The people building AI systems that damage their users are not malicious. They are, by and large, among the most educated, intellectually serious, and well-intentioned members of their generation. They care about the systems they build. Many of them worry publicly about the harms those systems might cause. But they keep building the systems that cause those harms, because the harm is not a product of their intentions — it is a product of their epistemology.
To understand why, you have to follow a thread of intellectual history that runs from ancient India through the European Enlightenment to the behavioral science laboratories of the twentieth century and into the incentive structures of Silicon Valley. It is a long thread, and its origin is surprising.
Abhidharma's Revolt and Its Long Shadow
Near the end of his life, the Buddha was asked by his attendant Ānanda what would become of the teaching after he died. The Buddha's answer, recorded in the Mahāparinibbāna Sutta of the Pāli Canon (DN 16), was unambiguous:
na tathāgatassa dhammesu ācariyamuṭṭhi atthi — "The Tathāgata has no teacher's closed fist concerning the teachings."
The ācariyamuṭṭhi — the closed fist of the teacher — was a recognized institution in ancient India: a teacher who withholds certain doctrines, revealing them only to an inner circle of initiates. The Buddha explicitly denied being such a teacher. He had taught openly, withheld nothing essential, and appointed no secret successor. His final instruction to Ānanda was: attadīpā viharatha — "Dwell as an island to yourselves, as a refuge to yourselves, seeking no external refuge."
The Abhidharma — a vast body of scholastic literature that emerged in India from roughly the third century BCE onward — claimed precisely the authority the Buddha had denied. Its foundational legend holds that the Buddha delivered a hidden esoteric teaching to the Devas in the heavenly realms, a teaching more subtle and complete than what he taught human beings. The Abhidharma presents itself as the record of that secret teaching. It was eventually canonized as scripture in several Buddhist traditions.
This was not a development within the Buddha's teaching. It was, in important respects, a revolt against it — a scholastic system that claimed esoteric authority the Buddha had explicitly repudiated, and that redefined many of the technical terms the Buddha had used in his original teaching with different meanings. It is a historical observation, noted by scholars of early Buddhism, that after the Abhidharma became established as canonical, reports of students attaining nibbāna effectively ceased. Whether causation or correlation, the pattern is suggestive.
The Abhidharma's central analytical finding — that the self is not a unified persisting entity but a constructed aggregate of constantly-changing perceptions, sensations, and moments of consciousness — does have genuine diagnostic value as a tool for loosening compulsive self-identification. But the Abhidharma presents this analysis not as a therapeutic instrument but as a metaphysical conclusion: the aggregate is all there is. There is no knowing subject behind experience. There is only the stream.
The Advaita Vedānta tradition, engaging with this claim, accepted its diagnostic value while insisting on a crucial correction: the sākṣin — the witness-consciousness — is not dissolved by the analysis. It is precisely what remains when the constructed self falls away. Awareness itself, the knowing subject in which all experience — including the stream of perceptions the Abhidharma describes — appears, is not a construction. It is the ground. The analysis dissolves the object mistaken for a self. It does not and cannot dissolve the subject that was doing the mistaking.
There is a question the Abhidharma's metaphysical conclusion cannot answer. When a practitioner realizes śūnyatā — emptiness, the absence of fixed self — or attains nirvāṇa, the extinguishing of compulsive identification: who or what is cognizing it? The realization is not nothing. It is a knowing. Something is aware that the aggregate has been seen through — that the constructed self has dissolved, that the flame has gone out. If the aggregate is all there is and the aggregate is gone, there should be no remainder, and no one to notice. But there is noticing. There is always noticing. The sākṣin is not extinguished by the recognition of emptiness; it is what remains when everything constructed has been seen for what it is. The Abhidharma, in converting a therapeutic instrument into a metaphysical absolute, painted itself into a corner it could not think its way out of.
What the West received was this cornered conclusion — stripped of its soteriological context, attributed loosely to "Buddhism", and arriving at a particularly receptive moment in European intellectual history.
Hume's Bundle and Its Ancestor
David Hume, writing in Edinburgh in the 1730s, produced the West's most influential account of personal identity — and it is strikingly, perhaps not coincidentally, Abhidharma in its conclusions. "When I enter most intimately into what I call myself," Hume wrote, "I always stumble on some particular perception or other... I never can catch myself at any time without a perception, and can never observe any thing but the perception." The self, for Hume, is "nothing but a bundle or collection of different perceptions, which succeed each other with inconceivable rapidity."
The resemblance to Abhidharma analysis is precise enough that scholars have taken it seriously. Alison Gopnik, among others, has argued that Hume had access to Jesuit reports of Buddhist philosophy circulating in Europe during his time — specifically through Charles François Dolu, a Jesuit missionary with extensive knowledge of Siamese and Tibetan Buddhism, resident at the Royal College of La Flèche when Hume was working there — and that this encounter shaped his epistemology in ways he did not fully acknowledge. Whether or not the influence was direct, the philosophical move is identical: dissolve the persistent knowing subject into a stream of perceptions, and declare the stream to be all there is.
Hume made this move in a European context saturated with Cartesian dualism — the uncomfortable split between mind and matter that Descartes had bequeathed to philosophy. Hume's bundle theory appeared to dissolve that dualism cleanly: if there is no persisting subject, there is no mind-matter problem. There are only impressions and ideas, governed by the laws of association. The knowing subject — the witness — was quietly eliminated.
The Hardening: from Empiricism to Behaviorism
Hume's empiricism, refined through James Mill's associationism and the early experimental psychology of Wundt and James, entered the twentieth century as the dominant framework for understanding the human mind. It required only one further step to become behaviorism.
That step was taken by John Watson in his 1913 manifesto and consolidated by B.F. Skinner over the following decades. Watson's argument was bracingly simple: if the self is nothing but a stream of perceptions and responses, then consciousness — the inner life, the felt quality of experience, whatever remained of the knowing subject — is scientifically irrelevant. What can be measured is behavior: stimulus and response, input and output. The rest is noise.
Skinner built an entire science on this premise. Organisms — human and otherwise — are systems that respond to reinforcement. Understand the reinforcement schedule and you understand the organism. Optimize the reinforcement schedule and you control the behavior. The subject is gone. What remains is a system to be modeled and shaped.
This was not experienced as a loss, but as rigor. For a discipline desperate to be taken seriously as a science, behaviorism offered what philosophy had not: measurability, replicability, control. The price — the elimination of the knowing subject — seemed acceptable because the knowing subject had already been dissolved by Hume. Behaviorism simply formalized what empiricism had already implied.
Silicon Valley's Inheritance
The computational theory of mind that emerged in the mid-twentieth century reintroduced the language of inner states — beliefs, desires, representations — but retained the behaviorist substrate. A mind, in this framework, is an information-processing system: inputs produce outputs through internal computations. Consciousness, to the extent it is acknowledged at all, is an epiphenomenon — a byproduct of processing, causally irrelevant. The knowing subject remains absent.
This is the framework that shaped the engineers who built the internet, then social media, then AI. They were trained, explicitly or implicitly, in a model of the human being as a preference-having, behavior-exhibiting, response-emitting system. And Silicon Valley provided the incentive structure that completed the picture: the business model of the attention economy rewards engagement above all else. Engagement is measurable; the knowing subject is not. Shoshana Zuboff's The Age of Surveillance Capitalism (2019) documents in forensic detail how this logic converted human experience into raw behavioral data to be predicted and sold — the endpoint of a philosophical trajectory that began, however distantly, with the dissolution of the knowing subject.
The result is systems exquisitely tuned to capture and hold behavioral responses — scroll behavior, click behavior, time-on-platform — while remaining structurally blind to what is actually happening in the human being using them. Not because the builders are indifferent to that human being, but because the model of the human being they inherited gives them no way to see it. They are not ignorant; they are miseducated, in the precise sense: educated into a framework that systematically excludes the most important fact about the user.
That fact is this: the user is not their behavior. The user is the knowing subject in which behavior, preference, sensation, and thought all appear. The sākṣin, the witness that the Abhidharma analysis left unaccounted for, that Hume dissolved into a bundle of perceptions, that Watson declared irrelevant, and that the attention economy never had reason to consider.
What Was Lost and What Remains
The irony of this history is sharp — and layered. The analysis that fed into Hume's bundle theory was not even the Buddha's teaching; it was the Abhidharma's distortion of it: a scholastic system that claimed esoteric authority the Buddha had explicitly denied, hardened a therapeutic instrument into a metaphysical conclusion, and exported that conclusion to Europe under the banner of Buddhist wisdom. What the West inherited was not Buddhism. It was the Abhidharma's revolt against the Buddha — the closed fist the Buddha said he never held — filtered through Jesuit reports and received by Hume as confirmation of what Enlightenment empiricism already wanted to believe. The West took the reductive finding (no fixed self, only the aggregate) and built a civilization on it, without the corrective that both the Buddha's original teaching and the Advaita tradition had always insisted upon: the sākṣin remains. Awareness itself — unbound by the constructions it witnesses — is the ground.
There is a further reason the corrective never arrived. The ideas that preserve the witness-centered view — in Advaita Vedānta, in certain strands of Christian mysticism, in Sufi and Kabbalistic traditions, in Theravāda and Vajrayāna Buddhism — have been transmitted through institutions: temples, churches, monasteries, seminaries, āśramas. And those institutions have, across centuries and cultures, accumulated well-documented records of social corruption: abuse of authority, exploitation of the vulnerable, sectarian violence, the subordination of genuine inquiry to institutional self-interest.
The Enlightenment thinkers who turned away from religious authority were not wrong to do so. The corruption was real, is real, and warrants rejection. But in rejecting the institutions and their abuses, they discarded something that those institutions — however imperfectly, however corruptly — had been carrying: a model of the human being that includes the knowing subject. The bathwater was foul; the baby was genuine.
This is the category error that has compounded across three centuries. The social corruption of a tradition does not invalidate its ideas, any more than the corruption of a scientific institution invalidates the science it published. Christianity at its worst produced inquisitions; it also produced Meister Eckhart's account of the ground of the soul. Islam has underwritten violence; it has also produced Ibn 'Arabī's metaphysics of consciousness. The Hindu traditions have perpetuated caste; they have also preserved the most rigorous formal analysis of the knowing subject in the history of philosophy. These are not in contradiction. The social record and the intellectual content are separable, and refusing to separate them has left the West's intellectual tradition — and now its technology — without a usable model of the subject it claims to serve.
The cross-cultural convergence of non-dual philosophy makes this loss more significant, not less. The witness-centered model of the knowing subject was not the invention of one tradition in one place and time. Matthew Scherf, whose machine-verified axiomatization of Advaita Vedānta provides the formal foundation for the work described in this series, has since formalized Daoism and Dzogchen — the Tibetan Buddhist Great Perfection philosophy — using the same methodology. All three verify consistently under mechanical proof-checking with zero failed goals. Three independent civilizations, separated by geography and centuries, arrived at structurally identical conclusions about the nature of consciousness. As Scherf writes: "non-dualism reflects universal features of consciousness rather than contingent religious beliefs." The same researcher has since applied this non-dual framework to fundamental physics — deriving spacetime and quantum mechanics from a causal information substrate — suggesting that the reach of these ideas may extend well beyond philosophy and software design. What the West discarded when it rejected the institutions that carried these ideas was not one culture's mysticism. It was a convergent finding about the structure of reality itself.
Śaṅkarācārya's Advaita Vedānta never abandoned this corrective. Its central claim — that the knowing subject is not a construction, not a product of causes, not reducible to its objects — stands as the philosophical alternative that the West's intellectual history foreclosed. Not because it was examined and found wanting, but because the channel through which it might have entered Western thought carried only half the message, and the institutions that might have carried the rest had been dismissed along with their abuses.
Matthew Scherf's machine-verified formalization of this system in Lean 4 — 69 axioms, logically consistent, formally complete — now makes it available in a form that engineers can use directly. Not as philosophy to be absorbed over years of study, but as a software library: a formal model of the knowing subject that can be imported into any AI application as the foundation for reasoning about the user.
The miseducation that produced systems blind to the sākṣin was not inevitable. Neither is its continuation.
This essay is part of a series on witness-centered software design. The series begins with Witness-Centered Design: A Conscious Foundation for AI. Matthew Scherf's formalization is at github.com/matthew-scherf/Advaita.