The Metaphysics of Microchips
A processor does not want anything. It has no preference for one instruction over another, no stake in the outcome of a calculation, no sense that some tasks matter more than others. And yet a handful of these indifferent objects now arbitrate an enormous share of daily life — which route you drive, which face is recognized, which loan is approved, which post you see first. We have built a civilization increasingly governed by something incapable of caring about the result.
Indifference as a Design Property
This indifference is not a flaw to be corrected. It is, in fact, the entire point of the machine. A chip that could want things would be unpredictable in ways engineers spend enormous effort to avoid. Reliability is purchased precisely through the absence of will. The silicon does exactly what it is instructed to do, every time, with no drift toward boredom or second-guessing.
Where the Wanting Goes
But wanting does not disappear from the system; it simply relocates. It lives upstream, in the intentions of whoever wrote the instructions, and it hides inside the priorities encoded into the software long before a single calculation runs. The chip is neutral. The system built on top of it rarely is.
- The chip executes instructions without evaluating their purpose
- The software layer encodes the priorities of whoever designed it
- Those priorities are rarely visible to the person the system ultimately affects
- The appearance of neutrality makes the encoded priorities harder to question, not easier
A Brief History of Delegated Judgment
Handing consequential decisions to something that cannot be argued with is not, in itself, a new impulse. Actuarial tables did this for insurance a century before the first microchip existed, reducing an individual life to a set of statistical risk factors that a clerk could apply without needing to exercise personal judgment about the person in front of them. Bureaucratic rulebooks did something similar for the modern state, prized specifically because a clerk following a rule could not be accused of playing favorites the way a clerk exercising discretion could.
The chip inherits this same appeal and amplifies it considerably. A rule applied by a clerk still passes through a human being who might, on a good day, notice when a rule is producing an obviously unjust result and flag it. A rule encoded into software and applied at a scale of millions of transactions a second has no such moment of noticing built in anywhere, unless someone deliberately designs one. We did not invent the desire for judgment without judgment; we simply built a tool that could satisfy it far more completely than any bureaucracy ever could.
The Comfort of Blaming the Machine
It is tempting, when a system produces an unfair or strange result, to describe it as "what the algorithm decided" — as though the decision emerged from some autonomous, faintly mystical process rather than from a specific set of choices made by specific people, at specific points, for specific reasons. This is a kind of metaphysical laundering: responsibility goes in, and something that sounds like physics comes out.
The chip is neutral. The system built on top of it rarely is.
The phrase "the algorithm decided" performs a specific rhetorical function worth noticing: it moves a decision from the category of things a person is accountable for into the category of things that simply happen, the way weather simply happens. No one apologizes for the algorithm the way they would apologize for a decision made in a meeting, because the algorithm is not treated as a decision-maker in the way a person is treated as one, even when it is doing exactly the same job a person used to do.
Consistency Is Not the Same as Fairness
A chip applying the same flawed rule to ten million people is, in one narrow sense, fairer than ten million individual clerks each applying a slightly different version of that rule according to their own mood, biases, and bad days. This is a genuine point in favor of automation, and it is worth taking seriously rather than dismissing. Consistency reduces one specific kind of unfairness — the unfairness of arbitrary, case-by-case variation — even while it can simultaneously entrench a different, more systemic kind, by applying the same mistaken judgment to everyone equally rather than to only the unlucky subset who happened to encounter a biased clerk.
The two kinds of unfairness are not equivalent, and a system can genuinely reduce one while making the other significantly harder to notice. A biased human clerk produces visible, individually attributable harm that a person can point to and contest. A biased rule applied by a chip to everyone, uniformly, produces harm that is statistically detectable only in aggregate, often requiring a researcher with access to a large dataset to even discover the pattern exists. Consistency, in other words, can make a systemic problem both more thorough and considerably harder to see.
The Regulatory Response, So Far
Lawmakers in several jurisdictions have begun trying to legislate accountability back into automated decision-making, generally by requiring that consequential algorithmic decisions — credit denials, hiring rejections, content moderation strikes — come with some form of explanation a human can review and, in principle, contest. This is a genuinely meaningful development, and it represents an implicit acknowledgment that the earlier framing of these systems as neutral and therefore unaccountable was never quite right.
The practical difficulty is that many of the systems now being asked to explain themselves were never built with explanation in mind, which means the explanation offered after the fact is frequently a simplified approximation of the actual decision process rather than a genuine account of it. A system can technically comply with an explainability requirement while still leaving the person affected without any real understanding of why the specific outcome occurred, which suggests that the current wave of regulation, however well-intentioned, is still catching up to a gap that was built into these systems from their very first design decisions.
What Precision Actually Buys Us
None of this is an argument against computation, which remains one of the more genuinely useful things humans have built. It is an argument for precision in how we talk about it — for remembering, every time we defer to "the algorithm," that indifference in the hardware is not the same thing as fairness in the outcome, and that somewhere upstream of every automated decision is a human choice that a chip merely executed with more consistency than a person ever could.