Preface: communication studies, part 2.

1) People Live in Their Imagined World

Before the late nineteenth century, people had long communicated in many ways: sermons, royal proclamations, rumor, word of mouth. None of this was yet communication studies in the modern sense. It lacked one key precondition: a mass mechanism of influence that could operate at scale, be used again, and produce effects that were statistically predictable.

Until then, the effect of any message depended heavily on its setting. Whether a sermon worked depended on the preacher's standing, the sanctity of the ritual, and the faith of the listeners. Whether a royal proclamation was carried out depended on force, hierarchy, and local power, not on persuasion. Rumor moved along unstable paths, at uncontrollable speed, and its content changed as it spread. These forms were not without influence, but they could not be broken down into stable variables or tested again in a different setting. No one could systematically answer the question: if the same message, built the same way, were delivered again in another city to another crowd of strangers, would it have a similar effect? Without that repeatability, theory was nearly impossible.

The Industrial Revolution changed not only how goods were made but how information related to the structure of society. Urbanization took people who had lived within networks of acquaintances and packed them into dense, lasting environments full of strangers. In the countryside, people understood the world mainly through those they knew. In the city, huge numbers of people who had never met read the same newspaper on the same day, saw the same poster, and were angered or reassured by the same story. Newspapers, the telegraph, and cheap printing did more than speed things up. For the first time, information existed apart from any personal relationship. You didn't have to know the reporter or any other reader, yet the same story could reach thousands of people at once.

A new object came into view: the masses. Neither a community nor an organization, the masses were a collection of strangers who could be triggered, psychologically, at the same moment. Under these conditions, communication began to show patterns that could be observed and compared. Similar message structures, narrative frames, and emotional cues tended to produce similar group reactions in different places. Fear, anger, nationalist feeling, and moral panic began to appear across many people at once. For the first time, communication became something that had to be studied systematically.

The questions this raised were not abstract. They bore directly on government, business, and war. If the masses could be triggered all at once, who was setting the triggers? Why did some narratives spread quickly while others were ignored? Why did the same facts, presented differently, produce completely different social reactions? Strikes, riots, revolutions, and financial panics were repeatedly found to be closely tied to how information spread. Information was no longer just reflecting reality. It had begun to take part, structurally, in producing it.

So the origins of communication studies are not romantic. The field did not arise because humanity finally wanted to understand itself. It arose from the anxiety of states, capital, and ruling groups about masses slipping out of control. Its starting point is not how long communication has existed, but the moment society clearly recognized that public opinion could be shaped systematically, and that if this force were not understood, it could turn against order itself. From then on, communication was no longer a side effect of culture. It became a force to be studied, modeled, and controlled. From the beginning, communication studies was bound up with power and fear.

One of the first people to see this clearly was Walter Lippmann. Born in 1889, he came of age as the United States industrialized rapidly and its media became highly concentrated. He had rigorous training in philosophy and politics and spent many years as a political columnist for major publications. This matters: he did not imagine "the public" from a library. He watched, again and again, how the public was pulled, mobilized, and misled by information in real newsrooms, real political fights, and real war narratives.

The First World War was decisive for him. During the war, the U.S. government mobilized public opinion on a massive scale, manufacturing stories of friend and enemy, emotional consensus, and moral justification, and Lippmann himself was part of this. He worked in the War Department, took part in the background research for Wilson's Fourteen Points, and later served as an army officer in France producing propaganda aimed at the enemy. He saw firsthand how a democracy that believed itself rational could be led by a single narrative in a very short time. The experience did not excite him. It alarmed him. He realized that modern society had become far too complex for ordinary people to understand through direct experience.

In 1922 he published Public Opinion, which introduced the much-cited concept of the pseudo-environment. The concept doesn't accuse the press of lying, nor does it belittle the public's intelligence. It points to a structural fact: reality has grown too large for any individual to experience directly. Politics, economics, war, finance, and international relations mostly happen far outside the radius of personal life, yet people still have to form judgments about them. So they live in a mediated reality. Headlines, photographs, statistics, story frames, and moral labels are not reality itself but versions of it that have been compressed, cut, and formatted. People act not on the world as it is but on their picture of it, and the material for that picture is continually supplied and reinforced by the media. The pseudo-environment is not reality, yet in practice it takes reality's place.

The crucial point is that none of this requires a conspiracy. Even if every journalist were perfectly honest, the pseudo-environment would still be unavoidable, because news as a form can only show fragments of reality, while public affairs demand systematic understanding. This is cognitive compression, not simply distortion. From here Lippmann reached a cold judgment: stereotypes are not a moral failing but a cognitive tool. In a complex world, people have to sort things into categories before they can quickly find where they stand. The question is not why people use stereotypes, but who supplies them, how they are reinforced, and what structure of power they serve. The public is not stupid. It is adapting rationally to a world it cannot fully grasp. The real danger to democracy is not that the public lacks reason, but that institutions pretend the public is judging the real world.

If Lippmann offered the diagnosis, Edward Bernays carried out the treatment. He was not purely a theorist but someone who worked inside the machinery of state and corporate opinion-making. During the First World War he worked for the Committee on Public Information, the U.S. government's wartime propaganda agency. Afterward he served corporations and governments as a public relations counsel. In the early 1950s, for example, he was hired by the United Fruit Company to build public legitimacy in the United States for action against Jacobo Árbenz's government in Guatemala. That campaign ended in the CIA-backed coup of 1954. His significance is not that "one man toppled a government," but that he showed how shaping public opinion became part of how modern power operates.

His three major works trace how public relations moved from a profession justifying itself to a full institutional capability. Crystallizing Public Opinion (1923) was nearly the first systematic theory public relations offered of itself. In it, Bernays denied that public opinion forms naturally and argued that it has to be "crystallized." This did not mean inventing facts. It meant organizing scattered, vague, emotional social attitudes, bringing them into focus, and giving them direction. The book still has a technician's restraint. It describes public relations as professional work that coordinates social relationships, and it tries to maintain a stance of neutrality and public interest. It is his mildest work.

Propaganda (1928) changes tone sharply. If the first book explained "what we do," this one openly declared "this is the reality of democratic society." Bernays wrote plainly that modern democracy cannot do without propaganda, and that the real question is not whether to manipulate but who manipulates and how. Guidance and manipulation are recast as necessary conditions for democracy to function. The danger of the book lies not in its moral stance but in its frankness. Here communication is understood explicitly as a technology of power, no longer an art of communicating.

After the Second World War, The Engineering of Consent went further. Bernays was no longer concerned with justification but with running the system. The phrase "engineering of consent" itself makes clear that public attitudes were now seen as something to be designed, tested, adjusted, and maintained over time. He likened opinion-shaping to an engineering project, emphasizing process, expert collaboration, and long-term planning. Communication was thus upgraded from a strategic act to an institutional capability, anticipating the full arrival of political consultants, campaign machines, and corporate branding systems.

This is where the two men part ways. Lippmann asked what this structural break meant. Bernays asked how power should operate given that the break could not be closed. Bernays did not misread Lippmann. He quoted Public Opinion directly in Crystallizing Public Opinion and turned Lippmann's diagnosis into an operating logic. In that sense, Bernays was Lippmann's most faithful and most dangerous heir. Lippmann offered no solution. What he did was keep dismantling the illusions of modern society: a rational public, a transparent press, a democracy that runs on its own, an international order grounded in morality.

Traced from the end of the First World War to the present, communication studies does not follow a line of steady intellectual progress. It has been forcibly rerouted, again and again, by the demands of power, the conditions of technology, and the structure of media. Each new form of communication has forced the field to rewrite what it studies, whom it serves, and what kind of creature it assumes people to be.

From the end of the First World War through the Second World War and the early Cold War, American communication research developed a strongly engineering-minded orientation. It had no unified theory at first. Driven by the practical needs of propaganda, opinion management, and attitude research, it gradually converged on one core question: can information be delivered effectively and produce statistically predictable effects? Propaganda analysis, attitude-change research, and effects measurement were the main concerns of the period. The masses were assumed to be something that could be stimulated, influenced, and adjusted. Media were treated as relatively neutral pipelines, and communication was broken into separable, controllable inputs and outputs. This view fit neatly with wartime and early Cold War state propaganda, psychological warfare, and domestic governance, and was reinforced as a result.

Harold Lasswell's contribution was not a specific theoretical model but the basic framework of questions for the field. His 1948 formula, "Who says what, in which channel, to whom, with what effect," broke communication down into a process whose endpoint is effect. In this framework, communication is first of all an act of influence, not an interaction. What matters is not whether meaning is shared but whether influence occurs. The audience appears only as the thing acted upon, and it matters only insofar as some measurable change results. Understanding is not the key variable. Change is. Once communication is modeled this way, dialogue has already been pushed to the margins of research, because dialogue means both sides might be changed, while an effects-oriented framework assumes change runs in one direction only.

At a deeper level, the model assumes that meaning is complete before communication begins. Meaning is treated as an object that can be encoded, transmitted, and delivered, not as something negotiated through interaction, dependent on context, and liable to drift. As long as the sender chooses the right symbols, channel, and timing and triggers the intended reaction in the audience, communication counts as a success. Negotiation, ambiguity, and uncertainty are not the focus of study but variables to be compressed and managed. The very concept of "effect" is therefore strongly normative. Communication is understood as an intervention with a predetermined goal, not as an open public process. As long as the goal is considered legitimate, whether the audience truly understands doesn't matter. This approach is highly practical for political propaganda, advertising, and psychological warfare, but it cannot explain any social process that requires public discussion, the negotiation of meaning, or collective judgment.

By the middle of the twentieth century, this assumption faced its first systematic challenge. Election studies beginning in the 1940s showed that the masses were not entirely passive and that information did not always have direct effects. Interpersonal networks, opinion leaders, and social structure clearly mediated and buffered communication. The approach later summed up as "limited effects" did not deny that communication had influence. It pointed out that influence is not linear, immediate, or all-powerful. On the surface, the field became gentler and more sociological, but its underlying goal did not fundamentally change. It did not give up on influence. It pursued influence by more indirect and more structural means. Communication moved from "acting directly on you" to "acting on you through the people and surroundings you trust."

At almost the same time, the medium itself entered the field's view. When Marshall McLuhan declared in the 1960s that "the medium is the message," he fundamentally challenged the field's existing assumptions. Before him, communication studies largely took for granted that the same content carried by different media was essentially the same message, differing only in efficiency and reach. McLuhan rejected that premise. For him, different media are not different packaging for the same message. Each introduces an entirely different structure of perception. What changes a society over the long term, deeply and irreversibly, is not what you say through a medium but how the medium itself changes the way people understand the world and relate to one another.

"The medium is the message" does not deny that content matters. It points out that a medium's most important effects happen before content is understood. By changing how attention is distributed, the sense of rhythm, the modes of participation, and the balance of the senses, a medium shapes in advance how people are able to understand and how far that understanding can go. By the time someone starts deciding whether to believe a piece of content, the structure of their perception has often already been set. In today's platform environment, this judgment looks especially prescient. When a medium is no longer just a tool but an environment people cannot leave, the power of communication lies less in any particular persuasive content than in the medium's capacity to rewrite a way of life.

2) Communication in the Age of TikTok

In the age of the internet and platforms, communication studies has not thrown out its existing theories, but its core assumptions have clearly begun to fail. Communication is no longer linear. It is networked, recursive, and heavily dependent on feedback. Agenda-setting, framing, and the spiral of silence did not originate in this period, but they have to be rethought under new conditions. When platforms control visibility, ranking, and distribution, the question of who holds the power to communicate is no longer simply a question of content.

In environments shaped by short-video platforms, the focus of research has shifted again. The question is no longer only how information persuades people, but how attention is continually shaped. Content is not actively chosen but systematically pushed. The agent of communication has become blurry. The algorithm does the filtering and distributing, yet it is neither visible nor easily held accountable. The basic unit of communication has shifted from information to behavioral feedback. Watch time, rewatches, and interaction count for more than understanding itself.

In this structure, the goal of communication moves from changing attitudes to shaping habits. The platform doesn't need you to believe any stable framework of explanation. It only needs you to settle into predictable patterns of use. Power is exercised less through consistency of discourse than through control of rhythm: when you watch, for how long, and in what emotional state. The main arena of communication slides from public reasoning to attention management, from producing meaning to occupying time. Different, even conflicting, narratives can coexist in the same feed as long as each one holds attention. Ideological consistency is no longer necessary. Occupying cognitive bandwidth is enough to keep the system running.

When communication runs on rhythm rather than meaning, public discussion faces a structural dilemma. People are not won over by a particular viewpoint. They are reshaped over time by a way of taking in information. Opinions can change often, but once patterns of attention, thresholds of patience, and rhythms of participation become fixed, the range of what people can understand and think through seriously shrinks with them. Occupying time ultimately becomes drawing the boundaries of what can be thought. The power of communication has completed a shift: from persuading people what to believe to determining how people exist within the media environment.

3) The Filtering System of TikTok

Following this line of thought further, Douyin is better understood as a real-time behavioral filtering system than as a content platform. What matters is not how the platform evaluates content itself, but how it judges whether a piece of content can reliably trigger a certain kind of user behavior. In this framework, a video is more like a probe. It tests whether, within a particular structure of users, a predictable and repeatable behavioral response appears. The system doesn't care what view you express. It cares only whether that expression produces a statistically stable behavioral signal.

Judging from the distribution patterns visible from the outside, a new video is usually first shown to a small but non-random sample of users. These users are more likely to have responded clearly to similar content or similar types of stimulus. This stage is less a recommendation than a stress test. The system uses very early behavioral indicators to judge whether it is worth investing more attention resources. It looks at the most basic cold-start signals, such as whether users stop scrolling the instant the video appears and whether they keep watching for a few more seconds. These signals are not the same as full views or interaction, but they decide whether the video meets the minimum bar for the next round of testing. If they fall clearly below the baseline, more complex data usually isn't examined closely at all.

The process, then, is closer to passing a series of gates than to receiving an overall score. The system doesn't need to decide whether a video is good overall. It only needs to keep checking whether the video meets the conditions for the next stage. Once key signals fall short, distribution tends to stop abruptly rather than taper off. This explains why some videos with decent early numbers suddenly stop spreading. One direct consequence is that the early sample carries far more weight than later performance. A video's fate depends mainly on the reactions of a small initial group of users, not on proving itself slowly over time. Once the system judges that a stimulus has no statistical value for expansion, even content with real potential has a hard time getting shown again.

After a video passes the initial test, distribution doesn't simply scale up. It more likely shifts to differentiated testing: the video is shown to user groups with different structures and higher thresholds to see whether the response still holds. At this point the system cares less about the number of responses than about whether the response carries over between different user structures. If a stimulus works only among low-threshold users and fails quickly among pickier users or users whose actions cost more, the system tends to classify it as a pattern that holds only locally and limits how far it spreads. This kind of content isn't removed, but it is usually confined to a fairly narrow band of distribution. Conversely, if content reaches only a modest audience but reliably triggers high-cost actions, such as comments that take a clear stance or shares, the system tends to treat it as a signal with clear positioning value. This doesn't mean the content is better. It means the content has formed a stronger bond with a certain kind of behavior. For the platform, a stimulus structure that can be reused and predicted has more engineering value than content that is merely "pretty good overall."

At this level, the system's core concern is still not what the content means but whether it reliably produces a change of state. The platform doesn't need to fully understand the logic of a viewpoint. It only needs to identify which users, under which stimulus, move from one pre-behavioral state to another. Emotion here is not the goal but an intermediate variable that can be recorded and put to use. Once the mapping from content to change of state to behavioral response becomes stable, the video is no longer something being evaluated. It becomes more like a tool with a known use. From then on it is shown continually to the users most likely to complete that behavior and is rarely used to explore entirely different user structures. This is less a naturally forming echo chamber than the result of precision targeting.

Over a longer time scale, the focus of evaluation gradually shifts from individual videos to the account as a whole. The account is treated as a fairly stable source of stimulus, and each video as a sample from that source. The system no longer asks "how is this video?" but "what kind of behavioral response does this account usually produce?" Once an account shows a statistically stable pattern of output, distribution relies more on prediction than on fresh testing. Content that fits the established pattern starts faster and gets more room for error. Content that clearly departs from the account's usual type of stimulus may get fewer chances to be tested because its uncertainty is too high. This is why changing an account's direction is so costly in practice. It isn't that viewers can't accept it, but that the system has to pay an extra cost to learn the account again.

By this logic, the platform has no inherent preference for so-called positive or negative content. It prefers content whose response curve is highly predictable. A response curve can be understood as a recordable time series: at which second the user stops, when an emotion arises, when an action is taken. The system doesn't judge the values the curve expresses. It judges only whether the curve is stable enough to be predicted accurately next time. Accounts with a clear stance, a stable structure, and a consistent pattern of output are more easily trusted by the system. Expression that is vague, wavering, or constantly trying out different kinds of stimulus is harder to model and amplify. At the engineering level, trust rests on stability, not on morality or viewpoint.

Seen this way, Douyin is more a filter than a stage. Creators are not simply expressing themselves. They are being tested, continually, on how stable a stimulus they provide. Content that keeps spreading is usually not deliberately pandering. It has simply, without anyone intending it, become highly compatible with the filtering mechanism. 📖