Formation vs. Training

Definition

Training equips skills for defined tasks—what you do. Formation develops character alongside intellect—who you become. Learning Python becomes obsolete in five years; cultivating judgment remains applicable indefinitely. The distinction has a specific mechanism: a trained engineer implements a specification efficiently; a formed engineer looks at the same specification and asks whether it should exist—then has the judgment to argue for a better one. As machines handle execution with superhuman competence, humans must handle intent. Formation is the educational paradigm for cultivating that intent, organized around four capacities the book calls the Four Pillars: systems thinking, intent design, taste formation, and AI fluency.

The principle, stated plainly: the gap between training and formation is the gap between executing specifications and questioning whether those specifications should exist.

The problem it solves

We have built an economy that runs on judgment while systematically training its workers to avoid exercising it. The dominant model is still, structurally, the 1890 Prussian system—fifty-minute periods, age-based cohorts, recall-based tests—designed to produce a population that could follow instructions, absorb standardized information, and execute defined tasks in the industrial economy. It succeeded at that mission. The mission is now obsolete. MIT researchers studying more than 300 enterprise AI deployments found that roughly 95% of enterprise GenAI initiatives deliver no measurable P&L impact—organizations can deploy capable agents but cannot tell them what to pursue, because most educated adults were never taught to define a worthy objective.

The barrier is not technical. As writing migrates to AI, the danger Derek Thompson names is cognitive de-skilling—"a quiet apocalypse" in which deep human thinking becomes scarce precisely as capable tools become abundant. We are deploying more capable AI into a workforce measurably less capable of directing it, because we trained the directing capacity out of people. Formation is the response.

Anatomy

Figure: The gap between training and formation is the gap between executing specifications and questioning whether those specifications should exist—a depreciating asset versus an appreciating one.

The Four Pillars are not independent competencies—they are a system, with intent design as the keystone.

• Pillar One — Systems Thinking. The AI-Born world behaves like a murmuration of starlings: autonomous agents where second-order effects matter more than first-order intentions. Optimize response time, get curt replies that damage trust. Systems thinking traces how an intervention ripples outward—stocks, flows, feedback loops, and the leverage points where small shifts trigger disproportionate change (Donella Meadows). It is the cognitive architecture to see what you're operating inside.

• Pillar Two — Intent Design. The keystone. In October 2021, Facebook's own researchers showed its engagement algorithm was amplifying outrage—doing exactly what it was instructed to do, because no one specified that "engagement" should exclude emotional manipulation. That is the AI alignment challenge in its purest form, and it is a human judgment problem, not primarily a technical one. When specification is hard, measurement becomes the substitute, and systems optimize the proxy while the underlying value degrades. Intent design is the literacy of defining worthy objectives—and the literacy that makes democratic oversight of AI possible.

• Pillar Three — Taste Formation. AI collapses execution costs, flooding markets with competent mediocrity. When a generative system produces ten thousand ad variations overnight, the differentiating asset is the judgment to choose well. The crucial distinction: taste as pattern recognition (which AI performs well) versus taste as judgment under contested values (which requires knowing what you're optimizing for, and why, when values conflict). Formation cultivates the second. Aesthetic consciousness requires duration—AI compresses execution time, but it cannot compress the time required for taste to develop.

• Pillar Four — AI Fluency and Collaboration. Knowing when an output deserves skepticism; prompt craft; recognizing confident hallucination; collaborative creation without surrendering authorial judgment. The design trap to avoid: if AI fluency becomes mainly about using AI tools fluently, it reproduces cognitive de-skilling in the name of solving it. The test is not whether students produce better outputs with AI, but whether they understand why an output is good and can judge it without the tool. Generate first, evaluate second.

Remove intent design and the other three have no north star.

Figure: The Four Pillars as an integrated system—systems thinking, intent design, taste formation, and AI fluency—cultivating the capacities machines cannot replicate.

How it works in practice

Two graduates make the distinction concrete. David graduated in 2020 with a 3.9 GPA—valedictorian, the pride of a working-class family. Sixteen years of solving well-defined problems taught him to identify correct answers and optimize for measurable outcomes. In 2025, assigned to implement an AI customer service system, his manager asked him to "figure out what we should optimize for." He waited for clearer specification, then defaulted to what he could measure: response time. Three weeks later response time had dropped sharply and everything looked beautiful—until customers called the system "cold," "dismissive," brand trust fell, and churn accelerated. The system did exactly what he'd specified. His education taught him to execute instructions flawlessly, not to question whether they captured what mattered.

Sofia graduated from a formation-focused school in 2022 with a portfolio instead of a GPA. Joining a healthcare AI startup in 2026, she faced a diagnostic-uncertainty dilemma: flag only high-confidence findings, or surface tentative possibilities? She recognized it as intent design. She didn't have an answer, but she knew the process—interviewing oncologists, emergency physicians, patients with conflicting risk tolerances—and proposed making confidence thresholds explicit and adjustable by context. Don't hide the trade-off; make it governable. David optimized a system. Sofia questioned whether it should exist.

The employer evidence backs the shift. In an August 2025 survey of 1,030 executives and hiring managers, more than 9 in 10 rated human judgment, critical thinking, and ethical reasoning as important or very important—figures that held or rose as AI tools were deployed. The skills employers can't automate away are precisely the ones the Prussian model never cultivated.

How to apply it

1. Start with teacher formation, not curriculum. The primary barrier isn't curriculum or technology—it's teacher capacity, and teacher capacity is a funding-allocation choice. You cannot give what you don't have. The Three-Tier model: experiential immersion, collaborative design communities, mentor-apprentice cycles over a rolling five-year schedule.
2. Redirect existing spend. U.S. teachers average 24 hours of professional development annually; Singapore entitles 100; Finland requires a master's with 1,600+ hours of pedagogical preparation. The U.S. already spends above the OECD average per student—the money flows into administration and facilities, not teacher formation. Redirecting a fraction toward Singapore-level intensity is the single highest-leverage intervention, and it requires a budget choice, not new federal legislation.
3. Assess formation, not just performance. Replace "where does this student fall in the distribution?" with "what is this person becoming, and what do they need next?" Portfolios, process documentation, authentic performance tasks, metacognitive self-assessment.
4. Use AI to handle content so humans handle formation. Khan Academy's Khanmigo never gives the answer—it asks the next question, scaling the two-sigma tutoring effect. Alpha School's two-hour model lets AI own content acquisition so humans own character formation. Mind the equity caveat: premium AI-enabled formation available only to the already-advantaged widens the gap.

Failure modes

• The timing mismatch. Formation properly resourced takes a decade or more to scale (the Prussian system took 30 years). The displacement window is 1–5 years. These timelines don't overlap—which is why formation cannot substitute for the The Three-Pillar Bridge for already-displaced adults. Three tiers, three timelines: the Bridge stabilizes in months, community forms over years, schools transform the baseline over decades. Conflating them is the consistent error of both optimists and dismissers.
• Hollow AI fluency. Teaching students to offload difficult thinking to language models erodes the very formation it claims to support.
• The scale limit. Formation is necessary for those obtaining Human Cortex roles; it is insufficient to absorb the majority. AI-Born firms create roughly 5–10% of traditional employment levels. Formation develops the capacities the remaining roles need—it does not generate those roles.
• Equity inversion. Without active investment, formation cultivates agency for the already-advantaged and deepens structural unemployment for everyone else.

What it is not

It is not job-specific reskilling. Training for a known task is a depreciating asset; formation develops capacities that travel—judgment across contexts, systems thinking in novel domains, ethical reasoning for unforeseen dilemmas. It is not a labor-market intervention only: formation is civic infrastructure, the difference between citizens who can interrogate an algorithmic system and subjects who cannot. And it is not a humanistic luxury—it is, increasingly, the thing employers cannot find.

Relationship to other frameworks

Formation is the individual-capacity precondition for the Economy of Doing → Economy of Being: freed time without formed judgment produces drift, not flourishing. The Four Pillars cultivate exactly the capacities the Machine Core + Human Cortex reserves for humans—the Architect, Guardian, and Force Multiplier roles of the The New Triumvirate are formed, not trained. Formation operationalizes the radical-reskilling pillar of the The Three-Pillar Bridge, reframing reskilling as formation rather than skills training. And it pairs with steward-ownership at two levels of resolution: legal charters constrain the drift toward extraction, but formed humans are the actual prevention—a steward-owned enterprise staffed by people never formed in judgment is a governance structure without a functioning conscience.

Origin note

Original to this manuscript. The formation/training distinction and the Four Pillars (systems thinking, intent design, taste formation, AI fluency) as an integrated system with intent design as keystone are original framing. The framework draws on John Dewey, Amartya Sen, Donella Meadows, Maxine Greene, Byung-Chul Han, Oliver Burkeman, Salman Khan, and Derek Thompson—named in citation—and grounds its claims in real institutions (High Tech High, Quest to Learn, Finland's phenomenon-based learning, Alpha School, Khanmigo) used as evidence rather than claimed as the framework's invention.