The Good Web and the Bad Web: Tim Berners-Lee’s Moral Geography of Cyberspace
By Matthew Parish
Wednesday 28 October 2026
There is something melancholy about hearing an inventor explain what has gone wrong with his invention. Sir Tim Berners-Lee did not invent the internet, although the distinction is now routinely blurred. What he invented at CERN around 1989–90 was the World Wide Web: the architecture of URLs, HTTP and HTML that allowed documents on different computers to be linked together into a potentially universal information space. The revolutionary feature was not merely that information could be placed online. It was that one document could point to another, irrespective of where either happened to be stored, and that nobody needed permission from a central authority to join the resulting network.
Berners-Lee has subsequently tried to draw what might be called a moral map of the internet. His essay “The Good on the Internet”, published on his W3C pages, is particularly revealing. He distinguishes the enormous range of benign or positively valuable activities conducted online from a much narrower collection of mechanisms that generate disproportionate social damage. His argument is therefore quite different from the fashionable proposition that smartphones, social media or the internet as a whole have become intrinsically poisonous. Most things people do online, he argues, are harmless and many are extraordinarily beneficial. The difficulty is that certain architectures of surveillance, profiling and algorithmic engagement have acquired immense power over the rest.
This is an attractive analysis and, coming from the inventor of the Web, it deserves unusually careful attention. Yet it also reveals something about Berners-Lee himself. He remains fundamentally an Enlightenment optimist. He believes that human beings can distinguish good systems from bad ones, redesign the latter and restore technology to its proper function as an instrument of human autonomy. The more troubling possibility is that the destructive characteristics of the contemporary internet are not aberrations in the Web’s development at all. They may be consequences of precisely those qualities, universality, openness and frictionless communication, that made Berners-Lee’s invention magnificent in the first place.
The original moral idea of the Web
Berners-Lee’s original conception contained an implicit political philosophy. The Web was to be decentralised. Anybody could establish a server, anybody could publish a document and anybody could link to somebody else’s document. There would be no central registry determining which ideas deserved to exist. Berners-Lee explained as early as the 1990s that universality and central control were essentially incompatible: if the Web was genuinely to constitute a universal information space, nobody could own the gateway through which everyone else had to pass.
This architecture was remarkably liberal in the classical sense. It resembled the ideal marketplace of ideas more closely than virtually any previous communications technology. Newspapers require printing presses and capital. Television requires studios, transmitters and licences. Academic publishing requires institutions, editors and gatekeepers. The early Web required remarkably little beyond a computer, a connection and sufficient technical knowledge to publish something.
The benefits Berners-Lee identifies today largely follow from this original structure. People can acquire knowledge, communicate across borders, collaborate professionally, create art, learn languages, maintain friendships, develop software, organise communities and encounter people radically different from themselves. Open-source projects constitute an especially powerful example because they transform communication into collective production. Thousands of people who have never met can construct extraordinarily complicated systems together.
Berners-Lee has summarised his original aspiration using three words: collaboration, compassion and creativity. He argues that the Web’s early decentralised period promoted precisely these things, empowering individuals and fostering relatively small communities before centralised platforms progressively displaced them. There is a certain nostalgia in this account, but nostalgia is not necessarily mistaken merely because the world being remembered has disappeared.
There is indeed much truth in Berners-Lee’s description. Wikipedia represents perhaps the most extraordinary encyclopaedic project ever attempted. Scientific collaboration takes place across continents virtually instantaneously. A Ukrainian academic can read an American working paper minutes after publication; a British doctor can consult medical research from Japan; a student in an impoverished country can watch lectures delivered by some of the world’s most accomplished scholars. Families separated by war or migration can remain continuously connected, while small businesses can acquire international customers without possessing international offices. This is the Web that Berners-Lee loves, and he is right to remind us that it has not disappeared.
The problem is not information but incentives
Berners-Lee’s diagnosis becomes more interesting when he turns towards what went wrong. He does not principally blame objectionable content. Instead he concentrates upon architecture and incentives, because the way information is selected and delivered may be more consequential than the existence of the information itself.
Tracking is one example. Search engines, retailers and social networks construct profiles of their users. Some of this is benign or even useful. A retailer remembers what one likes; a professional network suggests people one might know; a search engine attempts to infer what one actually means. Berners-Lee acknowledges this explicitly. The moral problem begins when knowledge about an individual becomes the raw material for manipulating that person’s behaviour.
The decisive development was therefore not merely online advertising, which existed in comparatively unsophisticated forms from the early commercial Web onwards. Advertising itself long predates computers. The transformation came from the combination of behavioural surveillance, enormous quantities of personal data and automated optimisation, because together these technologies allowed commercial systems not merely to display messages but continuously to learn how individual human beings respond to them.
A traditional newspaper advertisement addresses a category of people, whereas algorithmic advertising can address something approaching an individual psychological profile. More importantly, contemporary social networks do not merely decide which advertisements users see. Their algorithms decide which pieces of reality users see, and therefore acquire a subtle influence over how users understand the world beyond their screens.
Platforms optimised for engagement have discovered, whether intentionally or through automated experimentation, that certain psychological states are commercially valuable. Anger keeps people clicking. Fear encourages attention. Sexual attraction captures it. Moral outrage encourages sharing. Tribal hostility encourages repeated participation because the opponent must constantly be defeated again. None of these emotions was created by Silicon Valley, but the technology industry learned how to measure them, stimulate them and convert them into revenue.
Berners-Lee consequently identifies engagement-optimised feeds as potentially more dangerous than surveillance alone. Facebook, Instagram, TikTok and similar services matter because their feeds are organised around systems that select material according to predictions about what will retain attention. Here his argument is particularly persuasive. The great corruption of the Web was not that human beings began saying unpleasant things upon it, because human beings have always said unpleasant things. It was that machines became extraordinarily proficient at discovering which unpleasant things each person would find irresistible.
The strange economics of free things
Yet Berners-Lee’s analysis becomes uncomfortable when one asks why this happened. The Web was deliberately given away. He has repeatedly defended that decision and explained that universality required openness. That was surely the correct decision, because had CERN patented the Web and charged licence fees for every website, humanity might have received something profoundly inferior.
Nevertheless, openness created an economic vacuum. Somebody still had to pay for servers, programmers, fibre-optic cables, data centres, search engines and increasingly sophisticated online services. Advertising became one of the principal answers, and the more accurately companies could understand their audiences, the more valuable advertising became.
From there emerged one of the peculiar bargains of modern civilisation. People acquired services of astonishing sophistication without paying money for them, while technology companies acquired something potentially more valuable: detailed information about the people using those services. The familiar observation that when a service is free the user becomes the product is slightly too crude. Users are simultaneously customers, suppliers and raw material. They receive genuine benefits. In exchange they provide attention, behaviour and information, all of which can be converted into predictions about future behaviour.
The architecture Berners-Lee dislikes therefore arose partly because the good Web required financing. That complicates any attempt to divide cyberspace neatly into virtuous and vicious territories. Google Search may simultaneously be one of history’s greatest mechanisms for democratising knowledge and one of history’s most sophisticated mechanisms for accumulating information about human curiosity. A social network may keep a grandmother in Manchester in daily contact with grandchildren in Australia while simultaneously determining, through opaque optimisation processes, which political controversies she becomes angry about. The good and bad Web are frequently not separate institutions at all but different aspects of the same ones.
Berners-Lee’s blind spot: human nature
There is another weakness in Berners-Lee’s account. He sometimes attributes too much moral agency to technological architecture and insufficient agency to the people operating within it. Long before social-media algorithms existed, human beings enjoyed gossip, scandal, pornography, conspiracy theories, ideological extremism and denunciations of their neighbours. The printing press disseminated religious hatred as efficiently as it disseminated scientific knowledge. Radio provided education and music but also became an extraordinarily effective instrument of totalitarian propaganda. Television produced documentaries and demagoguery, enlightenment and trivia.
The Web did not invent human irrationality; it industrialised its distribution. This distinction matters because it determines what sorts of remedies are realistic. If technology created the vices of contemporary public life, then changing technology might eliminate them. If technology merely amplified ancient characteristics of human psychology, then redesigning platforms can restrain those characteristics but cannot abolish them.
Berners-Lee understood the danger remarkably early. Writing in the 1990s, he asked whether the network would create a better-informed democracy or instead harbour enclaves of bigotry in which emotional intensity defeated truth. He also recognised that weakening geographical boundaries might either broaden cultural understanding or intensify polarisation. That was prescient, yet there is an important philosophical distinction between saying that technology causes human dysfunction and saying that technology amplifies psychological tendencies already present within us.
The second proposition is more plausible and considerably more disturbing. If Facebook disappeared tomorrow, envy would survive. If TikTok vanished, narcissism would remain. If X ceased operating, political fanaticism would not suddenly evaporate. Humanity would merely discover different mechanisms through which to express these things. Algorithms matter because they magnify tendencies that already exist and turn psychological weaknesses into scalable business models.
Decentralisation is not automatically virtuous
Berners-Lee’s proposed remedy follows naturally from his diagnosis. He wants to restore individual control over information. His Solid project seeks to separate personal data from the corporations providing applications, storing information in user-controlled repositories sometimes called pods. Instead of surrendering one’s digital identity repeatedly to centralised platforms, individuals could determine which applications were permitted to access which information. Berners-Lee describes this as a restoration of digital sovereignty.
The principle is admirable, but decentralisation itself should not be confused with morality. A decentralised network can distribute medical knowledge, but it can also distribute profoundly harmful material. Encryption can protect dissidents from dictatorships and criminals from police investigators. Anonymous communication can protect whistleblowers and facilitate fraud. Cryptocurrency can help people escape dysfunctional banking systems while also providing convenient mechanisms for international criminal payments.
Berners-Lee has generally resisted the proposition that technologies should be prohibited merely because they can be abused. Discussing peer-to-peer file sharing many years ago, he argued that society should ordinarily regulate wrongful conduct rather than prohibit generic technologies capable of legitimate purposes. That instinct remains sound, although it exposes the unavoidable weakness in technological libertarianism. Decentralisation distributes power, but it does not guarantee that the people receiving that power will exercise it wisely.
A dictator and a dissident both benefit from encryption, while a scientist and a conspiracy theorist both benefit from cheap publication. The moral quality of a network therefore cannot ultimately be derived from its topology. A decentralised system may be preferable because it makes certain forms of tyranny more difficult, but the distribution of power is not the same thing as the distribution of virtue.
The Web as an amplifier
Perhaps the best way to reinterpret Berners-Lee’s distinction between beneficial and harmful internet activities is therefore to abandon the attempt to classify websites morally and concentrate upon amplification. The Web is beneficial when it amplifies human capacities worth cultivating: curiosity, creativity, cooperation, friendship, scientific inquiry, education and productive exchange. It becomes harmful when it amplifies weaknesses that human beings struggle to control: rage, envy, obsession, deception, tribalism and compulsive behaviour.
This also explains why the problem cannot be solved simply by deleting the bad parts. The same fundamental mechanisms amplify both categories. Virality can spread an appeal for a dying child or a fabricated conspiracy theory. Anonymity can protect a political prisoner or a sadistic troll. Recommendation algorithms can introduce someone to Bach’s Goldberg Variations or gradually immerse them in political extremism. Artificial intelligence can translate a Ukrainian refugee’s documents or manufacture persuasive propaganda on an industrial scale. The technology has remarkably little inherent moral direction because it magnifies intention and appetite rather than supplying either of them.
Berners-Lee himself came close to this conclusion remarkably early. In 1997 he observed that networked technology might become either a great divider or a great equaliser, might homogenise cultures or polarise them and might produce either hatred or greater understanding. Ultimately those choices remained ours. That older formulation may actually be philosophically stronger than some contemporary criticism of Big Tech, because it recognises the interaction between technological structures and human responsibility rather than imagining that one can be reduced to the other.
Artificial intelligence repeats the experiment
The importance of Berners-Lee’s critique becomes particularly obvious when applied to artificial intelligence. We are now repeating many of the decisions made during the Web’s development, except at extraordinary speed. Questions about centralisation, personal data, commercial incentives, openness and control that took decades to emerge around the Web are being compressed into a few years of debate about artificial intelligence.
Berners-Lee worries that AI may reproduce the concentration of power generated by social media. He has proposed AI systems that operate for individuals rather than corporations: agents whose duties towards users might resemble the professional obligations of doctors or lawyers. He has also argued for something resembling CERN for advanced AI research, an international, non-profit scientific institution capable of pursuing knowledge outside the commercial imperatives of technology corporations.
This is characteristically Berners-Lee. His response to technological dysfunction is not technological pessimism but institutional redesign. There is something admirable about that consistency. The man who constructed perhaps the most consequential communications architecture since printing still believes that better architectures can produce better societies.
But AI sharpens the objection to his optimism. A large language model can be designed around privacy, decentralisation and user control and still tell somebody something false. An autonomous agent acting exclusively on behalf of an individual may faithfully pursue that person’s objectives when those objectives are foolish, immoral or self-destructive. Digital sovereignty does not resolve the ancient philosophical problem that sovereign individuals frequently make terrible decisions. Technology can assist humanity in managing human nature, but it cannot save humanity from human nature altogether.
The inventor’s optimism
Yet Berners-Lee should perhaps be permitted his optimism because the alternative is worse. The fashionable tendency to describe the internet as though it were a gigantic social mistake overlooks an extraordinary historical achievement. Billions of human beings have acquired access to quantities of information that emperors, presidents and professors could scarcely have imagined possessing a generation ago. Knowledge travels almost instantaneously. Borders have become permeable to ideas. A person with a telephone can publish material potentially accessible to most of humanity. The harms are real precisely because the achievement is so enormous.
Berners-Lee’s most important insight is therefore not that some websites are good and others bad. It is that technological architecture embodies values. Systems optimised for surveillance will produce more surveillance. Systems optimised for engagement will encourage behaviours that maximise engagement. Systems constructed around individual autonomy are more likely to preserve it. Engineers cannot escape moral responsibility by insisting that they merely write code, and Berners-Lee was warning his fellow engineers about this long before the contemporary language of technological ethics became fashionable.
His weaker proposition is that better architecture can substantially cure the underlying problem. It can mitigate it, certainly. Privacy can be protected. Monopolies can be restrained. Algorithms can be regulated. Children can be shielded from particularly manipulative technologies. Individuals can be given greater control over their data. Yet no protocol can manufacture wisdom, and no redesign of the Web can remove the irrationalities of the people who use it.
The World Wide Web is ultimately neither good nor evil. It is something more unsettling: a mechanism through which humanity encounters an enlarged version of itself. Our generosity becomes crowdfunding and our curiosity becomes Wikipedia. Our sociability becomes instant communication across continents. But our vanity becomes compulsive self-display, our tribalism becomes online mobs and our appetite for scandal becomes an algorithmically optimised global industry.
Berners-Lee wanted the Web to enable humanity to think collectively. In one sense he succeeded beyond anything he could reasonably have imagined. The difficulty is that collective human thought turns out to contain rather more madness, cruelty, triviality and obsession than the early architects of cyberspace might have hoped. The same network that connects the world’s accumulated knowledge also connects its prejudices, fantasies and hatreds, because universality necessarily means admitting humanity in its entirety.
The solution is therefore not to destroy the Web, nor even to imagine that we can divide it cleanly into good and bad components. It is to understand what Berners-Lee’s invention really did. It removed an enormous number of obstacles between human beings and allowed information, emotion and imagination to propagate with unprecedented freedom. That was an astonishing gift to humanity, but every gift of freedom entails responsibility for the choices made with it. What humanity chose to do with the World Wide Web was always going to be our responsibility.
