Who Chooses the Conversation

By the time you encounter a headline, many decisions have already been made for you.

Someone decided the event was worth covering. Someone assigned time and money. A source returned a call while another did not. An editor selected a frame. A distribution system decided whether the result would reach your screen. You may then decide to read, dismiss or share it.

Your choice is real. So is the sequence that made this particular choice available.

A society’s conversation does not simply rise from events in proportion to their importance. It is produced by institutions under pressures. Some of those pressures improve the work: evidence, professional standards, competition to uncover something significant. Others reward repetition, haste, access to influential people or the kind of conflict that keeps an audience watching.

We need to understand those pressures without deciding in advance that every journalist is dishonest or every reader easily controlled. The question is how public knowledge gets made, and who can afford to keep making it.

Herman and Chomsky’s challenge

Edward S. Herman and Noam Chomsky’s 1988 propaganda model argued that large commercial news organizations operate through structural filters. Their original account emphasized ownership and profit orientation, advertising, reliance on institutional sources, organized pressure against disfavored coverage, and the ideological role of anticommunism in its historical setting.¹

The model directs attention to selection and incentives rather than requiring a secret instruction behind every story. It is an attributed theoretical account, not a demonstration that every report fits it or that audiences lack independent judgment. Its original ideological filter should not be silently rewritten as a claim the authors made about today’s platforms.

The useful challenge is to investigate the conditions under which some accounts of the world become routine and others struggle to obtain sustained attention. That inquiry should be open to evidence about audiences, professional autonomy, competing owners and institutional change.

I find the challenge powerful because it asks us to look beyond a journalist’s declared intention. Good intentions can coexist with a structure that regularly leaves important things unseen.

The audience is part of the institution

An owner-centered explanation is incomplete. Audiences can reward coverage that confirms their political outlook and punish coverage they find uncomfortable. An outlet trying to survive may respond to those preferences even without an owner dictating a line.

Matthew Gentzkow and Jesse Shapiro’s 2010 study of U.S. newspapers used political language in 2005 to estimate slant. Their model found that audience preferences explained a meaningful share of variation, while ownership identity explained much less. The measure concerned relative political language, not accuracy, and the print-market setting does not settle how current platforms or every media owner behave.²

Ownership can matter too. Gregory Martin and Joshua McCrain’s 2019 study compared local television stations acquired by Sinclair with others in the same markets, using 2017 transcripts. After acquisition, the stations increased national political coverage, reduced local political coverage and shifted rightward on the authors’ text-based measure. This was a historical acquisition study with a particular design, not proof of a direct effect on voting or a license to invent the owner’s private motives.³

These findings point toward a more exact account. Different institutions can be shaped by different combinations of ownership, audience demand, production costs, professional practice and distribution. A serious criticism should identify the mechanism operating in the case before it.

That is harder than dismissing the entire press with one label. It is also much harder for a powerful institution to evade.

Some stories are expensive to know

A dramatic confrontation is immediately visible. A long administrative failure may require months of work to establish.

Imagine a newsroom choosing between an easily obtained political exchange and an investigation into why workers in several businesses are repeatedly underpaid. The first offers recognizable speakers, ready-made footage and a clear conflict. The second requires records, legal understanding, careful interviews and protection against errors that could harm the people involved.

A conscientious editor may want both. A limited budget may not permit both.

Now imagine the labor story requires contact with people who fear losing their jobs. Interviews take place outside normal hours. Documents are incomplete. The employer has a communications team able to respond quickly. The workers need time to check what they can safely disclose.

A deadline treats these sources unequally. The institution with a prepared spokesperson becomes easier to quote. The person with direct experience becomes harder to include.

This is an organizational problem, not evidence that the journalist secretly dislikes workers. It can be addressed through time, specialist knowledge, source protection and funding that supports difficult reporting. Without those resources, a declared commitment to representing ordinary people may remain a sentiment rather than an ability.

The cost of discovering a fact helps determine whether the fact becomes public.

A camera finds the protest

Public action also enters this selection process. A march creates an event the media can recognize. A negotiation over a technical enforcement rule may change more lives while producing little that looks like news.

Movements therefore face a temptation to optimize for visibility. A sharp line travels. A dramatic confrontation creates footage. A careful explanation of how a procurement condition can be enforced does neither particularly well.

Visibility has value. People cannot join a cause they never encounter. A movement that refuses every form of effective communication can become a private seminar about a public problem.

But a cause can be damaged when attention becomes its principal measure of success. A campaign may learn to reproduce the grievance in increasingly vivid forms while neglecting the less visible work needed to resolve it. Leaders may be rewarded for being recognizable rather than accountable.

The remedy is to connect communication to a concrete purpose. What should someone understand after seeing the story? What can they do? Which decision is at stake? How will public attention help move that decision? What evidence will show whether it did?

The point of a microphone is to carry a demand, not to become the demand.

Distribution is a form of governance

A platform can affect whether a publication reaches readers, whether a seller reaches buyers or whether a worker finds clients. Its rules concern ranking, eligibility, payment, data and access. These decisions can alter livelihoods even when the platform describes itself as merely hosting interactions.

There are legitimate reasons for moderation and ranking. A service without protection against abuse or fraud can become unusable. A chronological feed is also a design choice with advantages and weaknesses. No arrangement is free of judgment.

The issue is whether the authority is sufficiently transparent and contestable for the stakes involved. A small business losing access to its customers needs a usable explanation and remedy. A creator whose income depends on a platform needs to understand the relevant rules. A public conversation should not depend entirely on systems whose objectives and changes are difficult to inspect.

This does not require disclosing every technical detail in a way that enables exploitation. It requires meaningful accountability: clear policies, notice of consequential changes, an appeal suited to the harm, independent scrutiny and practical alternatives where possible.

A platform’s scale can make its choices socially important without making every choice a matter for direct government command. We need institutions capable of drawing that distinction carefully.

Exposure does not equal persuasion

A system that changes what people see does not necessarily change what they believe. The empirical effects require study.

A large preregistered experiment during the 2020 U.S. election reduced like-minded content for consenting Facebook users over three months. The intervention changed exposure but found no measured change in eight preregistered attitude outcomes. The study, published in Nature in 2023, involved collaboration with Meta and disclosed academic control over analysis and publication. Its bounded null result does not show that every algorithm or longer-term pattern is politically harmless.⁴

A separate Nature study published in 2026 examined an experiment on X conducted in 2023. Turning on the algorithmic feed for users previously using a chronological one shifted selected policy and current-event opinions in a conservative direction and increased engagement. Turning the algorithm off did not show a corresponding reversal. Neither switch significantly changed party identification or affective polarization. The study was specific to its users, period and intervention; it does not establish the effect of the platform’s 2026 system or anyone’s intent.⁵

Together these studies counsel precision. Algorithms can change exposure; some interventions affect some attitudes, while others do not show those effects. We should neither assume irresistible manipulation nor use one null finding to absolve every design.

The democratic case for oversight is broader than persuasion alone. Access, visibility, harassment, economic dependence and the ability to challenge an error can matter even where a measured political attitude does not move.

The expert and the outsider

A working-class critique of institutions must resist an easy trap: treating all expertise as a disguise for domination.

Powerful organizations can use expertise selectively. They can fund favorable questions, promote convenient measures or hire specialists to make a weak argument appear inevitable. These practices deserve examination.

The answer is better evidence and more independent expertise. A person selling an unsupported claim does not become trustworthy because they present themselves as an outsider. A wealthy media personality can perform rebellion while operating a business that rewards the same attention-seeking habits they denounce.

The relevant questions remain concrete. What is the claim? What evidence supports it? What would count against it? Who benefits from my accepting it? Is the person willing to correct an error in public?

Credentials can help identify relevant competence, but they do not replace scrutiny. Lack of credentials can coexist with valuable experience, but it does not remove the need for evidence. A worker may understand a production problem that a consultant has missed. That knowledge should enter the inquiry on its merits.

Democratic knowledge is strongest when experience and expertise can challenge and improve one another. It weakens when either side is declared inherently incapable of learning.

Follow the funding without pretending it answers everything

A funding relationship creates a question, not an automatic verdict.

If a study is paid for by an interested party, examine its methods, disclosures, access to data and publication rights. If a newsroom depends on a donor, ask what protects editorial independence. If a public broadcaster is funded by government, ask what prevents ministers from using the budget as a threat.

Different sources of money produce different vulnerabilities. Commercial revenue can reward audience capture. Advertising can create sensitivities around major clients. Donations can concentrate influence in a few patrons. Public funding can become vulnerable to political retaliation. Subscriber funding can exclude people who cannot afford it and reward a narrow audience’s worldview.

There is no immaculate funding stream. There are designs that distribute risk, disclose dependencies and protect the work from retaliation.

A durable information institution may need diverse revenue, governance that separates funding from editorial decisions, transparent corrections and independent review. Local and public-interest reporting may require forms of shared support because its social value is difficult to capture through individual purchases.

We should be ambitious about providing that support and exacting about the safeguards. Public money should not buy the government’s preferred story. Private money should not buy immunity from investigation.

An information institution earns trust partly by making its own vulnerabilities visible.

A culture of correction

Correction is often treated as an embarrassment to be minimized. It should be a normal sign that an institution takes truth seriously.

That does not mean every error is trivial. A false allegation can do lasting harm. A correction that appears long after a widely shared claim may not reach the same audience. Repeated errors in one direction can reveal a deeper problem.

A serious correction practice should say what was wrong, what the evidence now supports and how the process will change. It should be easy to find. Where feasible, the correction should travel through the same channels as the original claim.

Readers have obligations too. If we share a claim because it confirms our anger, we should be prepared to share the correction with equal visibility. Otherwise, our criticism of institutional dishonesty becomes selective.

This matters politically because movements need a reputation that survives disagreement. An organization able to admit a mistake can ask others to trust its next well-supported claim. An organization that treats correction as betrayal teaches its members to choose loyalty over reality.

Working people have too much at stake to organize around convenient falsehoods.

An ordinary reader’s independence

You cannot investigate every statement you encounter. A demand that everyone become their own full-time fact-checker is another way of transferring institutional labor onto individuals.

You can, however, develop a few habits that make you less easily steered. Slow down before spreading an emotionally satisfying allegation. Distinguish a report of an event from a claim about its cause. Notice when a headline’s certainty exceeds the evidence described beneath it. Seek the relevant primary record for a consequential claim when it is available.

It also helps to read people who disagree without making disagreement itself a mark of wisdom. A contrarian can be just as dependent on an audience as a conventional commentator. What matters is the quality of the argument and the willingness to engage contrary evidence.

More importantly, support institutions that do the work individuals cannot sustain alone. A local reporter who follows a public contract for months may provide more democratic value than another hour of national outrage. A union publication that explains an agreement clearly may equip workers better than an endless sequence of inspirational messages. A public data office can make scrutiny possible for many organizations at once.

Independence is partly a personal practice and partly a shared infrastructure.

Make room for a different conversation

Imagine a public conversation in which the conditions of ordinary work are treated as a continuing subject of serious reporting. A conversation in which a policy is followed through implementation, and the people living with its consequences are heard after the announcement has faded.

Imagine fewer debates in which one participant supplies a moral claim and the other supplies an unexplained assertion that the economy will not permit it. Both should have to show their work. The aspiration deserves a mechanism; the claimed constraint deserves evidence.

Such a conversation would still contain conflict. It should. People disagree about values, risks and priorities. The aim is not a frictionless consensus. It is a public capable of seeing the choices clearly enough to govern them.

The powerful will continue to speak. Their knowledge and preferences should be heard. They should not be allowed to define the entire range of sensible thought merely because they can afford to be present every day.

Ordinary people need more than a chance to react to someone else’s story. They need institutions through which their experience can help determine what the story is about.