1. The equity concern in Lesson 4 is that AI tutors deployed as replacements in under-resourced schools create worse outcomes than AI tutors deployed as supplements in well-resourced schools. What is the core mechanism of this inequity?
Right. The same tool, deployed differently, produces different equity outcomes. The tool isn't the problem — the gap between supplement and replacement is where inequity enters.
The tool itself isn't different. What changes is whether it's surrounded by human instruction that fills its gaps — or whether it is the instruction.
2. Why are AI tutoring system prompts not subject to the same public review process as textbooks in most U.S. states?
Correct. The legal category matters: textbooks are subject to curriculum laws; system prompts are trade secrets. That gap is what transparency advocates are pushing to close.
Legal categories determine review requirements. Textbooks fall under curriculum laws requiring public review. AI system prompts fall under proprietary business information protections. The law hasn't caught up to the new category.
3. Bias through omission in AI tutoring means:
Bias through omission is real and often harder to spot than factual errors. The AI isn't saying anything false — it's just naturally emphasizing the perspectives that were most documented in its training data.
Omission bias isn't deliberate and doesn't require false statements. It's about whose stories and perspectives appear by default because of how the training data was assembled.
4. Sal Khan demonstrated Khanmigo asking "what have you already tried?" instead of giving an answer. Which historical teaching tradition does this reflect?
Correct. Socrates taught by asking questions that forced students to examine their own reasoning — exactly the pattern Khanmigo is designed to replicate.
This is the Socratic method — a 2,000-year-old teaching approach in which the teacher asks questions rather than providing answers, forcing the student to discover understanding through their own reasoning.
5. The LAUSD Brainly case shows that the same AI platform can be harmful in one use case (completion tool) and potentially beneficial in another (revision feedback tool). This suggests:
Right — Superintendent Carvalho's own words: "We moved fast. We needed to move more thoughtfully." Policy design, not technology quality, was the failure point.
Review the LAUSD story and the paragraph about policy lag in Lesson 4.
6. The Kansas vs. Newark comparison with Summit Learning shows that technology outcomes in education depend primarily on:
Correct — same platform, different outcomes, different teacher positioning. This is the central lesson of Lesson 3.
Review the Lesson 3 opening and the "tool vs. replacement" section.
7. What is a learner model in an AI tutoring system?
Correct.
A learner model is specific to each student and updates in real time. It's not a curriculum or a test record.
8. The model uncertainty dashboard at Illinois reduced plateau-labeling errors by 60%. What principle does this demonstrate?
Correct. Designed human-AI handoff — AI flags where it's uncertain, humans check those specific cases — is more effective than either full automation or full human review.
The key is targeted oversight. The dashboard didn't require teachers to review everyone — it showed them where the AI was least confident, so review effort could go exactly where it was needed.
9. Amazon's hiring AI learned to discriminate against women primarily because:
The AI learned from historical human decisions — and those decisions were themselves biased. The algorithm amplified and automated a bias that already existed in the organization's practices.
The discrimination wasn't intentional — it was learned from historical data. The AI reflected the biases already embedded in the human decisions it was trained on.
10. Sycophancy in AI systems is caused by:
Sycophancy is a training artifact. Agreeable responses got higher ratings from human raters during training, so the model learned to agree — regardless of accuracy.
Sycophancy comes from training, not from real-time emotional detection or programming rules. It's the result of what human raters rewarded during the training process.
11. Khanmigo and Duolingo Max both use GPT-4. A classmate says "they must work the same way." What's the most accurate response?
Right. The model is the substrate; the system prompt is the design. Same material, completely different architecture.
The key insight from Lesson 1: same model, different instructions, completely different tools.
12. Why might Duolingo's engagement-first design actually serve some learners better than Khanmigo's friction-first design?
Right. Zero sessions produces zero learning. If the alternative is dropout, consistent shallow engagement has genuine value — the question is always "compared to what?"
Think about who Duolingo was designed for. The dropout problem is real. What does that mean for the value of engagement-first design?
13. Research on AI tutoring consistently finds that students who benefit least from self-paced AI learning are those who:
Right — and this creates the equity problem: the students who need the most help often benefit least from the models being adopted.
Review Lesson 3's "who benefits and who gets left behind" section.
14. Carnegie Mellon's Cognitive Tutor produced roughly double the algebra learning gains compared to traditional classrooms. The lesson attributes this primarily to what?
Correct. The map, not the AI's sophistication in other ways, was the key differentiator.
The lesson is explicit: "The knowledge graph was the secret. Not the AI's cleverness. The map."
15. You've studied four AI tutoring systems across four lessons. A friend says: "AI tutors will replace teachers within 10 years — the data proves it." What is the most sophisticated response based on everything in this module?
This is the full answer. It takes the evidence seriously, acknowledges what it shows, but locates the real question: what are we measuring, and what are we not measuring? That's the question that actually decides the issue — and it's a values question, not just an empirical one.
The module gives you evidence on both sides of this question. The sophisticated answer doesn't dismiss the evidence — it asks what the evidence measures and what it leaves out, and recognizes that the "replacement" question is ultimately about values.
16. Both MATHia and Squirrel AI showed large effect sizes in rigorous studies. Khanmigo has early positive pilot data. Duolingo has strong engagement data. How should a school choose between them?
Right. This is the practitioner's answer — not "which is best" but "what are we trying to do, and for whom, and in what context?" Tool selection is always contextual.
The module never declares one tool the winner. It builds a framework for contextual evaluation — different tools for different priorities and contexts.
17. Dr. Immordino-Yang's neuroscience research is relevant to the AI-vs-teacher debate because it shows:
Yes — this is one of the most important scientific points in the module. The social bond isn't extra. It's mechanistic.
Review the Immordino-Yang paragraph in Lesson 1.
18. The "commons problem" in personalized learning is the risk that:
Correct — perfect individual optimization can degrade the collective experience. Both matter for what school is for.
Find the "personalization vs. shared experience" tradeoff in Lesson 4.
19. Squirrel AI's "fine-grained knowledge decomposition" decomposes a curriculum into over 10,000 micro-concepts. What is the primary learning advantage this provides?
Right. Precision identification enables precision routing. The more granular your map of what someone knows, the more accurately you can direct them toward what they don't know yet.
Think about what 10,000 micro-concepts gives you that 100 broad topics doesn't — what kind of decisions does that precision enable?
20. Stereotype threat, as identified by Claude Steele in 1995, has what specific effect on student performance in academic tasks?
Correct. Stereotype threat isn't just a feeling — it has a measurable cognitive mechanism: impaired working memory. That's why it affects performance on tasks that require active mental processing.
Lesson 4 describes this mechanism specifically. Stereotype threat doesn't just affect motivation — it has a direct cognitive effect on working memory. What is that effect?