Chapter 24 Exercises

Learning in the Age of AI: What's Still Worth Knowing When Machines Can Look It Up

These exercises are designed to move beyond recognition toward genuine understanding and application. Some of these exercises deliberately involve AI tools — but always in ways that require you to do the thinking. Others deliberately exclude AI. The distinction is the lesson.

Part A: Conceptual Understanding

These questions test whether you can define and explain the chapter's core concepts. Aim for your own words, not quoted definitions.

A1. Explain the knowledge paradox in your own words. Why does knowing more about a subject make AI more useful, and knowing less make it more dangerous?

A2. Define cognitive offloading and give three examples from your everyday life (not from the chapter). For each example, assess whether the offloading helps or hurts your long-term skill development.

A3. What is the difference between knowledge retrieval and knowledge construction? Give an example of each. Then explain why this distinction matters for evaluating whether AI is helping you learn.

A4. Define automation complacency. How does it relate to the metacognitive monitoring skills you developed in Chapter 13?

A5. The chapter introduces the concept of "metacognitive delegation." Explain what this means and why it's problematic for learning — even when the AI you're delegating to is giving correct answers.

A6. Explain the difference between AI as a cognitive tool and AI as a cognitive replacement. Why is the same technology capable of being both? What determines which role it plays?

A7. What is a "hallucination" in the context of AI? Why can't you rely on the AI's confidence level to tell you whether its output is accurate?

A8. The chapter claims that "prompt engineering is metacognition." Explain this claim by connecting at least two specific aspects of prompt engineering to metacognitive skills from Chapter 13.

Part B: Applied Analysis

These questions present scenarios and ask you to analyze them using the concepts from this chapter.

B1. Scenario: Aisha has a chemistry exam in three days. She asks an AI to generate a study guide covering all the topics from the past four weeks. The AI produces a detailed, well-organized guide. Aisha reads through it twice and feels prepared.

Using at least two concepts from this chapter, analyze what's likely to go wrong. What would a metacognitively informed approach look like instead?

B2. Scenario: Diego is writing a history essay about the causes of World War I. He asks an AI to write his thesis statement and introduction. Then he writes the body paragraphs himself, using the AI's framing.

Is Diego using AI as a tool or a replacement? Analyze both sides — what learning is he missing, and is there any learning value in what he's doing? How would you redesign this approach?

B3. Scenario: Priya is learning French. She uses an AI translation tool to check her homework. Whenever she doesn't know a word, she asks the AI to translate it rather than looking it up in a dictionary or trying to reason it out from context.

Analyze this using the generation effect and cognitive offloading. What's the difference between using AI to check her completed work versus using it to produce answers she hasn't attempted?

B4. Scenario: Raj asks an AI: "What's the theory of relativity?" and gets a clear, two-paragraph explanation. He reads it and moves on. His classmate Kai asks the same AI: "I think I understand special relativity — it's about how time passes differently for objects moving at different speeds. But I'm confused about why the speed of light is constant for all observers. Can you help me understand specifically why that's the starting assumption and not a derived conclusion?"

Compare the quality and learning value of these two interactions. Connect your analysis to the Explain-Before-You-Ask Protocol from the chapter.

B5. Scenario: A professor announces that students may use AI tools for all assignments in her course, with no restrictions. Some students celebrate. Others worry.

Using the concepts from this chapter, explain why unrestricted AI use in a learning context could be both a gift and a trap. What metacognitive safeguards would you recommend for students in this course?

B6. Scenario: Marcus (from the chapter) asks an AI to debug his Python code. The AI fixes the bug and explains what was wrong. Marcus reads the explanation, says "oh, that makes sense," and moves on.

A week later, Marcus encounters a similar bug in different code. He doesn't recognize it. Analyze what went wrong, using at least one concept from this chapter and one concept from a previous chapter (specify which chapter).

Part C: Real-World Application

These questions ask you to apply chapter concepts directly to your own life.

C1. Conduct an AI Use Audit on your past week of learning. List every interaction you had with an AI tool (if any). For each interaction, classify it on the AI Learning Ladder (Rungs 1-5). Then calculate your ratio of tool-use (Rungs 3-5) to replacement-use (Rungs 1-2). What patterns do you notice? If you don't currently use AI tools, describe how you imagine you would use them, and predict where on the ladder most of your interactions would fall.

C2. Choose one topic you're currently studying. Use the Explain-Before-You-Ask Protocol: write down what you currently know about the topic, identify where your understanding breaks down, and formulate a specific question. Then — and only then — ask an AI tool (or, if you don't have access, imagine what you would ask). Compare the quality of this targeted question with the kind of question you would have asked without the protocol.

C3. Try a generation-first experiment. Take a concept you need to learn for a current class or project. First, try to learn it entirely on your own — read the textbook, work through examples, take notes. Rate your understanding on a 1-10 scale. Then use AI to fill in gaps, clarify confusion, and test your understanding. Rate your understanding again. Finally, wait 48 hours and try to explain the concept from memory. Reflect on how AI contributed to (or detracted from) your final understanding.

C4. Identify your top three vulnerability points — situations where you're most tempted to use AI as a replacement rather than a tool. For each one, design a specific "trigger response" — a behavioral rule that kicks in when you notice the temptation. (Example: "When I'm tempted to ask AI to write my essay introduction, I'll first write three terrible introductions by hand. Then, if I still want AI input, I'll ask it to critique my best attempt rather than generate a new one.")

Part D: Synthesis and Critical Thinking

These questions require you to integrate multiple concepts, evaluate arguments, or think beyond what the chapter explicitly stated.

D1. The chapter argues that "AI is the world's best shallow processor." Do you agree? Can you think of ways in which AI might actually facilitate deep processing — not by doing the deep processing for you, but by creating conditions that encourage you to process more deeply? Make the strongest case you can, and then evaluate your own argument.

D2. Consider the equity implications of AI in education. If metacognitive skills are the key to using AI well for learning, and if metacognitive skills are unevenly distributed (as we've discussed throughout this book), what happens when AI tools become widespread in education? Does AI narrow or widen existing educational inequalities? Defend your position with reasoning from this chapter and at least one previous chapter.

D3. The extended mind thesis (Clark & Chalmers) suggests that cognition extends beyond the brain into tools. If this is true, is there a meaningful difference between "knowing" something in your head and "knowing" how to access it through AI? Argue both sides, then state your own position. Connect your argument to the distinction between knowledge retrieval and knowledge construction.

D4. The chapter identifies five things that remain "uniquely human" in the age of AI. Pick one and argue against the chapter: make the strongest case you can that AI might eventually develop this capability. Then evaluate your own argument — how convinced are you? What would change your mind?

D5. Imagine a student in the year 2035 reading this chapter. What might they find dated or naive about our current concerns regarding AI and learning? What advice might they add that we can't yet foresee? Then reflect: does the metacognitive framework the chapter proposes (tool vs. replacement, the knowledge paradox, the importance of monitoring) still hold even if the specific technology changes? Why or why not?

Part M: Mixed Practice and Spaced Review

These questions deliberately pull from previous chapters to promote interleaving and long-term retention.

M1. (Spaced Review — Chapter 12) Without looking back, explain the difference between deep processing and shallow processing. Give an example of each. Then connect this distinction to AI use: how does the depth of your processing determine whether an AI interaction is valuable or wasteful?

M2. (Spaced Review — Chapter 18) In Chapter 18, you explored how identity and mindset affect learning. How might a learner's identity shift if they become heavily dependent on AI? Contrast a "learner identity" with a "prompter identity." Which one supports long-term growth, and why?

M3. (Integration — Chapters 7, 10, 13) The chapter mentions retrieval practice (Ch 7), the generation effect (Ch 10), and metacognitive monitoring (Ch 13) as skills that interact with AI use. Choose one of these three and explain, in detail, how it either enhances or is undermined by AI use. Include a specific scenario.

M4. (Cross-chapter analysis) The chapter argues that the knowledge paradox means "you cannot outsource the foundation." How does this connect to the concept of schema formation (Chapter 5)? Why does having schemas make AI more useful, and how does AI use without schemas create problems?

M5. (Metacognitive check) Rate your confidence that you understand the main ideas of this chapter on a 1-10 scale. Then, without looking, list the six key concepts from this chapter. How many can you list? How many can you define accurately? Now compare your confidence rating with your actual recall performance. What does any gap tell you about your calibration (Chapter 15)?

Part E: Research and Extension (Optional)

These questions go beyond the chapter content for students who want to explore further.

E1. Search for empirical research on the effects of AI tutoring on student learning outcomes. Find at least one study (published in a peer-reviewed journal) that compares AI-assisted learning to traditional learning. Summarize the study's design, findings, and limitations. Does it support or complicate the claims in this chapter?

E2. The extended mind thesis (Clark & Chalmers, 1998) is referenced in this chapter. Locate and read the original paper (or a substantive summary). What were the original arguments? How do modern AI tools challenge or extend the thesis beyond what Clark and Chalmers originally envisioned?

E3. Research the concept of "automation complacency" in aviation safety or medical diagnosis. Find a documented case where over-reliance on automated systems led to errors. Write a brief (300-word) analysis connecting the aviation/medical case to the risks of AI-assisted learning discussed in this chapter.

E4. Design a hypothetical study to test the knowledge paradox. Specifically, design an experiment that would test whether students with more prior knowledge in a domain get more learning benefit from AI assistance than students with less prior knowledge. Specify: (a) research question, (b) participants, (c) independent and dependent variables, (d) how you'd control for confounding variables, (e) what you'd expect to find and why.

End of Chapter 24 Exercises. As you work through these, notice which exercises you're tempted to use AI for — and whether that temptation reflects a legitimate tool-use or a desire to skip the cognitive work. That awareness is the point.