Let the crowd deliver your message.
Asch, chapter 8. Around three quarters of participants conformed at least once to a unanimous majority giving an obviously wrong answer about the length of a line, and conformity ran at roughly a third of critical trials.
Deutsch and Gerard separated the reasons: informational influence, where the target concludes that so many people are unlikely to be wrong, and normative influence, where the target knows the majority is wrong and complies to avoid being the deviant. Normative influence is stronger in public, which is the relevant condition online.
The operational point is that the persuasive force is prevalence rather than argument. The message does not need to be good. It needs to appear widely held.
Which is why fabricating prevalence is cheap and why chapter 238 exists. The Internet Research Agency's method, as set out in the February 2018 indictment, was to operate accounts posing as Americans — activists, veterans, church groups, secessionists — at a scale sufficient to constitute an apparent distribution of opinion. A reader cannot audit account authenticity. Nobody can sample public opinion directly; everyone infers it from what they see.
Asch's own most useful finding is the operator's constraint. A single visible dissenter reduced conformity by around three quarters, and it did not matter whether the dissenter was correct — an ally who gave a different wrong answer freed the subject just as effectively. What the technique requires is not a majority. It is unanimity, or the appearance of it.
Which means the manufactured version has a characteristic signature: it is too clean. Genuine consensus among many people is untidy and contains complaints, qualifications and people who agree with reservations. A synthesized one does not, because nobody built the dissent.
The boundary is where manufactured prevalence enters democratic processes. Astroturfed comment campaigns to regulators, fabricated constituent correspondence, coordinated inauthentic accounts in an election — these corrupt a signal that institutions rely on to determine what people actually want, and several are criminal offenses.
The commercial version — purchased reviews, fake testimonials, undisclosed influencer relationships — is subject to enforcement in most jurisdictions and is the subject of chapter 253.
The counter is chapter 306, and it is also the cheapest contribution anyone can make: be the visible dissenter.
Countered by The One Dissenting Voice, Trust Inside the Tribe