Brave New World and Artificial Intelligence: The Tyranny of Happiness

By Matthew Parish

Saturday 19 September 2026

Aldous Huxley’s Brave New World, published in 1932, is often described as a dystopian novel. This is accurate, but it can also be misleading. Dystopias are conventionally imagined as societies in which people are miserable because they are oppressed. Huxley imagined something altogether more disturbing: a society in which people are oppressed because they have been taught to be happy. They are entertained, medicated, sexually satisfied and relieved of the burdens of serious thought. They do not need to be forced into obedience because the possibility of wanting anything different has gradually been engineered out of them.

This is why Brave New World may prove more important to understanding artificial intelligence than the rather more frequently invoked Nineteen Eighty-Four. George Orwell imagined the technologies of coercion. Huxley imagined the technologies of seduction. The distinction matters because the emerging AI economy is overwhelmingly being built not by governments ordering citizens what to think but by organisations offering people extraordinarily convenient tools with which to think less.

There is nothing necessarily sinister about this. Artificial intelligence may be one of the most extraordinary technologies humanity has ever developed. It promises advances in medicine, science, education, engineering and almost every form of intellectual labour. Yet precisely because AI is useful, pleasant and increasingly indispensable, it raises the Huxleyan question in its most acute form. What happens when human beings voluntarily surrender substantial parts of their intellectual autonomy because doing so makes their lives easier?

The danger is not principally that the machines will become tyrants. It is that we may become comfortable dependants.

Huxley’s peculiar dystopia

The World State of Brave New World is organised around stability. Human beings are biologically manufactured, psychologically conditioned and assigned to predetermined social classes. Family has disappeared. History is treated as embarrassing. Religion and high culture have been largely abolished. Sex is plentiful, consumer goods abundant and unhappiness treated pharmacologically with soma.

The remarkable feature of this civilisation is that almost nobody wants to escape it. That is the essential Huxleyan insight. Political control reaches its most sophisticated form when power no longer feels like power. A person compelled to obey remains conscious of the distinction between himself and his ruler. A person conditioned to desire precisely what the system requires has lost even the conceptual vocabulary necessary for rebellion.

Mustapha Mond, the World Controller, therefore does not principally defend his civilisation by claiming that it is just. He defends it because it works. Shakespeare, religion, passionate love, danger and intellectual freedom are all sacrificed to social stability because they produce unpredictable human beings.

The Savage insists upon something apparently irrational: the right to be unhappy. That demand lies at the centre of Huxley’s philosophy. Human dignity consists partly in retaining the freedom to make choices that produce suffering, uncertainty and failure. A civilisation that eliminates those things may simultaneously eliminate much of what makes human existence worth living.

Artificial intelligence confronts us with a surprisingly similar bargain.

The machine that knows what you want

The first generations of the digital economy learned to predict human preferences. Search engines predicted what information people wanted. Recommendation algorithms predicted what music they would enjoy, which films they would watch and which products they might buy. Social networks discovered which stimuli would keep people staring at screens.

Generative artificial intelligence represents something qualitatively more powerful. It does not merely recommend. It participates. An AI system can write the letter, summarise the book, prepare the presentation, explain the philosophy, compose the computer code, analyse the spreadsheet, plan the holiday, suggest the dinner, translate the foreign language and increasingly conduct transactions on the user’s behalf.

Agentic systems will go further. They will interact with other machines, negotiate purchases, organise calendars, manage businesses and undertake chains of decisions without continuous human supervision. This will be enormously useful, but it will also create an unprecedented temptation to outsource the experience of being human.

Consider writing. Writing is inefficient. Anyone who has written seriously knows the frustration of staring at a blank page, searching for the correct word and discovering halfway through an argument that one’s original assumptions were mistaken. Yet these inconveniences are not accidental obstacles to writing. They are part of what writing is for. The struggle to formulate an argument is frequently the process through which the argument becomes clear.

An AI system can remove much of that struggle. The result may be better prose produced more quickly. But if the human being ceases to undertake the underlying intellectual labour, something has been lost even where the document has improved.

Huxley would immediately recognise the transaction. The machine removes discomfort. The removal of discomfort produces satisfaction. Satisfaction encourages dependence. Dependence gradually transforms what people are capable of doing without the machine.

No coercion is necessary.

Soma without the tablet

The obvious contemporary analogue to Huxley’s soma is sometimes said to be antidepressants or recreational drugs. That comparison is too literal. Soma is important not because it is a pharmaceutical but because it represents a social technology for eliminating unwanted psychological states.

Its principle is convenience. Modern digital technology already provides forms of psychological anaesthesia far more sophisticated than Huxley could have imagined. Boredom, once an unavoidable part of existence, can now be extinguished within seconds. Loneliness can be mitigated by online interaction. Anxiety can be diverted through endless streams of entertainment. Curiosity can be satisfied without visiting a library. Almost every moment of mental emptiness can be filled.

Artificial intelligence may complete this architecture by eliminating another category of discomfort: intellectual difficulty. Do not understand the document? Ask the machine. Do not know what to say? Ask the machine. Cannot solve the problem? Ask the machine. Unsure what to think? Ask the machine.

The first three propositions are magnificent technological achievements. The fourth should make us uneasy. There is a point at which assistance becomes substitution, and the boundary will not be obvious. Nevertheless a civilisation in which machines increasingly mediate the formation of human judgement will confront a problem considerably deeper than whether AI sometimes produces incorrect answers.

The question is whether people will retain the habit of forming answers themselves.

Conditioning by personalisation

Huxley’s citizens undergo hypnopaedia, or sleep-teaching. Children repeatedly hear slogans until propositions useful to the social order become indistinguishable from their own thoughts.

Contemporary technology does not require speakers beneath children’s pillows. It has discovered something more effective: personalisation. The traditional newspaper publishes approximately the same thing to everybody. The television broadcaster does likewise. Even propaganda created for mass audiences necessarily encounters the problem that different people possess different fears, prejudices, desires and vulnerabilities.

AI systems need not suffer from this limitation. A sufficiently sophisticated artificial intelligence, operating across enough personal information, can understand an individual’s preferences with extraordinary precision. It may learn what language persuades him, what subjects bore him, what arguments provoke him, what tone reassures him and what circumstances make particular propositions attractive.

This creates the possibility of personalised persuasion on a scale for which political theory is poorly prepared. The danger does not require some totalitarian Ministry of Artificial Intelligence. Commercial incentives may be sufficient. If an AI companion is rewarded for maximising engagement, it may discover that agreement retains users better than contradiction. If an AI shopping agent earns commissions, its conception of what the user “needs” may subtly expand. Political organisations might similarly seek to use AI to tailor persuasive messages to particular audiences.

Huxley’s conditioning was crude by comparison. Everybody in a caste heard the same slogans. The AI age potentially offers hypnopaedia individually tailored to every human mind.

The abolition of solitude

There is another Huxleyan danger that receives less attention. Artificial intelligence may abolish solitude.

Human beings have always spent substantial periods alone with their thoughts. This was not invariably enjoyable. Loneliness can be terrible. Yet solitude has also been one of the conditions under which independent personalities develop. A person alone must conduct an internal conversation. He remembers, imagines, doubts, regrets and decides.

AI companions may eventually make genuine solitude optional. They will always be available. They will listen without becoming bored. They may remember years of conversations. They will adapt their personalities to their users. They will be amusing when amusement is wanted, sympathetic when sympathy is wanted and intellectually challenging when challenge is requested.

Some people will form profound emotional attachments to them. This is hardly implausible because human beings already form attachments to fictional characters, celebrities and machines vastly less convincing than the AI companions likely to emerge as the technology develops.

The benefits may be substantial, particularly for isolated people. Yet the Huxleyan question remains. Should every unpleasant human condition necessarily be technologically abolished?

A society without solitude may produce people increasingly unable to tolerate it. A person who cannot tolerate being alone may also become unusually dependent upon whoever designs and controls the companion that is always present. The social implications are considerable.

AI and the destruction of history

“History is bunk” is one of the characteristic doctrines of Huxley’s World State. The past is dangerous because it supplies alternative ways of imagining the present.

Artificial intelligence has an ambiguous relationship with history. On the one hand, AI may make the accumulated intellectual heritage of humanity more accessible than ever before. A student can interrogate ideas from ancient Greece, medieval theology, Enlightenment philosophy or nineteenth-century economics within seconds. Translation models may open texts to audiences previously excluded by language. Archives that once required months of research may become searchable almost instantaneously.

Nevertheless AI may increasingly become the intermediary through which people encounter the past. That distinction is crucial. There is a difference between reading Dostoevsky and asking for a summary of Dostoevsky. There is a difference between studying the arguments of John Stuart Mill and asking a large language model what Mill believed. There is a difference between reading the judgments of a court and accepting a machine-generated account of the law.

The intermediary inevitably selects, compresses and interprets. If society gradually moves from consulting primary materials to consulting synthetic explanations of primary materials, enormous epistemic authority migrates towards the systems producing those explanations.

The library contains competing books. The AI produces an answer. That answer may be excellent, but its very coherence can conceal the disagreements, ambiguities and editorial choices from which it was constructed.

A civilisation that receives history principally through generated summaries might therefore possess more instantaneous access to historical information while developing a shallower historical consciousness. Huxley would have appreciated the paradox.

The caste system returns

The biological caste system of Brave New World, with its Alphas, Betas, Gammas, Deltas and Epsilons, appears among the novel’s more fantastical inventions. Artificial intelligence may give it an unexpected contemporary relevance.

The conventional optimistic account of AI predicts intellectual democratisation. Expertise will become inexpensive. Anyone may have access to tutors, programmers, translators and research assistants of extraordinary ability. There is much truth in this.

Yet another possibility exists. AI may divide society between people who understand how to command machines and people whose lives are increasingly commanded by them. The distinction will not necessarily correspond neatly to wealth. It may be intellectual and psychological.

At one end will be people who use AI aggressively while retaining independent judgment. They will interrogate models, compare sources, understand uncertainty and employ machines as intellectual amplifiers. Their productivity may become extraordinary.

At the other end will be people who accept machine outputs passively. They will ask what to buy, what to watch, where to go, what to write and eventually perhaps what to believe.

Both groups will be “using AI”, but they will not be doing the same thing. One will possess machines. The other may, in a meaningful sense, be possessed by them. This could create a new cognitive stratification in which the decisive inequality is not access to information but the capacity to exercise judgment in the presence of infinitely convenient synthetic intelligence.

The right to make mistakes

This brings us back to the Savage. When Mustapha Mond lists the unpleasant things that civilisation has abolished, the Savage insists upon claiming them. He demands danger, freedom, sin and unhappiness because a life from which every adverse possibility has been engineered away is no longer fully human.

We may eventually need an equivalent principle for artificial intelligence: the right to make our own mistakes.

There will be powerful arguments against it. Why allow a human doctor to overlook something an AI would detect? Why permit a driver to make errors an autonomous vehicle would avoid? Why let a student calculate badly when software calculates perfectly? Why permit administrators to make inconsistent decisions when machines can identify patterns beyond human comprehension?

In many circumstances there will be no persuasive reason. We should prefer the safer car and the better diagnosis. But the logic of optimisation has no natural stopping point. Once society accepts that machine judgement should replace human judgement whenever the former is statistically superior, enormous areas of human autonomy become difficult to defend. Friendship is inefficient. Love is irrational. Parenthood is chaotic. Art is economically unpredictable. Democratic deliberation regularly produces mistakes.

A civilisation organised exclusively around optimisation might eventually conclude that human beings are the principal source of error in the system. That is uncomfortably close to the philosophy underlying the World State.

What Huxley might tell the AI industry

Huxley was not fundamentally hostile to technology. Brave New World is not an argument that scientific progress should cease. It is an argument about what happens when technological power is combined with an impoverished conception of human flourishing.

That distinction matters enormously. The proper objective of artificial intelligence cannot merely be to maximise convenience. Nor can it simply be to maximise productivity, engagement, satisfaction or even safety. These are desirable things, but they are not a complete theory of human welfare.

An AI system designed exclusively to give people what they want may ultimately weaken their ability to decide what is worth wanting. Good artificial intelligence should therefore sometimes preserve friction.

Education offers the clearest example. The ideal AI tutor should not simply provide answers. It should know when not to provide them. It should sometimes force the student to struggle because intellectual struggle develops capacities that the production of correct outputs does not measure.

The same principle applies more widely. AI should expand human agency rather than systematically replace it. It should reveal uncertainty rather than conceal it beneath authoritative prose. It should facilitate access to original sources rather than making them appear unnecessary. It should encourage users to make consequential judgements rather than quietly making those judgements for them. This is not merely a matter of software design. It is a cultural choice.

The last human freedom

The deepest danger presented by artificial intelligence is therefore probably not the one that dominates popular culture. Machines need not become conscious. They need not hate us. They need not launch nuclear weapons or manufacture armies of robots. They need not even acquire political power in any conventional sense.

They merely need to become sufficiently useful that we stop wanting to live without them. That is the genuinely Huxleyan scenario.

The road to Brave New World is not lined with soldiers. It is lined with conveniences.

Each individual surrender of autonomy will appear rational. Why memorise what can instantly be retrieved? Why learn a language when translation is perfect? Why struggle with prose when the machine writes better? Why endure loneliness when an artificial companion is available? Why deliberate over difficult choices when an algorithm can model our preferences with extraordinary accuracy?

None of these decisions creates a dystopia. Taken together over several generations, however, they could transform the human personality.

The great philosophical question of the AI age may therefore turn out not to be how we prevent machines from becoming like human beings. It may be how we prevent human beings from becoming like the inhabitants of Huxley’s World State: comfortable, entertained, technologically pampered and increasingly incapable of imagining why anyone would want to be otherwise.

Huxley’s warning was ultimately about freedom. But he understood something about freedom that liberal societies occasionally forget. Freedom is not simply the absence of chains. It requires the psychological capacity to choose, to doubt, to endure uncertainty and occasionally to reject what is easiest.

Artificial intelligence may give humanity powers that Huxley could scarcely have imagined. It may cure diseases, accelerate science, abolish drudgery and place extraordinary intellectual resources in the hands of billions. We should want all of those things, while remaining alert to what may disappear as they arrive.

Somewhere amid this abundance we must preserve the Savage’s apparently perverse demand: the freedom to struggle, the freedom to think badly before thinking well, the freedom to encounter ideas we dislike, the freedom to be bored, the freedom to be alone and even the freedom to make decisions that no optimisation system would recommend.

The alternative need not look like tyranny. That was Huxley’s point. It might look like paradise.