13 The Reconstruction of Study
When what needs to be learned changes, how we learn changes too. This chapter covers curriculum debates, experiments in educational settings, and shifts in the path along which expertise grows. This very course is itself one case of that experiment.
13.1 What to Learn
Pre-draft. Check ownership and status in TOPICS.md at the repository root.
Brief. In an age when AI writes syntax and boilerplate for us, what should we teach and learn? This section examines the claim that the center of gravity in programming education is shifting from memorizing syntax toward the abilities to define problems, decompose them, and verify solutions, and it maps out which positions clash in actual curriculum debates. Its task is not to reach a conclusion in advance but to check the grounds for each position against interviews and primary sources. This section serves as the introduction to all of Part 5, and it keeps in mind that this course itself is one experimental answer to this debate.
Seed Questions
- What grounds does each side offer: the claim that “memorizing syntax no longer matters” versus the counterargument that “without basic syntax you cannot even verify”?
- Do curriculum efforts that explicitly try to teach problem definition, decomposition, and verification actually exist, and what form do they take?
- Should human capacities such as collaboration, building trust, reasoning about others’ perspectives, and dealing with uncertainty be taught as separate course content, or should they be cultivated through pedagogical methods such as team assignments and discussion?
- How does the shape of this debate differ between data science education and traditional computer science education?
- How much do educators actually agree on what should be taught in the age of AI, and where do opinions diverge sharply?
- What position does this course’s own curriculum design (seminar format, collaborative book writing) take within this debate?
Research Pointers
- Verify cases of curriculum reform at universities and bootcamps using primary documents (syllabi, department announcements).
- Starting from David Deming’s webinar The Value of Human Skills in an AI Economy (2026-07-28), verify the claim about the comparative advantage of social learning and that human capacities should be addressed through pedagogy, using the original paper and follow-up research.
- Interview programming education researchers and current professors and instructors to gather primary sources.
- Divide roles with §13.2 (Experiments in the Classroom): this section covers the debate over “what” to teach, while §13.2 covers actual experimental cases.
13.2 Experiments in Educational Settings
Not yet drafted. Check TOPICS.md at the repository root for owner and status.
Brief. In the AI era, educational settings have largely split into three camps: those that ban AI use, those that mandate it, and those that overhaul their evaluation methods entirely. This section finds concrete cases of each approach and compares the actual outcomes they produced. A special task of this section is to include this very course (seminar-style, collaborative book writing, full AI use permitted) as one of the experimental cases and self-document it. Interviews and primary data collection are the core methods.
Seed Questions
- What cases exist of educational institutions or courses that banned AI use, and was the ban actually observed in practice?
- What cases exist of mandating or fully permitting AI use, and how did evaluation methods change?
- What outcomes have been reported by cases that overhauled evaluation methods themselves (reviving oral exams, strengthening process-based assessment, etc.)?
- Do educational research studies or large-scale surveys exist that compare these three approaches?
- How should the design and progress of this course itself be documented as an experimental case, and what can be confirmed through student interviews?
Research Pointers
- Collect official AI policy announcements from domestic and international universities and high schools as primary sources.
- Prioritize searching empirical studies in education journals.
- Directly collect this course’s syllabus, operational records, and student interviews as primary sources.
13.3 The Ladder of Expertise: How Do Beginners Become Experts
Pre-draft. Check ownership and status in TOPICS.md at the repository root.
Brief. The basic premise of deliberate practice theory is that experts build their skills by repeating easy tasks. What happens to this ladder if AI does those easy tasks instead? This section summarizes the core claims of deliberate practice theory, and seeks out research and observations on how beginners actually learn in AI-assisted environments to argue where they conflict. It pairs with §9.4 (The Junior’s Paradox), but this section places more weight on learning theory. Directly interviewing developers with different levels of experience is an important method.
Seed Questions
- Are the conditions required by deliberate practice theory (immediate feedback, gradually increasing difficulty, repetition) met or undermined in AI-assisted learning environments?
- Is there research comparing the speed and quality of skill acquisition between beginners who learned using AI tools and those who learned without them?
- Even with the same AI tool, do long-term learning outcomes differ between being assisted from the start versus first solving problems on one’s own to establish a baseline and only then using AI for feedback?
- Is there evidence for the concern that “delegating easy tasks to AI leaves only verification ability intact while generative ability atrophies”?
- When experts look back on their own learning path, what do they say would have been different if AI had existed?
- What recent academic critiques or revisions of deliberate practice theory itself exist?
Research Pointers
- Check academic databases for the original research on deliberate practice and the critical literature that followed it
- Prioritize searching empirical research in educational psychology and computer science education on AI-assisted learning
- Track down experimental studies testing the claim David Deming presented in the The Value of Human Skills in an AI Economy webinar (2026-07-28), that one should “first go through productive struggle and only then use AI as a thinking partner”
- Interview developers with under 3 years of experience and developers with over 10 years of experience separately, and compare them as primary sources