The Shared Record: What Democracy Can’t Survive Without

American democracy does not require consensus, but it does require a shared understanding of current events. This foundational condition is deteriorating. Asked what is necessary for democracy to thrive in the coming decade, most people point to increased participation, reduced polarization, and greater trust in institutions. However, underlying these issues is a more basic challenge: Citizens are increasingly unable to agree on the baseline set of facts necessary for debate, voting, or accountability. 

While democracy can withstand disagreement, it cannot function without a verifiable record of events that people can examine and contest. Reestablishing that baseline requires four concrete measures: funding for local journalism, transparency in platform ranking, verification skills in education, and labeling of synthetic media. Each of these targets a distinct point at which the shared record breaks down.

It is worth examining how this fragmentation came to exist and why it is a matter of design rather than circumstance. 

In previous decades, most individuals encountered news through a limited number of overlapping sources, such as major broadcast networks, local newspapers, and regional affiliates. Although coverage varied across these channels, it was generally based on a common pool of reporting, providing a shared foundation for public discourse. Today, information circulates through algorithmic systems designed to favor engagement rather than accuracy. Two individuals can follow the same election, protest, or policy decision and emerge with entirely different understandings of the events, each reinforced by the information systems they use.

This fragmentation is not accidental. It is a structural feature of the contemporary information environment, and its architects are aware of its impact. Internal Facebook research disclosed by whistleblower Frances Haugen in 2021 showed that for several years the platform’s algorithm weighted “angry” reactions more heavily than simple likes when ranking content, and that by 2019 data scientists found those posts skewed toward misinformation and toxicity. Testifying before the Senate, Haugen stated that content provoking extreme reactions is more likely to be clicked and reshared, regardless of its veracity. 

These consequences are tangible. In May 2023, a single fabricated AI-generated image caused a brief dip in the stock market. The image, which falsely depicted an explosion near the Pentagon, was disseminated by an account impersonating a Bloomberg news feed and included a verification checkmark. By the time officials confirmed that no explosion had occurred, the image had been broadcast as legitimate news on Republic TV, a major Indian television network. No firsthand photo or video ever surfaced. The false image alone was sufficient. The ability to verify authenticity has not kept pace with the capacity to generate fake content.

Independent journalism is among the few institutions designed to produce a verifiable record that can be checked, corrected, and referenced over time. Its decline has measurable consequences. Political scientists Danny Hayes and Jennifer Lawless tracked the same voters over time and found that decreases in political knowledge and civic participation consistently followed reductions in local news coverage. 

This decline has accelerated. The 2025 State of Local News Report found that 213 U.S. counties lack any local news source and another 1,524 have only one, leaving approximately 50 million Americans with limited or no access to reporting on their communities. Newspaper circulation has fallen by about 70 percent since 2005. The loss of a newsroom is not merely the loss of a record of local decisions; it is a reduction in the public’s capacity to participate in democracy.

At the same time, individuals now play an increasingly active role in shaping the information environment. Every share, repost, or omission influences what others encounter. Participation in democracy increasingly entails participation in information distribution, regardless of intent. While individuals are not responsible for the design of these systems, rebuilding a shared factual baseline requires more than institutional reform. Individuals must have the skills to critically evaluate information before disseminating it.

If the problem is structural, the response must be as well. What follows are four concrete interventions, each already underway in some form. 

1. Funding. Several states are experimenting with how to sustain local journalism. Illinois introduced a tax credit to incentivize local news organizations to hire and retain journalists. In its most recent funding round, the program distributed $4 million, with half allocated to outlets outside Chicago and an increasing share directed to nonprofit newsrooms. Another strategy is nonprofit ownership. The Texas Tribune, established in 2009 with philanthropic and local support rather than advertising revenue, serves as a model for subsequent nonprofit statewide newsrooms, while outlets like the Chicago Sun-Times show that existing commercial newsrooms can convert to nonprofit status rather than close. Although none of these initiatives fully reverses the 70 percent decline in print circulation since 2005, collectively they constitute a tangible, tested funding infrastructure rather than a hypothetical solution.
   
2. Transparency. The European Union’s Digital Services Act, effective since February 2024, mandates that large platforms explain in clear terms how their recommender systems rank and display content and provide at least one feed option without personalized profiling. This gives users a real choice about how their information is filtered. For the largest platforms, the law requires assessment and mitigation of risks related to elections and civic discourse. In response, several major platforms have committed to improving their handling of election-related content and to more clearly labeling AI-generated material. In the United States, a comparable proposal known as the Platform Accountability and Transparency Act has been introduced in the Senate across three Congresses since 2022, and, despite consistent bipartisan support, has yet to pass. This legislation would direct the Federal Trade Commission to require platforms to disclose which features use ranking algorithms, the data informing those algorithms, and information about the most widely distributed content. Both approaches rest on the same principle: platforms should be required to explain how they shape civic information, without dictating what that information should say.
   
3. Education. A technique known as lateral reading has proven effective in improving online source evaluation. Stanford researchers observed that professional fact-checkers assess unfamiliar sources by consulting other sources, rather than scrutinizing a site in isolation. This approach was tested across an entire school district, where just six fifty-minute lessons led to measurable improvements in students’ ability to evaluate online information. Unlike funding or regulation, this intervention asks for neither money nor new law—only that schools choose to teach it.
   
4. Labeling synthetic media. The implementation of the C2PA technical standard, supported by major technology companies such as Google, Microsoft, Adobe, and OpenAI, embeds signed records in AI-generated files to identify their origins. Regulatory measures have also been enacted. The European Union’s AI Act has required clear labeling of AI-generated content since August 2, 2026, with a grace period until December 2026 for generative systems already on the market. California’s AI Transparency Act imposed similar requirements beginning the same day. Some engineers are supplementing with invisible watermarking that persists through compression and re-uploading. Unlike other proposed solutions, the infrastructure for labeling synthetic media is already established and being implemented at scale.

No single intervention is sufficient or rapid enough to address the challenges to information integrity. Funding local journalism does not resolve algorithmic issues, and labeling synthetic media cannot restore closed newsrooms. Yet, each measure addresses a distinct vulnerability in the information ecosystem, and each already has actors behind it. 

Organizations like Rebuild Local News have helped legislators adapt Illinois’ funding model into law. State legislators and advocacy groups like Media Literacy Now are doing similar work for lateral reading: Twenty-five states now have media literacy laws, though implementation still lags behind legislation. Transparency legislation faces a steeper climb. The Platform Accountability and Transparency Act’s failure to reach a floor vote is a reminder that sponsorship alone does not guarantee passage. Of the four, funding has the best near-term chance of taking hold nationally as it is already replicating state-by-state rather than waiting on a federal vote. 

What unites these efforts is the underlying premise that democratic societies can withstand disagreements about policy, but not disputes over factual events. The primary objective is not to enforce consensus, but to reestablish conditions in which agreement or disagreement is grounded in a shared set of facts. This is not a partisan endeavor; it is a fundamental prerequisite for functional political discourse.