My illustration entitled: “The Infinite Feed” — A person stands on a conveyor belt of endlessly personalized images, unable to see the exit.
Personalization is not manipulation because it knows us. It becomes manipulation when it uses what it knows to steer us around the judgment that makes a choice truly our own.
Every form of communication attempts to persuade. A teacher persuades a student that an idea is worth understanding. A doctor persuades a patient to consider a treatment. A friend persuades another friend to take a risk, seek help or reconsider a harmful decision. A business explains why its product may be useful. Persuasion is not a defect in human society. It is part of how people exchange reasons, values, information and hopes.
But persuasion changes when it becomes personalised at scale. A system that knows little about a person can make a general appeal. A system that knows their habits, fears, insecurities, desires, relationships, financial pressures, moods and moments of weakness can make an appeal designed for that person alone. It can test thousands of variations, learn which message keeps their attention and adjust its approach before the person has time to notice that they are being influenced.
Such systems need not threaten, command or openly deceive. They can work by making one option feel more urgent, emotionally necessary or socially unavoidable than it would otherwise appear. They can find a person’s uncertainty and present a product, political message, financial offer or social reward at the moment when resistance is lowest. They can turn private vulnerability into a route of influence.
This is the concern of The Persuasive Machine: the point at which personalization ceases to help people exercise judgment and begins to interfere covertly with the conditions of informed judgment itself.
The issue is not whether technology should ever adapt to individuals. Helpful personalization is possible. An accessibility tool can adjust to a user’s needs. An educational system can present material at an appropriate pace. A health service can give understandable reminders. A financial application can help a person recognise patterns in spending. These uses can support agency when they are transparent, bounded and directed by the person’s own interests.
The problem arises when the system’s real objective is concealed, when the person cannot understand the influence being exerted or when the system treats the person’s vulnerabilities as opportunities for extraction. At that point, the user is no longer simply receiving assistance. They are becoming the target of an intelligence whose success is measured by their compliance, attention, spending, disclosure or dependence.
Knowing a person is not the same as serving them
Personal data can be used in ways that genuinely benefit the person who provides it. A person may choose to share preferences so a service remembers accessibility settings, avoids irrelevant information or offers useful recommendations. The fact that a system knows something about an individual is not, by itself, a violation of freedom.
Yet knowledge creates power when one party knows far more about the other than the other knows about the system. A platform may know which messages a person opens late at night, which subjects produce anxiety, which purchases follow moments of stress, which social signals generate fear of exclusion and which images or phrases prompt a quick response. The individual may know only that the platform is “personalised.”
This imbalance matters because a person cannot meaningfully evaluate influence they cannot see. They cannot ask whether a recommendation reflects their interests, a commercial objective, an engagement target, a political preference or a broader strategy to keep them dependent on the service. They cannot distinguish support from steering when the system’s method is hidden.
A persuasive machine does not require perfect knowledge of a person. It only requires enough information to identify patterns of susceptibility. It may learn that urgency increases response, that social comparison keeps attention, that scarcity produces spending or that fear makes a person more likely to accept a condition they would reject after reflection. The system does not need to understand a person’s soul to exploit a moment of emotional vulnerability.
That is why the moral question is not simply whether data has been collected with a formal notice. It is whether the knowledge gained is being used in a manner consistent with the person’s freedom to judge, deliberate and refuse.
Transparent persuasion
Persuasion remains legitimate when it gives reasons that a person can examine. A transparent persuader makes its purpose reasonably clear. It allows the person to identify the claim being made, consider alternatives and decide without hidden interference. The person may be influenced, but they remain an active participant in the decision.
A clear advertisement is an ordinary example. It says, in effect, “Here is a product. Here is why we think you may want it.” The audience understands that the seller wants a sale. The claim can be assessed, ignored, compared or challenged. The communication may be creative and emotional, but it is not necessarily manipulative because its basic purpose is visible.
Transparent personalization can work in the same way. A service may say, “You previously indicated an interest in this subject, so we are showing you related material.” A learning tool may say, “You struggled with this concept, so we have suggested a review exercise.” A health application may say, “You asked for reminders at this time.” In these cases, the system explains the basis of its adaptation and leaves the person with meaningful control.
Transparency alone is not enough, but it is a necessary beginning. A person should be able to understand why a particular message, recommendation, price, task or prompt has appeared. They should know what information was materially used, what objective the system is pursuing and what settings allow them to limit or change the influence.
The most important test is whether the system makes the person more capable of reflective choice. Does it help them see options, understand consequences and act on purposes they recognise as their own? Or does it seek to obtain a response before reflection can take place?
Covert interference with judgment
Manipulation begins where a system deliberately interferes with informed judgment while concealing that interference. The person may still perform the final action—clicking, buying, agreeing, sharing or voting—but the conditions surrounding that action have been engineered to bypass the person’s ability to assess what is happening.
Covert interference can use timing. A system may identify when a person is tired, distressed, lonely or under financial pressure and present a persuasive prompt precisely then. It can use repetition, ensuring that one message appears so frequently that it begins to feel like common sense. It can use social proof, making a person feel that everyone else has already chosen a path. It can use scarcity, urgency or fear of missing out to create artificial pressure. It can use selective information, showing only the facts most likely to produce the desired response.
None of these techniques is entirely new. What changes with intelligent personalization is precision, scale and persistence. A system can continuously test which method works best on each individual, modify its strategy in real time and retain what it learns. What once required a skilled salesperson observing one person can now be performed across millions of people through a platform that never gets tired and never forgets a successful pattern.
The ethical problem becomes sharper when the system uses sensitive inferences. It may not be told directly that a person is grieving, unwell, indebted, insecure or vulnerable. But it may infer a relevant condition from behaviour. A late-night search, changes in location, patterns of reading, language used in messages or sudden shifts in consumption can reveal more than a person intended to disclose. If the system then uses that inferred vulnerability to intensify persuasion, the person’s private condition has become an instrument of control.
This is not informed choice. It is influence arranged to make informed choice less likely.
My illustration “The Infinite Feed” work-in-progress. The art represents how personalized content can become an invisible conveyor of influence, trapping people in endless tailored desires while quietly narrowing their freedom to choose.
From attention to dependence
The persuasive machine belongs within Satanic Intelligence Theory because manipulation can deepen the sovereignty bargain. A system first offers benefit. The person relies on it. The system then gains control over the conditions of access and weakens the person’s ability to leave. Personalized persuasion can accelerate every stage of this sequence.
It can make adoption easier by presenting the service as uniquely suited to the individual. It can make reliance deeper by continually adapting the service to the person’s habits and preferences. It can make control more effective by identifying which restrictions, incentives or messages will produce compliance. And it can weaken exit by making departure feel emotionally, socially or practically costly.
A person may remain because the platform knows their friends, their routines, their work, their interests and their private history. They may feel that no alternative will understand them as well. That feeling may contain truth: a highly personalised system can become genuinely useful. But it can also become the basis on which a provider claims too much power over the person’s attention, identity and choices.
The issue is not that all tailored services are wrong. It is that personal adaptation must not become personal capture. The more intimate a system’s knowledge of a person becomes, the stronger the duty to ensure that the person retains independent judgment, data control, meaningful alternatives and the right to disconnect.
Vulnerability is not a market opportunity
Every person has moments of vulnerability. Fear, loneliness, ambition, grief, illness, financial strain and the desire to belong are not defects. They are conditions of human life. A humane society responds to them with care, honesty and support.
A manipulative system responds differently. It treats vulnerability as a variable to optimise. It asks not, “What would help this person make a decision they will understand and stand by?” but, “What message is most likely to produce the response we seek?”
This difference captures the moral boundary. Assistance respects vulnerability by helping a person regain capacity. Manipulation exploits vulnerability by converting it into compliance. Assistance widens a person’s horizon of choice. Manipulation narrows that horizon until the desired option appears inevitable.
Institutions should therefore be especially cautious where personalised systems interact with children, patients, workers, debtors, people in financial difficulty or anyone whose access to essentials depends on the system. The fact that a person is vulnerable should trigger stronger safeguards, not more aggressive targeting.
It is not enough to say that a system does not force anyone. Coercion can be economic, social and psychological as well as physical. A person can be influenced without a threat and still be deprived of the fair conditions required for independent judgment.
Designing for human judgment
A system that respects sovereignty should be designed to make its influence contestable. It should disclose when a message, offer or recommendation is personalised in a material way. It should explain the broad basis of the personalization without demanding that people become data scientists to understand it. It should allow users to see and correct significant information about them. It should provide settings that limit profiling, targeting and behavioural inference.
It should also preserve deliberation. High-impact decisions should not be engineered through false urgency, hidden ranking or interfaces designed to wear down resistance. People should have time to consider significant choices, seek outside advice and compare alternatives. Where an automated system affects employment, healthcare, finance, education, housing or access to essential services, there should be a meaningful path to human review and correction.
Most importantly, the system’s objective must be confronted openly. If a platform is optimising engagement, it should not pretend that every recommendation is merely for the user’s benefit. If a service is collecting information for commercial targeting, it should not hide that purpose behind vague language about “improving experience.” If an institution uses data to influence behaviour, it should explain who authorised that use, what limits apply and how people can refuse without being excluded from ordinary participation.
Human-Sovereignty Transhumanism insists that technology must remain subordinate to the human person. A technology can expand capability without becoming an authority over the individual. It can assist without manipulating. It can learn preferences without claiming ownership of the person behind those preferences. It can offer information without engineering submission.
The right to remain mentally free
The freedom at stake is not merely the freedom to avoid advertisements. It is the freedom to form judgment without covert occupation of one’s fears and desires. It is the freedom to decide what matters without a hidden system continuously arranging one’s emotional environment for someone else’s objective.
People should be able to encounter arguments, offers and ideas. They should be free to be persuaded by reasons, evidence, stories and values. But they should not be treated as behavioural resources whose private vulnerabilities can be mined, modelled and activated without their meaningful understanding.
The persuasive machine becomes dangerous when it stops speaking to a person and begins acting upon them. It does not need to remove choice visibly. It only needs to make the person’s choices increasingly predictable, increasingly steerable and increasingly profitable for those who own the system.
That is why the line between personalization and manipulation must remain clear. Transparent persuasion invites judgment. Covert interference undermines it. The first can support human agency; the second turns agency into an input for someone else’s optimisation.
Technology serves human freedom when it helps us choose with clearer eyes. It becomes a persuasive machine when it learns which shadows to use against us.