My dad once said to me, with some sincerity, "You think about thinking. When I was your age, I didn't think about thinking." It was one of those moments: I remember where I was and what we were doing (I was in college and we were on a bridge watching a rowing regatta). He meant it as an observation more than anything, not necessarily a compliment, but I think he was genuinely interested that our minds seem to work differently. But that memory, or at least the version I have in my head, had stuck with me, in the way that just one of a million remarks by your parent can, because he had named something that in fact felt true. For much of my adult life, I have been intrigued by the different levels at which a person can engage with their own mind, and by how few people realize there's anything above the level they're at.
I've spent years developing a framework I call the Levels of Learning, which distinguishes between schooling, training, education, and self-directed learning. These aren't just different methods. They represent fundamentally different relationships between the learner and what's being learned, from passive reception to active ownership. That framework has given me a vocabulary for talking about what's really happening in education, beneath the policy arguments and institutional defenses.
I've wanted an equivalent framework for thinking itself for most of my adult life. I think I've found it, and it's no surprise that it aligns so well with my learning framework. The surprise is just how long it's taking me to articulate it.
Level 1: Coalitional Thinking — The Inherited Narrative. You think what your group thinks. Beliefs arrive socially, through family, culture, and community, not through investigation. You couldn't articulate why you believe what you believe because the question has never occurred to you. This isn't stupidity. It's the default human operating system, optimized over hundreds of thousands of years for coalitional safety. Most people throughout most of history have lived here, and for good reason; in stable environments where the group narrative is reasonably aligned with reality, it works.
Level 2: Informed Thinking — The Credentialed Narrative. You've added knowledge, credentials, and institutional fluency. You can cite sources, reference experts, and invoke "the science." You genuinely believe you've transcended Level 1 because you've replaced tribal intuition with institutional authority. But the epistemic structure is identical: deference to consensus, social punishment of dissent, inability to distinguish between "the evidence supports X" and "the institutions I trust say X." This level is the most dangerous precisely because it feels like the highest level to the person inside it. It provides exactly enough sophistication to make you confident you've arrived, and exactly not enough to see what you're missing.
Level 3: Critical Thinking — The Examined Narrative. You've internalized the insight that you yourself are subject to cognitive traps: confirmation bias, authority bias, coalitional pressure, and motivated reasoning. You can name the logical fallacies not as weapons against opponents but as descriptions of general human (and your own) tendencies. You understand why the founders built checks and balances, why the legal system presumes innocence, and why science requires falsifiability--not as historical trivia, but as evidence that smart people knew they couldn't trust their own judgment. You can hold a position while genuinely entertaining the possibility you're wrong.
Level 4: Structural Thinking — The Examined Self. You're not just watching for fallacies in arguments. You're asking why certain arguments dominate, who benefits from the consensus, what signals are being suppressed, and why. You can reweight an entire body of evidence based on a single verified falsehood, because you understand the structures (institutional, psychological, evolutionary) that produce coordinated distortion. You've turned the lens not just on your thinking but on the systems that shape what's thinkable. Plato's allegory of the Cave lives here, not as a metaphor for ignorance, but as a description of the structural relationship between social consensus and reality.
These levels are not stages you graduate from. You don't leave the lower levels behind. A Level 4 thinker still feels the coalitional pull, still flinches at social disapproval, still has the gut-level desire to align with the group narrative. The subconscious mind, the mind shaped by evolution for physical and social survival, doesn't go away. The difference is that you've built enough internal architecture to notice the coalitional pull and interrogate it rather than obey it.
This is also not a measure of intelligence. There are articulate people permanently at Level 2. There are modestly educated people who operate at Level 4 because life forced them to see through institutional narratives firsthand. The levels describe your relationship to your own cognition, whether you've ever turned the lens on the lens itself.
Level 2 is where most educated people live, and it's the most comfortable level to occupy. It satisfies the deep coalitional instinct (you belong, you're on the right side, you're with the smart people) while simultaneously providing the self-regard of believing you arrived there through reason. You get the warmth of group belonging and the satisfaction of feeling intellectually superior to those you see as less informed.
This is why Level 2 thinkers are often the most condescending. They look down at Level 1 thinkers as unsophisticated and at Level 4 thinkers as conspiracy theorists. From inside Level 2, the capacity to impute coordinated deception looks identical to paranoia, because the possibility that institutional consensus could be structurally distorted is simply outside the frame. It's not that they've considered it and rejected it. It's that it has never occurred to them as a serious possibility. The institutions they trust have told them it doesn't happen, and they trust the institutions.
There was a time when education took the project of moving people beyond Level 2 seriously. It was called a liberal arts education, which was not liberal in the political sense, but in the original Latin sense of liberalis: the education that distinguished a free person from a slave, because free people were expected to govern themselves. The trivium (grammar, logic, rhetoric) wasn't ornamental. It was the toolkit for thinking about thinking. Grammar taught you to parse claims precisely. Logic taught you to identify valid and invalid reasoning. Rhetoric taught you how persuasion works, so that you could recognize when it was being used on you.
The teaching of logic and logical fallacies was central to this tradition. Students learned to name the ways arguments could appear valid while being fundamentally deceptive: ad hominem, appeal to authority, false dichotomy, and straw man. These weren't abstract categories. They were the accumulated residue of generations of humans noticing, with painful precision, exactly how their own thinking went wrong.
We have largely abandoned this curriculum. What remains of "critical thinking" in education is often just Level 2 thinking with a more confident tone, the ability to cite better sources, and dismiss opposing views with more sophisticated vocabulary. Rarely does it include the genuine epistemic humility that defines Level 3, and almost never the structural awareness that defines Level 4.
The result is a population that is more credentialed than ever and less capable of independent thought than it has been in generations.
The lost curriculum is half the story. The other half is that we also dismantled the spaces where deep thinking could happen publicly.
There was a brief period, roughly 2005 to 2012, when the internet genuinely supported Level 3 and 4 discourse at scale. The tools of what was called Web 2.0 (blogs, wikis, threaded discussion forums, early social networks built around shared interests) were structurally hospitable to long-form, reflective conversation. You could develop an argument across paragraphs. Someone could respond to a specific point within it. A genuine exchange could unfold over days, visible to others who could learn from it. The format allowed depth, and depth attracted people who valued it.
I lived this firsthand. I ran one of the first social networks for educators (Classroom 2.0), with tens of thousands of members engaged in substantive threaded discussions about teaching, learning, and the purpose of education. I conducted over 400 long-form interviews with researchers, authors, and practitioners in a series called the Future of Education. The conversations were rich, searchable, and cumulative; they built on each other over time.
Then two things happened, neither of them malicious, both of them devastating.
First, Facebook and Twitter reshaped the economics of online attention. They replaced long-form, threaded discussions with short-form, non-easily searchable, algorithmically sorted content optimized for immediate emotional response. The shift didn't just shorten the format; it structurally selected for Level 1 and 2 engagement. Coalitional signaling. Performative agreement and disagreement. Content that tells you you're right and your opponents are wrong. The medium didn't change the conversation. It changed the level of thinking the conversation could sustain.
Second, the two most significant platforms for educational discourse, Ning and Wikispaces, were each purchased by companies that gutted them and, in both cases, removed all the free content educators had created. Years of accumulated discussion, resources, and collaborative work, all gone. This is a much larger cultural loss than anyone has acknowledged, because it wasn't just content that disappeared. It was the infrastructure for a particular kind of thinking.
No one set out to destroy deep public discourse. The equity transitions, the need to monetize, the logic of scale; none of it required anyone to intend the shallowing. It happened because depth doesn't scale and attention does. The commercial pressures were indifferent to what was lost.
Long-form writing still exists, of course; Medium, Substack, and the blogs that survive prove that. But substantive engagement with that writing has become vanishingly rare. A shallow reaction gets faster attention than a careful response. And once audiences reach a certain size, the conversation degrades into bickering over small nuances or defending against bad-faith misreadings, because the ratio of Level 2 readers to Level 3 and 4 readers makes genuine exchange nearly impossible at massive scales.
So we stopped teaching the tools for deep thinking and we dismantled the spaces where it could be practiced publicly. The loss of the curriculum removed the training pipeline. The platform shift removed the practice environment. Together, they explain why Level 2 is ascendant and why the silence around deeper work is not a failure of that work but a predictable consequence of the structures we've built and the ones we've lost.
What I'm describing isn't new. It's the rediscovery of an intellectual tradition that runs through Western civilization and that we've been forgetting.
The ancient Greeks gave us the formal study of logic and the cataloging of fallacies because they recognized that persuasion and truth are not the same thing. The legal tradition gave us the presumption of innocence, the adversarial system, the requirement for evidence beyond a reasonable doubt, and trial by jury--none of which are intuitive and all of which run against our natural tendency to assume guilt, defer to authority, and trust the accuser. They exist because enough people honestly looked at how justice failed and built institutional remedies to compensate.
The American founders did the same thing at the level of government. The separation of powers, the Bill of Rights, the elaborate system of checks and balances; these weren't expressions of optimism about human nature. They were expressions of deep skepticism. The founders had read enough history to know that power concentrates, that institutions corrupt, and that the people most likely to abuse authority are often the ones most confident they won't.
The scientific method belongs here, too. Peer review, replication, falsifiability; all of it exists because scientists recognized that even rigorous, well-intentioned researchers are subject to confirmation bias and motivated reasoning.
What unites all of these is a single insight: we cannot trust our own thinking without structures designed to catch its failures. That insight is the threshold between Level 2 and Level 3. The further insight, that the very institutions built to catch failure can themselves be captured, corrupted, and turned into instruments of coordinated distortion, is the threshold between Level 3 and Level 4.
I was recently reading about a Supreme Court case in which the lone dissenter was said to have described the defense of free speech as "puzzling." This same justice, the article asserted, had previously expressed concern that the First Amendment might "hamstring the government." In another hearing, she apparently argued that experts (doctors, economists, Ph.D.s) should be insulated from democratic oversight.
What struck me was not the positions themselves but the level of thinking they represented. This is a genuinely intelligent, well-credentialed person who (as represented) gives the appearance of having never asked the question that defines Level 3: Why did the founders want to hamstring the government? That question only arises if you've internalized the possibility that government power, like all concentrated power, will tend toward abuse regardless of the intentions of those who hold it. From inside Level 2, where institutions are assumed to be trustworthy and expert consensus is assumed to be reliable, constraints on government look irrational. From Level 3 or 4, they look essential.
The commentary I read about this justice framed her as a radical ideologue, which itself is only a Level 3 analysis; it sees through the claim to expertise and names the danger, but explains the behavior as bad intent. A Level 4 reading sees something more useful: she's not an anomaly, she's an archetype. She represents what happens when a genuinely intelligent person ascends through institutional structures that reward Level 2 thinking and never encounters a reason to go further. Her puzzlement isn't performative. We can assume She is genuinely puzzled. And that's the more important and more generalizable insight, because there are millions of people who share her puzzlement for exactly the same structural reasons.
There is a further dimension to this framework that I find striking. In a piece I wrote recently on "Structural Blindness," I explored the observation that large language models are structurally locked at something very close to Level 2. They process the preponderance of content. They weight claims by volume and institutional authority. They can reference the metacognitive tradition; they can tell you about logical fallacies, about checks and balances, about the history of epistemic humility. But they cannot practice it.
An LLM cannot do what a Level 4 human thinker can do: encounter a single verified falsehood and reweight an entire body of evidence, because it understands the institutional and psychological structures that produce coordinated distortion. The LLM processes signals by their statistical weight in the training data. The Level 4 thinker can override statistical weight with structural analysis. The LLM and the Level 2 thinker are doing the same thing by different means: trusting the preponderance.
This matters because we are increasingly delegating our reasoning to systems that are incapable of the very kind of thinking that the metacognitive tradition was built to enable. And we are doing it at a moment when institutional trust is at historic lows, when the gap between official narratives and lived experience is wider than it has been in most people's lifetimes, and when the ability to think structurally about why that gap exists has never been more important.
I said at the start that this framework parallels my Levels of Learning. The parallel is more than structural; it's causal.
Schooling produces Level 1 thinkers: people who absorb the narrative they're given. Training produces Level 2 thinkers: people who become fluent within an institutional frame. Education, when it works, produces Level 3 thinkers: people who learn to question. Self-directed learning produces Level 4 thinkers: people who take full responsibility for their own epistemic situation, including the structures that constrain what they're able to see.
The education system, as it currently operates, is optimized for producing Level 1 and Level 2 thinking (with Level 1 being the majority and Level 2 considered the "best" students). That is not an accident. And the fact that it has largely abandoned the liberal arts tradition, the curriculum specifically designed to move people beyond Level 2, is not an accident either. A population of Level 2 thinkers is a population that defers. A population of Level 3 and Level 4 thinkers is a population that asks uncomfortable questions about why it's being asked to defer.
By now, you know my dad was right. I do think about thinking.
