The Machine Beneath the Mission

The origin story most educators know goes something like this: Horace Mann, appalled by inequality in 1830s Massachusetts, championed common schools to forge democratic citizens. Education would be the "great balance wheel of society" — equalizing conditions, lifting the poor, forging shared civic values. This narrative is not false. But it is dangerously incomplete.

Mann modeled his vision on the Prussian education system, which he praised in his 1843 Seventh Annual Report as possessing the most distinguished reputation for school excellence. What Mann admired had been built after Prussia's humiliating defeat by Napoleon in 1806 — a system explicitly designed to produce disciplined, obedient subjects who would follow orders as a unit.

In America, after Shays' Rebellion and the Whiskey Rebellion, Thomas Jefferson wrote to James Madison that schools should teach children that violence is an illegitimate way for citizens to express discontent. The common school was, from its inception, a technology of social control dressed in democratic language.

The architects were sometimes startlingly candid. Edward A. Ross, whose Social Control (1901) provided the theoretical framework for education-as-management, described the goal as collecting "little plastic lumps of human dough from private households and shaping them on the social kneading-board." Ellwood Cubberley, Dean of Stanford's School of Education, wrote in 1916: "Our schools are, in a sense, factories in which the raw products — children — are to be shaped and fashioned into products." Cubberley collaborated with Lewis Terman, creator of the Stanford-Binet IQ test and an avowed eugenicist, whose testing instruments became the primary mechanism for sorting children into academic tracks. These were not fringe figures. They built the institutional architecture American schools still inhabit.

A finding that should unsettle any narrative of progressive improvement: the Committee of Ten's 1893 report — often portrayed as elitist — actually insisted on rigorous academic education for all students regardless of post-graduation plans. It was the "progressive" Cardinal Principles of Secondary Education in 1918 that introduced differentiated tracking, using the democratic-sounding language of "meeting students' needs" to sort children by class and perceived ability. By 1920, most big-city high schools offered four tracks: college preparatory, commercial, vocational, and general. The language of equity was used to build a sorting machine.

When Sean Reardon's team at the Stanford Center for Education Policy Analysis built the Stanford Education Data Archive — analyzing over 300 million standardized test scores from virtually every public school district in America — they did not find a meritocracy. They found that three-quarters of achievement gap variation can be explained by two factors: local racial and ethnic differences in parental income, and patterns of residential segregation.

AI and machine learning models trained to predict educational success consistently identify the same dominant variables. A 2025 study using Random Forest and Gradient Boosting models on 2,549 first-graders identified the top predictors from 142 candidate variables: teacher perceptions of attention, early literacy skills, and parental education level — which outperformed all other socioeconomic indicators.

An NLP analysis of 60,000 college application essays found that essay content predicted household income with measurable accuracy. This means class markers are embedded even in the parts of applications perceived as qualitative counterweights to testing.

The 2024 NAEP results confirmed a pattern that should alarm anyone who believes education is the "great equalizer": the lowest-performing students — overwhelmingly poor, disproportionately Black and Latino — declined the most after the pandemic, while students near the top changed little. The system is not merely unequal. It is divergent, pulling the top and bottom further apart over time.

America funds its schools primarily through local property taxes — a structural choice dating to 1647 that no other industrialized nation replicates. The Supreme Court affirmed this in San Antonio Independent School District v. Rodriguez (1973), ruling 5–4 that education is not a fundamental right under the U.S. Constitution and that wealth is not a protected class. The consequences are staggering and specific.

A 2025 systematic review of 322 comparisons confirmed that while school-level funding disparities have slightly narrowed since 2005, student-level inequities have persisted or worsened. The gap in teacher experience and effectiveness between high-poverty and low-poverty schools has actually widened.

Richard Reeves at Brookings describes the mechanism as "opportunity hoarding" — the top quintile using exclusionary zoning, legacy admissions, and resource concentration to protect their children's advantage while framing the results as meritocratic. Seventy percent of students at elite colleges come from the top fifth of the income distribution.

When AI examines this funding structure, it identifies a revealed preference. Whatever America says about equal opportunity, the structure says something different: a child's educational resources should be determined by their parents' real estate, not by their needs. The system treats education not as a universal right but as a market commodity distributed through property markets.

If the system is a sorting machine, what does it sort for? Paulo Freire described what he called the "banking model" of education — the teacher deposits information into passive student-containers — and argued that it "anesthetizes and inhibits creative power." Samuel Bowles and Herbert Gintis, in Schooling in Capitalist America (1976), provided the empirical foundation: studying 237 New York high school seniors, they found that grades correlated more strongly with personality traits than with academic ability. Low grades were associated with creativity, aggressiveness, and independence. High grades were associated with perseverance, consistency, and punctuality.

A 2023 meta-analysis by Terrin and Triventi in the Review of Educational Research, analyzing 53 studies from 2000 to 2021, found that academic tracking has no statistically significant effect on overall educational efficiency but a significant positive effect on inequality. Tracking does not help anyone learn more. It reliably ensures that those who start behind stay behind.

Jeannie Oakes documented in Keeping Track that lower-track classes emphasize compliance and rote skills while upper-track classes emphasize critical thinking — and that race and class, not ability, predominantly determine placement. A 2024 Brookings analysis confirmed that grouping students by ability, however done, will inevitably separate students by race, ethnicity, native language, and class.

What AI Inherits From the Data It Trains On

When researchers at the University of Texas built AI models to predict college completion, they found the models incorrectly predicted failure for Black students 19% of the time and Hispanic students 21% of the time — compared to 12% for white students. The algorithms did not create this bias. They inherited it from the data the system produced.

Baker and Hawn's 2022 review of algorithmic bias in education found these systems function as "sorting mechanisms that can either target or exclude" students, remaining "both obfuscated and legitimized through perceptions that statistical models are objective."

The most provocative evidence surfaces when we ask not what the system does to students but what it reveals about how America views children. Georgetown Law's Center on Poverty and Inequality found that adults perceive Black girls as young as five as less innocent, more adult-like, needing less nurturing, less protection, and less comfort than white girls of the same age. This has a name — adultification bias — and it translates directly into institutional behavior.

Walter Gilliam's 2016 Yale Child Study Center experiment used eye-tracking technology on 135 preschool teachers watching videos of four children playing normally. Teachers were told to watch for "challenging behavior." There was none. Yet eye-tracking revealed teachers spent significantly more time surveilling Black children, particularly Black boys — who received the most sustained attention of any child in the video. When teachers were given information about a child's difficult home circumstances, empathy increased only when teacher and child shared the same race. When they differed racially, severity ratings escalated. Knowledge of adversity, combined with racial difference, triggered not compassion — but a sense of hopelessness and intractability.

The sorting operates in both directions simultaneously. Black students comprise 9% of gifted program enrollment despite representing 15% of the student population. A Black student with identical test scores to a white student is only half as likely to be placed in a gifted program. The single strongest mediating factor is teacher race: Black students with Black teachers are identified as gifted at rates approaching those of white students; with white teachers, identification falls by two-thirds.

Meanwhile, Black students are twice as likely to be labeled with "emotional disturbance" and 1.5 times as likely to be classified with "intellectual disability" — the most subjective and stigmatizing categories in special education. The system simultaneously excludes Black children from enrichment and includes them in stigmatizing categories, producing a dual sorting mechanism that channels children into "promising" and "problematic" bins based substantially on race.

John Dewey envisioned schools as "miniature communities" modeling democratic participation — where education meant growth itself, not preparation for some future purpose but the cultivation of intelligence, curiosity, and cooperative life in the present. In a 1915 debate with David Snedden, Massachusetts Commissioner of Education, Dewey argued that vocational education should not "adapt" workers to the existing economy but teach children how the world works so they could reshape it. Snedden's vision won. The Smith-Hughes Act of 1917 and the Cardinal Principles of 1918 institutionalized tracking. A century later, AI analysis of what schools actually reward confirms the correspondence principle: the social relations of schooling mirror the social relations of production, and the system's outputs reflect its inputs with uncomfortable fidelity.

Herbert Kohl offered a concept that deserves more attention than it receives: creative maladjustment — the idea that sometimes a student's refusal to learn is an intelligent response to a toxic or oppressive educational environment. Children who appear to "fail" may be making a deliberate, if unconscious, choice not to assimilate into a system that devalues them. This reframing turns the deficit narrative inside out. What if some portion of the "achievement gap" is not a failure of children but a refusal — a form of resistance that the system is structurally incapable of recognizing as rational?

The convergence of historical evidence, contemporary data, and computational analysis points to an uncomfortable synthesis. American public education simultaneously operates on multiple purposes, but the dominant ones — as revealed by what the system's own data encodes — are sorting and stratification, compliance enforcement, and credentialing as positional competition. It was not designed primarily for liberation, critical thinking, or the cultivation of human potential. It was designed to manage populations, allocate social position, and reproduce existing hierarchies while speaking the language of democracy and opportunity. That conclusion will generate objections. It should. What follows is a direct engagement with the most substantive of them.

The arguments presented here are grounded in peer-reviewed research, institutional data, and a century of documented educational scholarship, but they are not without complexity. What follows is a direct engagement with the most substantive critiques this paper is likely to receive.

None of these objections, taken individually or collectively, refute the core finding: that the patterns produced by the American public education system — its funding disparities, its tracking outcomes, its discipline data, its gifted enrollment demographics — are not accidental. They are structural. Engaging that finding honestly, professionally, and with the students' interests at the center is not a political act. It is the work.

This does not mean your classroom is a factory. It does not mean your relationships with students are fraudulent. It means you are doing deeply human work inside a machine that was not built for the purposes you bring to it. The distance between your intentions and the system's architecture is the space where the most important questions live — and that space is exactly where ForwardEd works.

Dewey, Freire, Ladson-Billings, and Kohl were not naïve about this distance. They named it, studied it, and insisted that educators must reckon with it — not to despair, but to act with clearer sight. That is the professional obligation this paper is asking you to accept. Not guilt. Not paralysis. Clarity.

Clarity requires us to have difficult conversations. It requires us to look at data that is uncomfortable, sit with conclusions that challenge what we want to believe about the systems we have dedicated our careers to, and resist the impulse to protect the institution at the expense of the child. These conversations are not radical. They are professional. They are precisely the conversations that well-trained, equity-minded educators have been having in staff rooms, graduate seminars, and coaching sessions for decades — often quietly, often at personal risk. This paper exists to bring those conversations into the open, to give them research grounding, and to insist that they belong in every school, at every level of leadership.

That is also why this paper was written in collaboration with AI — and why that collaboration is disclosed fully and without apology. ForwardEd's commitment to transparency is not a formality. It is a statement of professional values. We used AI as a research and synthesis tool because it allowed us to move further, faster, and with greater breadth than any single author working alone. We verified every source independently. We made every editorial judgment ourselves. And we named all of it — because modeling transparent, ethical AI use is itself part of the work.

If we are asking educators to interrogate the systems they work within, we must be willing to interrogate our own tools. AI is not neutral. It carries the same biases this paper documents. Used carelessly, it amplifies them. Used thoughtfully — with human judgment, clear purpose, and full transparency — it can extend our capacity to serve students better. That is the standard ForwardEd holds itself to, and it is the standard we are asking the profession to adopt.

Now we arrive at the harder conversation. The one that does not get had at faculty meetings because it is too uncomfortable. The one that gets deferred to the next strategic planning cycle, or buried in a committee, or quietly abandoned because the person with the power to act has decided that acting would cost them something personally. The research is unambiguous on this point: educational change fails not primarily because teachers resist it, but because leadership will not sustain it. Michael Fullan, whose work on educational change spans five decades and every continent, is direct: the single greatest obstacle to meaningful school improvement is not lack of knowledge about what works — it is the unwillingness of those in positions of authority to use that authority in service of students rather than in service of their own institutional survival.

Richard Elmore of Harvard, whose framework of internal accountability remains one of the most cited in educational leadership research, draws a clear line between leaders who use their authority to protect the conditions for learning and those who use it to protect their own position. He calls the latter "pragmatic compliance" — the performance of leadership without its substance, calibrated to satisfy external demands while leaving the instructional core untouched. Hargreaves and Fullan name it more bluntly in Professional Capital: the difference between a leader who has moral purpose and one who has political purpose. They are not the same thing. And students pay the price when we confuse them.

This paper is not accusing anyone. It is naming a structural reality that every honest educator already knows: there are people in this profession who are willing to lead the change — who have studied the data, developed the frameworks, built the relationships, and are ready to do the work. And there are people with more institutional power who are stopping them. Not because the evidence is weak. Not because the approach is unsound. But because change is threatening to people who have built their identity, their reputation, or their career around the way things currently are. That is an ego problem. And students should not be the ones paying for it.

The obligation of leadership — real leadership, not its performance — is to get out of the way of the people who are doing the work, and to use whatever authority you hold to protect them while they do it. Elmore put it plainly: if you require something of someone, you are obligated to build their capacity to do it. If you ask teachers to reach every child and then block every structural support that would make that possible, you are not leading. You are performing leadership for an audience of adults while the students in front of you wait. Darling-Hammond's decades of research on teacher quality and student outcomes arrives at the same conclusion from a different direction: the single most powerful thing a leader can do for students is to build and protect the conditions in which excellent teaching can happen. Not evaluate teaching into existence. Not mandate it. Build the conditions for it — and then defend those conditions against every political convenience and ego-driven obstruction that comes for them.

Yes, this is a broken system. Yes, we work within real constraints, limited budgets, political pressures, community skepticism, and a structure that was not built for the goals we are trying to achieve. None of that excuses inaction. None of it justifies using institutional authority to suppress the educators who are willing to lead. Fullan's research on change leadership is clear that schools that improve sustainably do so because their leaders create coherent conditions for innovation and then protect those conditions with moral courage — not because they waited for permission from a system that was never designed to give it. If you have educators in your building or district who are ready, trained, willing, and positioned to move students forward, your job is simple: remove the obstacles and stand behind them. Every day you spend protecting your comfort at their expense is a day a student does not get what they need.

The students in our classrooms right now — the ones in the lower track, the ones who have already been suspended before they learned to read, the ones whose potential has been filtered through a system that was never designed to see it — do not have time for institutional timidity or professional self-protection. They need adults who are willing to act on what they know. That is not a political statement. It is the most basic definition of what it means to be an educator.