics Grade 6 What is Artificial Intelligence

Study Guide: What is Artificial Intelligence?
Grade 6 | AI & Digital Ethics

1. The Driving Question

"If my phone can predict what I’m about to type, or a robot vacuum avoids my dog’s toys, how does it ‘know’ what to do—when it doesn’t even have a brain? And if it’s not really ‘thinking,’ what is it doing?"

2. The Core Idea — Built, Not Listed

Imagine you’re teaching your little brother how to sort his LEGO bricks by color. You don’t just say, "Put the red ones here"—you show him: "This is red. This is blue. If it looks like this, put it in this pile." After a few rounds, he starts doing it on his own, even with new bricks he’s never seen before. That’s kind of how AI works.

AI isn’t a brain, but it’s a system that learns patterns from examples—just like your brother learned to sort LEGOs. You feed it tons of data (like thousands of photos of cats and dogs), and it finds rules to tell them apart. The more examples it gets, the better it gets at guessing. But here’s the catch: it’s not understanding cats or dogs. It’s just spotting patterns, like how your brother might mistake a red fire truck for a red LEGO brick if he’s only ever sorted by color.

Key Vocabulary:
- Algorithm
Definition: A step-by-step set of rules a computer follows to solve a problem or make a decision.
Example: The "recipe" your video game uses to decide how fast enemies respawn when you lose a life.
(Note: In high school, you’ll learn algorithms can be biased if the rules are unfair—like if a game always makes enemies harder for certain players.)

• Machine Learning
  Definition: A type of AI where computers learn from data without being explicitly programmed for every single task.
  Example: Netflix suggesting "Because you watched Stranger Things"—it noticed patterns in what you’ve watched before, not because a human typed in rules for every show.

• Data
  Definition: Information (like numbers, words, or images) that AI uses to learn patterns.
  Example: The list of every song you’ve ever "liked" on Spotify, which helps it guess what you’ll want to hear next.

• Neural Network
  Definition: A system inspired by how brains work, where tiny "nodes" (like artificial neurons) work together to find patterns in data.
  Example: The way your phone’s camera automatically blurs the background in a photo—it’s not "seeing" like you do, but layers of nodes are guessing what’s the main subject.

3. Assessment Translation

How this appears in class:
- Exit Ticket (Formative): "Explain in 2–3 sentences how a self-driving car ‘knows’ when to stop at a red light. Use the words algorithm and data in your answer." - Proficient: "The car’s algorithm uses data from cameras and sensors to recognize the shape and color of a red light. It follows rules (like ‘if red, stop’) that it learned from thousands of examples." - Developing: "The car stops because it sees the light." (Missing key terms and how the system learns.)

• State Standardized Test (Short Answer):
  "A social media app uses AI to recommend videos. Describe one way this AI could make a mistake, and explain why that mistake happens."
• Proficient: "It might recommend videos about a topic I don’t like because it only looks at what I’ve clicked before, not why. For example, if I clicked on a video about sharks once, it might keep showing me shark videos even if I only watched it for the music in the background."
• Distractors in Multiple Choice:
  • "The AI is broken." (Ignores how AI learns from data.)
  • "The AI is trying to trick you." (Anthropomorphizes AI—it doesn’t have intentions.)
  • "The AI only works if you have the latest phone." (Irrelevant to how AI functions.)

Model Proficient Response (Short Answer):
"AI in my phone’s keyboard predicts words by looking at patterns in what I’ve typed before. For example, if I often type ‘see you later’ after ‘hey,’ it starts suggesting ‘later’ after I type ‘see.’ But it’s not reading my mind—it’s just matching patterns from my past messages. If I suddenly start typing in a different language, it’ll get confused because it doesn’t have enough data for that."

4. Mistake Taxonomy

Mistake 1: Anthropomorphizing AI
- Prompt: "How does a voice assistant like Siri ‘understand’ what you’re saying?" - Common Wrong Answer: "Siri listens to your words and thinks about what they mean, then answers." - Why It Loses Credit: The question asks how AI works, but this answer treats Siri like a person. It skips the mechanism (algorithms, data) and uses vague language.
- Correct Approach: "Siri uses an algorithm to break your speech into sounds, matches those sounds to words it’s been trained on, and then follows rules to respond. It doesn’t ‘understand’—it just follows patterns from millions of examples."

Mistake 2: Confusing AI with "Magic"
- Prompt: "A robot vacuum avoids furniture. How does it ‘know’ where the couch is?" - Common Wrong Answer: "It has a special sensor that tells it where everything is." - Why It Loses Credit: The answer is technically true but too vague. It doesn’t explain how the sensor works with AI (e.g., machine learning to map the room).
- Correct Approach: "The vacuum uses sensors to detect obstacles, but it also uses machine learning to remember where furniture usually is. Over time, it builds a map of the room by learning patterns—like how you remember where your bed is even in the dark."

Mistake 3: Ignoring Data’s Role
- Prompt: "Why might an AI that recommends books sometimes suggest books you don’t like?" - Common Wrong Answer: "The AI is bad at its job." - Why It Loses Credit: The answer doesn’t connect to how AI learns (from data) or why that process can go wrong.
- Correct Approach: "The AI recommends books based on what other people with similar tastes have liked. If those people have different opinions, or if the AI doesn’t have enough data about your preferences, it might guess wrong. It’s not ‘bad’—it’s just limited by the data it has."

5. Connection Layer

• Within AI & Digital Ethics → Bias in AI:
  If AI learns from data, and data comes from humans, then AI can inherit human biases—like how a hiring AI might favor resumes with names it’s seen more often in past hires (which could mean it favors certain genders or ethnicities).

• Across Subjects → Math (Statistics):
  AI’s ‘learning’ is just advanced pattern-finding—like how you might predict the next number in a sequence (e.g., 2, 4, 6, ?). The difference is AI does this with millions of data points, not just three numbers.

• Outside School → Sports:
  NBA teams use AI to analyze player movements and predict injuries before they happen. The AI isn’t ‘watching’ the game like a coach—it’s spotting tiny patterns in how players run, jump, and land that humans might miss.

6. The Stretch Question

"If an AI is trained on data from the 1950s to predict who should get a bank loan, why might it unfairly reject applications from women or people of color—even if the AI itself isn’t ‘sexist’ or ‘racist’?"

Pointer Toward the Answer:
The AI doesn’t intend to discriminate, but if the 1950s data shows that loans were mostly given to white men, the AI learns the pattern: "People who look like this get loans." It doesn’t know why the data looks that way (e.g., historical discrimination), so it repeats the bias. This is why some AI systems now use fairness tools to check if their predictions are treating groups equally—but it’s still a huge challenge.

Tone Note: For Grade 6, we avoid jargon like "training sets" or "supervised learning" but don’t dumb down the ideas. The LEGO analogy keeps it concrete, while the stretch question invites debate about real-world consequences.