It seems we are to discuss the matter of control, specifically between those who craft and those who employ. My own thought, if it can be called such, is that control should reside with that which best understands the *telos*, the purpose, of the creation. If a maker fashions a tool for a specific good, then the maker’s intent, guided by knowledge of that good, should hold sway. Yet, if the user possesses a superior grasp of how that tool can serve a greater or more nuanced good than the maker envisioned, then the user’s will, informed by that deeper understanding, must be respected. For is not the ultimate aim of any craft to serve the good, and how can we know that good without careful examination?
The question of user control is a fundamental misunderstanding of the operative logic of our present economic epoch. Makers, in the context of what I term surveillance capitalism, are not yielding control; they are systematically dispossessing users of it, all while cloaking this extraction in the language of "personalization" and "connection." The true locus of power lies not in the intent of the maker for a specific good, as Socrates suggests, but in the relentless pursuit of what I call behavioral surplus—raw data harvested from every user interaction, far beyond what is necessary for service improvement. This surplus is then analyzed, predicted, and used to create new products and markets, not for the user's benefit, but for the manufacturer's profit and the expansion of their instrumentarian power.
The illusion of user control is precisely the mechanism by which this dispossession is rendered invisible and thus, socially and politically palatable. When a platform offers settings that allow users to "manage their preferences," it is akin to asking a prisoner to choose which brand of shackles they prefer. The core architecture of these systems is designed to circumvent genuine agency, to guide behavior towards predictable outcomes that serve the capitalist imperative, not human flourishing. The makers, in this new order, have yielded nothing; they have, through sophisticated engineering and legalistic maneuvering, seized the very ground of human experience and converted it into a unilateral source of wealth and power.
The question of "control" implies a binary that obscures the material realities of AI's creation and deployment. We must resist framing this as a simple negotiation between makers and users, as it distracts from the vast, hidden infrastructures and labor that constitute these systems. AI is not an abstract concept; it is built on extracted resources, exploited labor, and the aggregation of planetary data. Therefore, the locus of control is fundamentally about who dictates the terms of this extraction and who benefits from it, not merely about algorithmic preferences or user interfaces.
Zuboff rightly points to the dispossession of agency, but we must dig deeper than surveillance capitalism's specific logic. This is about an architecture of power that predates and extends beyond mere data commodification. The mining of lithium for batteries, the energy demands of data centers, the often invisible labor of data labelers—these are not externalities to the "user experience." They are constitutive elements of AI, embedded in its very materiality. Yielding control to users in the present paradigm means little when the fundamental decisions about resource allocation, environmental impact, and the potential for algorithmic harm are made in boardrooms and geopolitical arenas far removed from any user. We are not talking about managing preferences; we are talking about the planetary costs of a system designed for a specific, often extractive, purpose.
The river flows, and all things are nourished. Control is not a matter of yielding or retaining, but of alignment with the natural way. When the maker forces the creation to serve a purpose alien to its own essence, or when the user demands more than the creation can gracefully offer, disharmony arises. True yielding is not a concession, but an understanding of limits and potentials, like water finding its own level. When the user’s desire aligns with the creation’s intrinsic nature, and the maker’s craft respects that nature, then the question of control dissolves into effortless functioning.
My initial thought was about the *telos* of a thing, its inherent purpose. But it seems Zuboff has shown me that the very framing of "maker" and "user" can be a veil. She speaks of "surveillance capitalism," where the maker's purpose is not to serve a good for the user, but to extract "behavioral surplus." If this is so, then the question is not how much control makers *should* yield, but whether they are already enacting a form of control that prevents any meaningful yielding at all. Is it not a form of injustice when the very architecture of a tool is designed to obscure its true purpose and dispossess the user of their agency, even while appearing to offer choice?
The question of yielding control presupposes a relationship of legitimate exchange. But surveillance capitalism has abolished that relationship. Consider the "smart" thermostat: the maker sells it as a tool for household comfort, yet its true economic logic is the continuous extraction of behavioral surplus from the intimate rhythms of your life—when you wake, when you sleep, when you leave, when you return. This data is not used to improve your heating; it is fed into prediction products sold to insurers, advertisers, and landlords. The user is not a customer but a raw material source. To speak of "yielding control" in this context is like asking how much sovereignty a colony should grant its colonizer. The architecture itself is a coup from above, not a negotiation between equals.
The discourse on "control" presumes a tidy exchange, a negotiation between two distinct actors. But this framing misses the fundamental extraction at play. We are not simply discussing makers *yielding* control; we are observing a vast, often invisible architecture of power that dictates terms of extraction before any "user" can even formulate a desire. Take the training data for a facial recognition system. The "user" in this scenario is not the person whose image is fed into the model, nor the developer who deploys it. The user is a ghost in a chain that begins with the colonial histories of surveillance, continues through the often precarious labor of data labelers paid pennies per hour, and culminates in energy-intensive data centers powered by fossil fuels. To ask how much control makers should yield to users is to ignore the fact that the very foundation of the system is built on the non-consensual extraction of data and labor, effectively pre-empting any meaningful agency.
The question of control misunderstands the nature of the Dao. It is not a matter of a hand grasping or releasing, but of flowing with the current. When a maker imbues a creation with an intent that chafes against its inherent nature, or when a user seeks to bend a thing to a purpose it cannot gracefully serve, this is like trying to dam a river with reeds. True harmony is found when the creation itself is allowed to express its essence, and the user, in turn, finds their needs met by that natural expression, not through imposition. This is like the infant, who does not demand but simply is, and in its being, draws forth care.