Media & Information

How News Cycles Work

A plane crash kills 200 people and dominates global headlines for weeks. Every single day those headlines are running, more than 6,000 children under five die from causes linked to undernutrition, by WHO and UN estimates, a toll that by the end of a two-week news cycle will pass 80,000. You hear about one and not the other, and the reason has nothing to do with which event matters more. News operates on a set of selection criteria that journalists call news values: novelty, conflict, proximity, scale, and human interest. A plane crash is sudden, dramatic, and unusual. Chronic hunger is slow, diffuse, and ongoing. The first fits news formats. The second does not. This is not a conspiracy; it is an economic and psychological reality. Audiences pay more attention to sudden, dramatic events, and media organizations need audiences to survive.

The phrase "if it bleeds, it leads" captures a genuine editorial tendency, but the selection process is more nuanced than pure sensationalism. Stories also get selected based on access; reporters cover events they can get to, with sources who will talk to them. Political stories dominate because political actors actively seek media attention and provide quotes, documents, and drama that make stories easy to produce. Complex policy issues get less coverage partly because they require more expertise to explain and partly because fewer readers engage with them. The result is a news landscape that systematically overrepresents conflict, crisis, and elite activity while underrepresenting slow-moving problems that affect far more people. The selection matters because it sets the public agenda. Studying undecided voters in Chapel Hill, North Carolina, during the 1968 campaign, Maxwell McCombs and Donald Shaw found that the issues voters ranked most important closely matched the issues the news covered most; they called the effect agenda-setting. Later research confirms that coverage shapes which issues people think about, while evidence on how far it changes what they conclude is mixed.

News cycles have also accelerated dramatically. In the era of daily newspapers, a story had roughly 24 hours of relevance. Cable news shortened that to hours. Social media shortened it to minutes. This acceleration means stories receive less verification, less context, and less follow-up. A misleading initial report can reach millions before corrections are published, and corrections rarely travel as far as the original claim. Speed and accuracy exist in tension, and the market overwhelmingly rewards speed.

A busy newsroom, where editors decide which of the day's events become news

The Collapse of Local News

Behind the high-volume conversation about national politics on cable and social platforms, something quieter and more consequential has been happening to local news. Northwestern's Medill School, which tracks this most closely, counts close to 3,500 American newspapers gone since 2005, almost 40% of the total. More than 270,000 newspaper jobs were lost over the same two decades. Its 2025 report counts 212 counties with no local news source at all - what researchers call "news deserts" - and another 1,525 counties down to a single one. Taken together, around 50 million Americans now live with limited or no access to local news, in places where city council meetings, school-board votes, county budget debates, and zoning decisions happen with no reporter in the room.

The collapse traces directly to what happened to the advertising business. Local newspapers historically funded themselves through classified ads, real-estate listings, and small-business advertising. Those revenue streams moved almost entirely to Google, Meta, Craigslist, Indeed, and similar digital platforms during the 2000s and 2010s. The local audience remained, but the money that paid for the journalism reaching that audience went to companies that did not produce journalism.

The consequences are substantial and slow-moving. Communities without local news have lower voter turnout, more documented government corruption (because fewer people are watching), higher municipal borrowing costs (because investors price in the absence of oversight), and increased political nationalization (because national outlets fill the attention vacuum and frame everything in national-political terms). When citizens are asked about local government, they increasingly answer using categories from national politics, because national news is most of what they consume. This is not a problem any single platform can solve, and the nonprofit, foundation-funded, and Substack-style alternatives that have emerged replace only a fraction of what was lost.

A local paper's former office, now for lease

Attention Economy

For most of media history, the business model was straightforward: create content, sell it to audiences through subscriptions, or sell audiences to advertisers. Digital media transformed this by making content abundant and attention scarce. When information is effectively free and infinite, the bottleneck is no longer production; it is your time and focus. Platforms that capture more of your attention can sell more advertising. This makes your attention the primary product being traded, and the content you see is optimized to capture and hold it, not necessarily to inform you well.

This creates perverse incentives. Content that triggers strong emotional reactions (outrage, fear, amusement, moral indignation) generates more engagement than content that is merely accurate or important. A nuanced analysis of trade policy gets fewer clicks than a headline about a politician's inflammatory remark. Platforms do not need to prefer outrage ideologically. They just need to measure what people click, share, and comment on, and then serve more of it. The result is an information environment that systematically amplifies the most emotionally provocative content, regardless of its accuracy or importance.

Creators respond to these incentives rationally. Journalists learn that provocative framing generates more traffic. YouTubers learn that controversy drives subscriptions. Politicians learn that extreme statements earn media coverage that moderate statements do not. None of these actors are necessarily acting in bad faith. They are adapting to an environment that rewards attention capture above all else. The result is an arms race of provocation that makes the information environment progressively louder and more extreme, not because anyone planned it, but because economic incentives point that direction.

Nearly every rider on a phone: attention is what the platforms sell to advertisers

Algorithmic Curation

When you open a social media feed, you are not seeing a random sample of content from people you follow. You are seeing a selection curated by algorithms optimized to maximize engagement, measured by time spent, clicks, likes, shares, and comments. These algorithms learn your preferences by observing your behavior: what you linger on, what you skip, what you react to. Then they serve you more of what you engage with. This sounds benign, even helpful. But it means two people following exactly the same accounts can see radically different content, because the algorithm has learned that different emotional triggers work for each of them.

Filter bubbles, a term coined by Eli Pariser, describe the result: personalized information environments that reinforce existing beliefs and shield you from contradictory perspectives. Echo chambers are a related concept where social dynamics further reinforce this filtering. When your feed shows you mostly content you agree with, and people in your network share mostly that same content, your sense of what most people think becomes distorted. Research on the actual extent of filter bubbles is more nuanced than popular accounts suggest. Some studies find that algorithmic curation does create measurable ideological sorting, while others find that people encounter more diverse content online than they do through offline social networks. The truth likely varies by platform, topic, and individual behavior.

What is less debatable is that algorithmic recommendation systems optimize for engagement, and engagement correlates with emotional intensity. YouTube's recommendation algorithm, for example, has been argued to progressively suggest more extreme content to users who watch political or conspiratorial videos. The argument is not that anyone programmed it to radicalize viewers, but that more extreme content may generate longer watch times. Recent research has complicated this picture: some studies find that the "rabbit hole" effect is weaker than initial reports suggested, and YouTube has changed its recommendations since 2019 in response to criticism. The broader point, however, holds across platforms. Engagement-based systems systematically favor content that grabs and holds attention, and extreme, emotional, or tribal content tends to do that more effectively than moderate, nuanced, or bridging content.

A reader at a laptop, in front of a wall of clippings joined by red string

Manufacturing Consent

In 1988, Edward Herman and Noam Chomsky published a model arguing that mass media in democratic societies functions to manufacture public consent for elite interests. Their propaganda model identified five filters that shape news coverage: concentrated media ownership, advertising as primary revenue source, reliance on official sources, organized pressure campaigns against unfriendly coverage, and a dominant ideology that frames acceptable debate (in 1988, anticommunism). The model predicts that mainstream media will systematically favor perspectives aligned with corporate and state power. The mechanism is not direct censorship but structural incentives that make challenging those perspectives professionally costly.

Parts of the model have empirical support. Media ownership is highly concentrated in many countries; the EU-funded Media Pluralism Monitor rates ownership concentration across the bloc a very high risk. Advertisers can bend coverage: a 2021 study found that US newspapers gave less coverage to safety recalls by carmakers that advertised with them. Reliance on official sources does give governments and corporations disproportionate ability to set news agendas. When a defense ministry provides reporters with embedded access, dramatic footage, and official briefings, covering war from that perspective is easier and cheaper than independent reporting from another angle. This is not censorship, but it shapes coverage in predictable ways.

Critics of the model, however, raise important objections. It tends to treat media as more monolithic than it is. Investigative journalism regularly challenges corporate and state power: Watergate, the Panama Papers, and the Snowden revelations all came through mainstream media. The model underestimates competition between outlets and the professional norms of journalism, and it predates the internet, which lowered barriers to entry. Critics on the right point to a different bias, toward the left. In a 2022 survey of US journalists, 36% identified as Democrats, 3.4% as Republicans, and about half as independents. A 2005 study that compared the think tanks cited by news outlets and by members of Congress placed all but two of the major US outlets it measured to the left of the average member. Other measurements complicate both critiques. A 2020 experiment found no partisan bias in which stories political journalists chose to cover. US newspapers' slant tracks their readers' politics far more than their owners' (a 2010 study), but when a conglomerate bought local TV stations, their coverage shifted toward national politics and to the right (a 2019 study). The manufacturing consent framework is a useful lens for understanding how incentives shape coverage, but treating it as a complete theory of how information flows in modern democracies overstates its explanatory power.

A boardroom reviews its media spending: ownership and advertising are the first two of the five filters

Misinformation vs Disinformation

These two words describe different problems. Misinformation is false or misleading content shared without intent to deceive. Your uncle sharing an inaccurate health claim on social media because he genuinely believes it is misinformation. Disinformation is false or misleading content created and spread deliberately to deceive. A state-sponsored troll farm producing fake news stories to influence a foreign election is disinformation. The distinction matters because solutions are different. Misinformation responds to better information, media literacy, and gentle correction. Disinformation responds to detection, platform enforcement, and geopolitical pressure. Treating all false information as deliberate deception alienates well-meaning people who simply got something wrong.

Disinformation campaigns have become sophisticated operations. State actors have created networks of fake social media accounts that pose as ordinary citizens, amplify divisive content on all sides of political debates, and exploit existing social tensions rather than creating new ones. Often the goal is not to convince people of a specific lie but to create confusion, erode trust in reliable information sources, and make people feel that nothing can be believed. When citizens conclude that all information is equally unreliable, they default to trusting whoever shares their identity or tribe, which is exactly what many disinformation campaigns intend.

But focusing exclusively on foreign disinformation misses a large part of the problem. Domestic misinformation, generated by politicians, pundits, influencers, and ordinary citizens, vastly exceeds foreign disinformation in volume and reach. In most democracies, the biggest source of bad information is not enemy agents but fellow citizens, prominent and ordinary, sharing content that confirms their existing beliefs without checking whether it is accurate. This is a harder problem to solve because it does not have a clear adversary to counter.

The popular image of a troll farm: one person, a basement, many screens

Parallel Realities

Something new has happened in recent decades. People living in the same city, working in similar jobs, speaking the same language can inhabit fundamentally different informational realities. They do not just disagree about policy; they disagree about basic facts. Is crime going up or down? Is the economy improving or declining? Is a particular event a crisis or a hoax? For much of the twentieth century, a small set of major news outlets in each country provided a common factual baseline that people could then argue about. That baseline has fractured.

This fracturing has multiple causes working simultaneously. Cable news channels discovered that partisan programming retained viewers better than balanced reporting. Social media made it easier to sort into like-minded clusters, though studies find that only a small, highly partisan minority lives in a true echo chamber. Trust in traditional gatekeeping institutions declined. In the Reuters Institute's 2026 survey of 48 markets, 37% of people said they trust news overall, the lowest since it began measuring in 2015. In the United States, Gallup's measure of trust in the mass media fell from about 70% in the 1970s to 28% in 2025, with Republicans at 8% and Democrats at 51%. Alternative media ecosystems emerged that explicitly define themselves against mainstream sources. And political identity became increasingly tied to information source preference; where you get your news became a tribal marker, not just a practical choice. The result is that corrections often do little when they come from a source the audience has already categorized as untrustworthy. The once-popular idea that corrections backfire, hardening the false belief, has mostly not held up; they usually shift beliefs somewhat toward the facts, when they reach people at all.

Living in parallel realities makes democratic governance extraordinarily difficult. Democracy requires enough shared understanding to have productive disagreements. You can argue about what to do about a problem only if you agree a problem exists. When citizens inhabit different factual universes, compromise becomes nearly impossible, because each side sees the other as not just wrong but delusional. Some researchers describe this as an epistemic crisis, not a crisis of values or ideology, but a crisis of shared knowledge. Solving it may be among the most important and most difficult challenges facing democratic societies, and no one has a convincing solution yet.

A protest shown two ways: a riot on one side, a peaceful march on the other

Everyone Is a Publisher Now

For most of history, publishing was expensive and access was controlled. Running a newspaper required presses, distribution networks, and capital. Broadcasting required licenses, transmitters, and studios. These barriers limited who could reach a mass audience, which concentrated power in the hands of media owners but also created a layer of editorial gatekeeping that filtered (imperfectly) for accuracy and relevance. Social media demolished these barriers. Anyone with a smartphone can now reach millions. This is simultaneously one of the most democratizing and most destabilizing developments in the history of communication.

On the positive side, marginalized voices that traditional media systematically excluded can now reach audiences directly. Citizen journalism has documented police violence, government corruption, and humanitarian crises that mainstream outlets missed or ignored. Movements that could never have organized through traditional channels, from pro-democracy protests to community mutual aid networks, have used social platforms to coordinate at unprecedented speed. Old gatekeepers were not neutral, and their removal has expanded whose stories get told.

On the negative side, the same infrastructure that lets a dissident bypass censorship also lets a conspiracy theorist bypass fact-checking. Traditional media had professional norms, imperfect but real, around verification, correction, and source protection. Social media has no comparable norms at scale. A rumor and a researched investigation look identical in a feed. Generative AI is accelerating this problem by making it trivially easy to produce convincing text, images, and video that depict events that never happened. The challenge going forward is not returning to old gatekeeping models (that is neither possible nor entirely desirable) but developing new mechanisms for quality and trust that work in a world where anyone can publish anything.

Generative AI is also changing who, or what, does the publishing. Writing tools and image generators now produce convincing articles, comments, reviews, and search-engine content at near-zero marginal cost. Some news sites, content farms, and product-review pages are now substantially or entirely AI-generated, built to capture search traffic. Bot-generated comments and replies on social platforms have become difficult to distinguish from human ones. The "dead internet theory," which began as a fringe conspiracy theory, has become a question researchers try to measure: how much of the web is now machines writing for other machines. The question shifts from "who is publishing?" to "how much of what I am reading was actually written by a human, and does that matter for whether I should trust it?" There is no settled answer. One estimate, from articles sampled across the English-language web, is that about half of new articles have been mainly machine-written since early 2025, a share that jumped after ChatGPT's release in late 2022 and has since leveled off.

Filmed on a phone, footage from a protest can reach millions without passing an editor

Deepfakes and the Crisis of Evidence

Synthetic media - face-swaps, cloned voices, fabricated video - has crossed from specialist tooling (the original adversarial networks, now increasingly diffusion models) into consumer apps, and the barrier to a convincing fake drops every year. What makes this uniquely corrosive for shared reality is not just that evidence can be fabricated; it is that the mere possibility undermines the trustworthiness of all evidence. Once any video can plausibly be faked, no video is automatically trustworthy. This is the "liar's dividend": a politician caught out can claim the evidence is fake, and some of the audience, especially supporters, will believe it. So far video has held up better than text: in experiments published in 2024, such denials helped politicians against written reports of a scandal but largely failed against video. Courts and journalism both face the same shift, from asking whether footage was edited to whether it was generated from scratch.

Detection and provenance systems (cryptographic signing at the camera sensor, content-authentication chains) are being built, but the arms race is lopsided. Each new generator can be tuned to slip past existing detectors, while detectors must keep up with every new generator. Many researchers expect detection to keep lagging, which means the long-run answer is less "is this real?" and more a rebuilt infrastructure of institutional trust in who is vouching for a piece of media. In the meantime, seeing is no longer believing, and societies built on shared evidence have not yet figured out what replaces it.

Mapping a face with a digital mesh, as face-swap tools do

What You Can Do

Media literacy is often presented as a set of simple rules: check your sources, read beyond headlines, be skeptical. These are fine as far as they go, but they underestimate the difficulty of the problem. You cannot fact-check everything you encounter. You lack the expertise to evaluate most technical claims. And your brain is running the same emotional-engagement software that platforms exploit, which means you are predisposed toward content that feels important rather than content that is accurate. Perfect media literacy is not realistic for anyone.

What is realistic is developing a few practical habits. First, notice your emotional reaction before sharing or believing something. If a story makes you feel intense outrage or vindication, that is exactly when your critical faculties are most compromised and when you should slow down. Second, diversify your information sources, not to achieve some imaginary balance point, but to develop a feel for how different outlets cover the same event. When you see how the same facts get framed differently by different sources, you develop a better intuition for where framing ends and facts begin. Third, pay attention to incentive structures behind information. Who benefits if you believe this? Who is paying for this content to reach you? These questions do not tell you whether something is true, but they help you calibrate how much skepticism to apply.

Perhaps the most important shift is accepting uncertainty. The information environment makes it harder, not easier, to know things with confidence. Being comfortable saying "I don't know enough about this to have an opinion" is itself a form of media literacy, and possibly the most useful one. In a world that rewards confident takes and punishes nuance, admitting uncertainty feels like weakness. But it is actually the most honest response to an information environment that pushes you toward certainty about things you have not had time to evaluate properly.

What you see through media is not reality but a version of reality filtered through economics, algorithms, and human psychology, and often optimized to hold your attention rather than to inform your judgment. Knowing that does not make you immune, but it does slow you down at exactly the moments when slowing down matters most. And it raises a deeper question: if this is how information reaches citizens, what does that mean for the governance systems that depend on informed participation?