Preface: Continuing from the previous piece. This article was co-written with ChatGPT.

Edward Bernays (1891–1995) was born in Vienna, Austria, and emigrated to the United States with his family at a young age. He was the nephew of Sigmund Freud, the founder of psychoanalysis. This familial connection is not a mere anecdote but a crucial background to the formation of his thinking. Bernays did not receive rigorous academic training; instead, he developed through work in journalism, propaganda, and government. During World War I, he participated in U.S. government war-propaganda efforts, directly witnessing and practicing how a democratic society could be mobilized to support war through media and symbolism. After the war, he systematized these experiences and went on to serve corporations, governments, and political forces, gradually translating “propaganda” into the more neutral-sounding term “public relations,” while personally laying the professional structure and methodology of the field. His life spanned World War I, the interwar period, and the early Cold War. He was both a witness to the full formation of mass society and an active shaper of it. Bernays did not see himself as a manipulator or conspirator; rather, he consciously positioned himself as a “social engineer,” attempting to answer what he regarded as an unavoidable question: once mass society arrives irreversibly, how is order to be maintained?

What Bernays called his “cold premise” was not an emotional value judgment but a structural one. In Propaganda, he repeatedly emphasized that he was not accusing “ordinary people” of stupidity, but describing the structural conditions confronting mass society. The scale of information in modern life far exceeds any individual’s capacity to process it; finance, diplomacy, technology, and law are all highly specialized domains, and no individual can realistically “figure everything out” alone. The problem, therefore, is not human ability but the fact that structure itself determines that judgment must be outsourced and delegated to intermediaries.

On this basis, Bernays drew a fundamental distinction between individual rationality and collective behavior. Drawing directly on crowd psychology, he held that while individuals may remain rational under certain conditions, once they enter large-scale groups their behavior rapidly becomes emotional, imitative, and polarized, driven by symbols, slogans, and identity. His concern was not moral evaluation but practical consequence: at scale, collective decision-making is more likely to spiral out of control than to converge on truth.

These risks were further amplified as democratic institutions expanded. From the nineteenth to the twentieth century, suffrage broadened, mass media spread, and urbanization and industrialization advanced simultaneously. Bernays argued that small-scale societies could rely on custom, familiarity, and informal constraints to maintain order, but mass societies could not depend on such “spontaneous order.” Without a unifying narrative and direction, society would fall into emotional cycles, fragmented agendas, and uncontrolled mobilization.

It was against this background that Bernays introduced the concept of the “Invisible Government.” This concept is easily mistaken for a conspiracy theory, but he made clear that it did not refer to secret meetings, black-box manipulation, or a single organization. Rather, it described a functional layer that inevitably emerges in mass society—not something deliberately designed, but something formed under structural pressure.

Bernays spoke very directly about the composition of this “invisible government.” It is not made up of politicians, but of a network of roles: media editors and gatekeepers who decide what counts as important news and how it is ordered, timed, and framed; experts, scholars, and think tanks that package value judgments as “professional conclusions,” providing the appearance of neutral authority; public-relations professionals and communications strategists who design narrative frameworks and connect interests with media systems; and opinion leaders and social symbols that translate abstract issues into perceptible images and offer models for imitation. These actors do not need to conspire; they only need to operate according to the same logic.

The power exercised by this structure does not lie in traditional state instruments such as law, police, taxation, or the military, but in control over the boundaries of cognition. Traditional government regulates behavior; the Invisible Government regulates the range of issues, acceptable modes of expression, and the spectrum of “reasonable” conclusions. It does not command, “You must think this way,” but ensures that alternative ideas are either invisible or not treated as serious options.

In Bernays’s view, this structure is unavoidable because its logic is closed: the public cannot process highly complex affairs; judgment must therefore be mediated; once mediation exists, it inevitably forms a structure; and that structure will either operate unconsciously, producing chaos, or be consciously designed to maintain order. Bernays clearly chose the latter. His concern was never whether manipulation should occur, but who should carry it out and according to what principles.

This leads to one of the coldest layers of his position. Throughout his writing, Bernays repeatedly suggested that transparency does not automatically produce rationality; it may instead generate panic, polarization, and systemic paralysis. For him, fully realized “public self-determination” was a dangerous illusion, and democracy had to be engineered in order to function within mass society.

What makes this logic genuinely unsettling is not whether it is “evil,” but its explanatory power, its close fit with modern social conditions, and its repeated effectiveness in practice. This is why Cold War psychological warfare, contemporary political communication, and platform-based algorithmic recommendation systems all align almost naturally with Bernays’s framework.

In compressed form: the Invisible Government refers to the layer of structure in mass society that determines how people see the world. It requires no additional legitimacy, because it decides in advance what counts as a legitimate question—what is worthy of serious discussion.

Bernays’s core judgment was that the systematic and engineering-based management of public opinion has not disappeared; it has been fully inherited and automated by contemporary platform algorithms. He designed a blueprint for “opinion engineering,” while today’s platforms have converted that blueprint into technical systems capable of real-time operation and massive scale. Human beings have not changed, nor has social structure in any fundamental sense; what has changed is the execution layer—from manual human operation to continuously running machines.

Within Bernays’s framework, the first step is always agenda selection. The public is not told directly what to think; it is confined within a pre-filtered range of discussion. Today, this function is performed by trending lists, recommendation feeds, and ranking systems. Platforms do not issue conclusions; they continually decide what is worth being seen. Content that is not pushed or displayed effectively ceases to exist at the public level.

The second key structure is third-party authority. Bernays emphasized that the public does not trust direct stakeholders, but trusts voices that appear neutral. Experts, scholars, journalists, and opinion leaders are therefore used to provide a credible shell for a given narrative. Today, this logic operates more efficiently through verified accounts, data-driven rhetoric, research findings, and creator ecosystems. People do not trust platforms themselves; they trust the intermediaries selected and labeled as “credible” by platforms.

The third layer is the priority of emotion over fact. Bernays never expected facts alone to drive public behavior; he stressed that fear, anger, shame, identity, and moral superiority are the true forces of mobilization. In the algorithmic era, this principle has been fully quantified. Click-through rates, shares, and dwell time consistently demonstrate that what spreads most easily is not information, but emotional structure. Algorithms do not judge truth; they amplify whatever most readily triggers reaction.

Next comes control of repetition and rhythm. When the same narrative appears repeatedly across different accounts, from different angles, and at different times, it gradually comes to be perceived as “social consensus.” This effect does not rely on persuasion, but on frequency of exposure. Where Bernays once had to coordinate editors and media timing, algorithms now accomplish this repetition at the millisecond level.

The decisive step is what Bernays called engineering. He treated public opinion as something that could be designed, tested, and adjusted, limited only by the tools of his time. Today, A/B testing, behavioral feedback, and real-time optimization have fully integrated public opinion into engineering systems—measured, compared, and continuously tuned. Opinion is no longer merely the result of guidance; it becomes a process of parameter adjustment.

Within this structure, the Invisible Government undergoes a change of form. Where it once consisted of editors, experts, and public-relations professionals, it now manifests as recommendation algorithms, ranking rules, and distribution mechanisms. It governs not by command, but by continually pushing individuals toward paths of least cognitive resistance.

Finally, it must be emphasized that whether platforms are “evil” is not the key issue. What truly inherits Bernays’s legacy is not subjective intent, but structural position. The core function of platforms is not to promote a particular stance, but to optimize what is easiest to believe, to share, and to align with. In this sense, Bernays invented the methodology, and algorithms completed its depersonalization and scaling.

Whether algorithms and “genuine public reason” can coexist admits of a clear answer: under current platform structures, they cannot; only if the structure itself is fundamentally altered does coexistence become theoretically possible. The problem is not that algorithms are “not good enough,” but that the objectives they are designed to optimize structurally conflict with the conditions required for public reason.

To discuss this, public reason must first be defined. In political communication and public deliberation, public reason does not mean that “everyone is smart.” At minimum, it requires delayed judgment rather than immediate emotional response; the absence of structural concealment of key information; the continued translatability and mutual intelligibility of opposing positions; and the containment of emotional noise. In other words, public reason does not demand more rational individuals, but a system that does not punish rational behavior.

By contrast, the true objective function of platform algorithms is not to improve the quality of public judgment, but to maximize engagement, dwell time, and behavioral predictability. Under this function, algorithms naturally select and amplify content that is emotionally intense, quickly digestible, sharply positioned, and easy to take sides on. Public reason exhibits the opposite traits: it is slow, complex, ambiguous, and cognitively costly. This is not algorithmic bias but a clear engineering outcome—under current objectives, rationality is a losing strategy.

A common misconception is that “smarter algorithms” could solve this problem. The issue, however, is not algorithmic intelligence, but the fact that algorithms can only optimize what is quantifiable and immediately fed back. The most important dimensions of public reason are long-term, counterfactual, and not instantly verifiable—for example, whether a judgment remains valid years later, whether it reduces social polarization, or whether it enhances cross-group understanding. These metrics cannot enter real-time feedback loops and are therefore almost invisible to algorithmic systems. Algorithms do not reject rationality; they structurally cannot see it.

From this perspective, the algorithmic age has not departed from Edward Bernays’s logic, but upgraded it. In Bernays’s time, consensus manipulation relied on editors, experts, public-relations professionals, and rhythm control. Algorithms completed two key leaps: the automation of opinion engineering and the real-time feedback of emotion. Even if platforms remain subjectively neutral, as long as they optimize engagement, they will repeatedly “discover” that emotion outperforms rationality. This is not conspiracy, but the natural result of reinforcement learning.

Accordingly, coexistence between algorithms and public reason is possible only if structural conditions change. Algorithmic objective functions would need to be rewritten to incorporate delayed evaluation, long-term stability, and de-polarization, so that rationality is no longer automatically disadvantaged. Systems would need to deliberately introduce “friction,” such as delayed sharing, reduced propagation speed for emotional content, and forced exposure to opposing views. Public reason requires slowness; slowness is technically feasible but directly conflicts with commercial efficiency. Finally, some portion of judgment must return to human institutions—editorial judgment, public bodies, and non-market agenda setting. Rationality does not emerge from algorithmic self-discipline; it depends on institutional protection.