The Sculpted Minds Inquiry

Introduction

How a learner engages with new information shapes whether the encounter remains fragile or leads to what cognitive scientists call a schema. In this account, a schema is a mental structure — new knowledge bound to old — organized for recall and use without support. The difference between fragility and schemas so formed is the difference between noting a historical event and highlighting its significance; between identifying parts of speech and deploying grammar to write complex sentences. This account selectively borrows from a ceramicist's tradition: the student's mind is both potter and clay, and a formed schema, like a kiln-fired piece, results from sustained work.

A student who has merely encountered new vocabulary can nod at the term on Tuesday and fail to summon it on Friday. Elsewhere a new proof merely scanned without consolidation can vanish from memory in a week. In both cases, the student has met the material without forming a schema for it; what is missing is the binding that would let that meeting hold. Fragility has further consequences. A student whose algebra never formed cannot reliably build calculus schemas. A student whose grammar was never worked into the clay is vulnerable to misconstructions when tasked to write long prose. These failures compound. What was never fired struggles to hold what follows.

Students who would once have spent hours on a lab report or a precis can now witness the finished piece manifest in seconds. By May 2025, 84% of high school students reported using generative AI for schoolwork (College Board, 2025). The artifact — a solved problem, a balanced redox equation, an admissions essay — arrives finished. The artifact more often came from a formed mind; with the tool it can arrive without. The tool is not in itself a problem, but in the realm of an unformed mind it produces a counterfeit. This inquiry is about the formation that lets a student direct such a tool without being consumed by it.

Section 1: The Sculpted Mind

Forming is not a download. Material does not flow from a teacher or a text into a waiting mind, leaving the student to receive something the mind did not help to make. The mind that ends up holding the material is the mind that did the work.

New material does not arrive ready for formation. The student prepares it through repeated contact that kneads the strangeness out. Material that has not been wedged cracks under the pressure of later work.

Rereading belongs to early contact: it builds familiarity and prepares the material for formation. Retrieval belongs to a later stage: it settles what has been prepared, returning the mind to the material and working the impression deeper. A student who rereads a page three times and calls it learning is mostly exercising recognition. The material may feel familiar, but familiarity without retrieval leaves the schema unformed (Roediger & Karpicke, 2006). Bjork (1994) identified the conditions that favor formation: spaced practice, mixed problem types, and the habit of producing answers from memory.

Scaffolding is temporary support that steadies the forming until the student can do it alone — worked examples, partial solutions, guided steps — and frees limited cognitive resources for the forming itself (Rosenshine, 2012; Sweller, 1988). Hmelo-Silver et al. (2007) argue that in problem-based and inquiry learning, effective scaffolding keeps students inside the difficulty rather than extracting them from it. Scaffolding, unlike a bypass, preserves the labor of forming rather than absolving the mind of it. At PROMYS, for instance, counselors redirect rather than answer students' questions (Boston University, 2024). A student who asks where to begin is sent back to prior knowledge and urged to conjecture, tinker, and look for patterns.

Firing is the moment when support ends and the student alone must draw on what has been formed — in a test, in an unfamiliar problem, in a claim to be argued without guidance. Sometimes the drawing holds. Sometimes the formed piece fractures at the edge of its reach. A student who has taken exponentiation to mean a multiplied by itself n times meets a^√2, where there is no n to count. A student who has learned grammar as rules meets a novelist whose fragments work; the rules cannot explain why. The schema has encountered a case it is not equipped for.

Such moments are occasions for kintsugi, the Japanese practice of repairing broken pottery with gold so that the fracture becomes visible and the repaired piece emerges stronger than the original, its history of difficulty worn openly in the seams. Working past a schema's limits produces an understanding the easy cases never demanded. The expert's studio is full of kintsugi.

Section 2: The Press Mold

Evading the labor of the studio is an ancient impulse. For generations students have traded the gravity of tomes for hollow summaries or coaxed others to produce a formed piece — an essay, a solved problem — while claiming it as their own. Generative AI, the latest instrument of mimicry, is also the first with the scale to reach billions of students at once. It reaches into a vast vault to answer each prompt with a roar of prose, producing a finished piece at the asking.

Bastani et al. (2025) found that students with unrestricted AI access scored higher on practice problems but 17% lower on tests than those who worked without it. Barcaui (2025) reported that students who studied with unrestricted AI access scored 11 percentage points lower on a retention test 45 days later. Those who used AI for research produced worse reasoning than those using conventional search methods (Stadler, Bannert & Sailer, 2024). A yearlong global study across fifty countries, drawing on interviews with over 500 students, parents, teachers, education leaders, and technologists, concluded that AI "turbocharges" cognitive offloading in schools (Burns et al., 2026).

The pattern is visible in individual classrooms. When a Fort Worth English teacher allowed her students to use AI to write a thesis statement for a literary analysis, those who did so could not explain the text they had supposedly analyzed (NPR, 2026b). A recent college graduate, reflecting on her own AI use, stopped after recognizing that the habit felt like "outsourcing my thinking" (NPR, 2026a).

The press mold, in yielding a finished piece without the striving that produces it, cleaves the mind from the work of formation.

Section 3: The Ceramics Teacher

Generative AI draws on an extraordinary repository of content knowledge. What it does not yet offer is the capacity that a tutor builds across many sessions with a student: a reading of that student's formation, a longitudinal eye distinguishing genuine fluency from performative confidence, and the trust that allows confusion to surface honestly (Bergin & Bergin, 2009; Bowlby, 1988).

This account holds that domain knowledge must be deep and coherent. Schemas form more readily when content is organized as a connected structure rather than a collection of isolated facts, each concept growing from the last and each procedure traceable to a definition. Hung-Hsi Wu, mathematician at UC Berkeley, has spent decades arguing that mathematics can be taught as a coherent story in which each concept grows from precise definitions and rigorous derivations (Wu, 2011). Classical traditions of language instruction treat grammar as foundational content, named and internalized before being pressed into use through composition and the reading of complex texts. Content arranged as a coherent story gives the mind something it can form around.

The forming conditions described in this inquiry — wedging, retrieval, scaffolding, firing, repair — are not unique to any single practice. A well-designed AI tool, with guardrails that preserve effortful engagement, can replicate some of these conditions. What it cannot replicate is the longitudinal reading of a single student's formation — the accumulated sense, built across sessions, of when the clay will hold and when it will not. The formed student holds what the press mold cannot give: a mind kiln-fired and ready for kintsugi.

References

Barcaui, A. (2025). ChatGPT as a cognitive crutch: Evidence from a randomized controlled trial on knowledge retention. Social Sciences & Humanities Open, 12, 102287. [link the URL: https://doi.org/10.1016/j.ssaho.2025.102287]

Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26), e2422633122. [link: https://doi.org/10.1073/pnas.2422633122]

Bergin, C., & Bergin, D. (2009). Attachment in the classroom. Educational Psychology Review, 21(2), 141–170. [link: https://doi.org/10.1007/s10648-009-9104-0]

Bjork, R. A. (1994). Memory and metamemory considerations in the training of human beings. In J. Metcalfe & A. P. Shimamura (Eds.), Metacognition: Knowing about knowing (pp. 185–205). MIT Press.

Boston University. (2024). PROMYS: A journey of questions. Boston University College of Arts & Sciences. [link: https://www.bu.edu/cas/promys-a-journey-of-questions/]

Bowlby, J. (1988). A secure base: Parent-child attachment and healthy human development. Basic Books.

Burns, M., Winthrop, R., Luther, N., Venetis, E., & Karim, R. (2026). A new direction for students in an AI world: Prosper, prepare, protect. Center for Universal Education, Brookings Institution. [link: https://www.brookings.edu/articles/a-new-direction-for-students-in-an-ai-world-prosper-prepare-protect/]

College Board. (2025). U.S. high school students' use of generative artificial intelligence [Research brief]. [link: https://research.collegeboard.org/media/pdf/AI%20Research%20Brief%201_vf_0.pdf]

Hmelo-Silver, C. E., Duncan, R. G., & Chinn, C. A. (2007). Scaffolding and achievement in problem-based and inquiry learning: A response to Kirschner, Sweller, and Clark (2006). Educational Psychologist, 42(2), 99–107. [link: https://doi.org/10.1080/00461520701263368]

NPR. (2026a, March 3). College students, professors are making their own AI rules. They don't always agree. [link: https://www.npr.org/2026/03/03/nx-s1-5716176/ai-college-students-professors]

NPR. (2026b, January 28). To keep AI out of her classroom, this high school English teacher went analog. [link: https://www.npr.org/2026/01/28/nx-s1-5631779/ai-schools-teachers-students]

Roediger, H. L., III, & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249–255. [link: https://doi.org/10.1111/j.1467-9280.2006.01693.x]

Rosenshine, B. (2012). Principles of instruction: Research-based strategies that all teachers should know. American Educator, 36(1), 12–19, 39.

Stadler, M., Bannert, M., & Sailer, M. (2024). Cognitive ease at a cost: LLMs reduce mental effort but compromise depth in student scientific inquiry. Computers in Human Behavior, 160, 108386. [link: https://doi.org/10.1016/j.chb.2024.108386]

Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257–285. [link: https://doi.org/10.1207/s15516709cog1202_4]

Wu, H.-H. (2011). Phoenix rising: Bringing the Common Core State Mathematics Standards to life. American Educator, 35(3), 3–13.