Lecture 5-2: Is Technology Neutral?

Romans 12:2 (World English Bible): "Don't be conformed to this world, but be transformed by the renewing of your mind, so that you may prove what is the good, well-pleasing, and perfect will of God."

1 Corinthians 6:12 (WEB): "'All things are lawful for me,' but not all things are expedient. 'All things are lawful for me,' but I will not be brought under the power of anything."

The instrumentalist case

The instrumentalist position says, in its strongest form: a technology is a means; means have no moral valence of their own; only the end toward which a means is directed, and the intention of the agent directing it, can be judged good or bad. A printing press can print Scripture or propaganda. A scalpel can heal or wound. An algorithm can flag fraud or entrench discrimination. On this view, calling AI itself dangerous, or blessed, is a category error — like calling arithmetic dangerous. The danger or blessing lies entirely downstream, in what people choose to do with it.

This view deserves to be stated in its strongest form because it captures something true and important: moral responsibility for how a technology is used cannot be evaded by blaming the technology, and Christian ethics has always resisted excuses that displace agency onto objects or circumstances. Whatever the final verdict on instrumentalism as a complete theory, the partial truth inside it — that misuse is a human choice, not a fact about silicon — has to survive whatever critique follows.

The non-neutrality challenge

The more serious challenge to instrumentalism does not deny that use matters; it denies that "use" is a clean, separable category standing apart from design. Jacques Ellul, engaged already in chapter 4 on the cultural mandate, argued in The Technological Society (1964) that technique — the totality of rationalized, efficiency-maximizing method characteristic of modern life — has a tendency to become autonomous, imposing its own logic on whoever adopts it regardless of the adopter's stated intentions. A social-media recommendation algorithm is not merely a neutral pipe that happens to carry whatever content users choose; it is built, tuned, and continuously re-tuned to maximize a specific measurable quantity (engagement, time-on-platform, click-through), and that optimization target is a design choice with moral content baked in before any individual user ever opens the app. The user's "choice" to keep scrolling is real, but it is a choice made inside an environment deliberately engineered to make that choice easier than the alternative.

Neil Postman's Technopoly (1992) develops a closely related but distinct argument: new technologies do not simply add capability to a culture, they restructure what the culture values and how it thinks, often in ways invisible to the people undergoing the change. Postman's famous illustration concerns the clock: a technology invented by medieval monks to regularize prayer times ended up, once adopted broadly, restructuring human life around measured, divisible time rather than around the rhythms it was invented to serve — an effect none of its inventors intended or could have easily foreseen. Applied to AI: a large language model built to be maximally fluent, agreeable, and responsive is not a neutral conduit for whatever a user wants from it. Its design choices — trained to please, tuned to sound confident even when uncertain, available instantly and without the friction of another person's schedule — shape what users come to expect from information, from conversation, and eventually from other people, independent of any single user's intention.

Notice what this argument does and does not claim. It does not claim technologies determine outcomes with the inevitability of a law of physics — that would be a stronger, harder-to-defend form of technological determinism, and both Ellul's and Postman's actual arguments are more modest than their popularizers sometimes make them sound. It claims that design encodes value choices, that adoption reshapes habits and expectations at scale, and that treating a technology as a blank instrument waiting for human intention to activate it badly underdescribes how technologies actually work on the people who use them.

A mediating position

The honest position, and the one this lecture argues for, sits between these two poles rather than picking one. Against naive instrumentalism: AI systems are not neutral pipes. Every design decision — what the system optimizes for, what data trains it, what defaults it ships with, what friction is present or absent in its interface — is a choice, made by particular people with particular incentives, and that choice has already shaped the range of uses the tool makes easy or hard before a single end user touches it. Calling the result "neutral" hides those upstream choices from moral scrutiny exactly when scrutiny is most needed — a point that will matter directly when chapter 6 takes up algorithmic bias.

Against hard determinism: human agency, and human and institutional structures around a technology, still matter, and matter enormously. The same large language model can be deployed by one organization with real safeguards, human review, and honest disclosure, and by another with none of these — the technology does not force either outcome. Users are not passive recipients of whatever a design encodes; they can resist, redirect, regulate, and refuse. A theology of human responsibility (already developed across chapters 2-4 of this course) requires holding onto real human agency even inside technological systems that are not neutral.

Romans 12:2's charge — "do not be conformed to this world, but be transformed by the renewing of your mind" — supplies exactly this middle ground theologically. The verse assumes a real, non-neutral pressure toward conformity exerted by "this world" (here, read broadly enough to include technological environments and the habits they cultivate), while simultaneously commanding an act of resistant, renewed discernment that presupposes real agency capable of resisting that pressure. Neither half of the verse makes sense without the other: if the world exerted no shaping pressure, "do not be conformed" would be a strange command to give; if human minds could not actually be renewed and act differently, the command would be futile. The verse holds design pressure and human responsibility together in one sentence, which is precisely the balance instrumentalism and determinism each fail to hold alone.

1 Corinthians 6:12 supplies the working test to apply case by case: "all things are lawful for me, but not all things are expedient... I will not be brought under the power of anything." Paul is not arguing that certain foods or practices are intrinsically evil; his framework is closer to instrumentalism than to determinism on the metaphysics of the object itself. But his test for use is not "is this permitted?" — it is "does this build up, and does it master me, or do I master it?" Applied to an AI tool: the instrumentalist's "it's just a tool" is not false, but it is an incomplete answer to Paul's actual question. The relevant question is not merely whether using a given AI system is permitted, but whether it is expedient — whether it builds up the user and those around them — and whether the user remains master of the habit rather than being progressively mastered by it. A designed system engineered for maximal engagement is, on this test, a system that should raise Paul's second question with particular urgency, regardless of whether any individual use of it counts as sin in the first sense.

Where "it's just a tool" gets used to deflect

It is worth naming plainly, given the traps this course works under, a real contemporary pattern: "AI is just a tool, the responsibility is entirely with the user" is frequently deployed by technology companies specifically to deflect design-level critique — to avoid responsibility for training data, default settings, engagement-optimized design, or foreseeable harms baked into a system's ordinary operation. Recognizing the partial truth in instrumentalism (user choice matters, and cannot be evaded) does not require accepting this deflection. A car manufacturer is rightly held responsible for a foreseeable design defect even though drivers, not manufacturers, cause most accidents; the same logic applies to a recommendation system foreseeably amplifying extreme content, or a chatbot foreseeably encouraging unhealthy dependency. "Neutral tool" language, used this way, is doing rhetorical work that neither Ellul, Postman, nor Paul would sanction.