Social Media and Democracy

Ronen Gradwohl, Yuval Heller, Arye HillmanarXiv 2022 · DOI · arXiv

Abstract

While Condorcet’s Jury Theorem suggests that large groups of voters are likely to make correct decisions, this paper demonstrates that a monopoly social media platform can systematically overturn this result. Using a Bayesian persuasion framework, the authors show that a platform can manipulate election outcomes by providing private, algorithmically filtered information to voters, even when those voters are aware of the manipulation.

Introduction

The "wisdom of the crowd" is a cornerstone of democratic legitimacy. Condorcet’s Jury Theorem (CJT) provides the mathematical basis for this, stating that if individual voters are even slightly more likely than not to choose the correct option, the collective decision of a large group will be correct with near certainty as the population grows. However, the rise of social media platforms as primary information gatekeepers challenges this assumption. Unlike traditional mass media, social media platforms utilize private algorithms to tailor information to individual users. This paper explores how such a platform, acting as a "natural monopoly," can exploit information asymmetry to force its preferred political outcome, even when that outcome is objectively worse for the electorate.

The Model of Algorithmic Persuasion

The study employs the framework of Bayesian Persuasion (specifically the model established by Kamenica and Gentzkow) to analyze the interaction between a platform (the "designer") and a continuum of voters.

In this model, there are two possible states of the world (e.g., Candidate A is better or Candidate B is better). The platform desires a specific outcome (Outcome A) regardless of the true state. Voters possess some exogenous information with an accuracy $q$. The platform commits to a messaging rule—an algorithm—that sends conditionally independent private signals to each voter.

Crucially, voters are assumed to be "sincere" Bayesians: they vote for the candidate they believe is most likely to be the better choice, but they are fully aware of the platform's manipulation strategy and account for it when updating their beliefs. This makes the platform's success particularly notable, as it occurs even without deceiving voters about its underlying motives.

Overturning the Jury Theorem

The authors find that a platform can design signals such that a majority of voters are nudged to believe Outcome A is more likely, even if the underlying reality supports Outcome B. Because the population is large, the Law of Large Numbers ensures that this manipulation becomes deterministic. The platform does not merely influence the probability of an outcome; it effectively dictates the result.

A critical discovery is the 70.7% Threshold. In a homogeneous population, the platform can perfectly manipulate the election if and only if the voters' initial information accuracy $q$ is below $\sqrt{2}/2 \approx 0.707$. If voters are more informed than this threshold, the platform's ability to overturn the "wisdom of the crowd" vanishes, as the exogenous information becomes too robust to be countered by algorithmic filtering.

The Paradox of Negative Bias

One of the paper's most striking insights is the Paradox of Negative Bias. To maintain credibility among relatively informed voters, a platform may need to send signals that are biased against its preferred candidate.

If the platform only sent "pro-A" messages, Bayesian voters would quickly discount them as noise. By occasionally sending "pro-B" signals, the platform maintains enough credibility to flip the specific marginal voters required to secure a majority for A. Specifically, when voter accuracy is between $2/3$ and $\sqrt{2}/2$, the platform must use a negatively biased signal to succeed. This "subtle" manipulation is more effective at shifting the needle than overt propaganda.

Social Media vs. Traditional Media

The paper highlights a fundamental difference between social media and traditional mass media. Traditional media sends a public signal—everyone sees the same information. This can shift the probability of an outcome but rarely guarantees it in a large, diverse population. Social media, through private persuasion, can target specific subsets of the population with different signals. This allows for a more surgical and deterministic manipulation of the democratic process, as the platform only needs to persuade a simple majority to achieve its goal.

Complexity in Diverse Populations

When the population is heterogeneous (a mix of highly informed and less informed voters), the platform's ability to manipulate becomes non-monotonic. It is often easier to control a population that is either mostly informed or mostly uninformed. A diverse mix creates a challenge: a single algorithmic strategy may fail to "tailor the lie" effectively to both groups simultaneously. In such cases, the platform might lose control of the outcome because it cannot satisfy the Bayesian updating requirements of both groups with one messaging rule.

Conclusion and Limitations

The research provides a sobering look at the power of algorithmic gatekeeping. While the model assumes sincere voting and a monopoly platform, it underscores the vulnerability of collective decision-making to sophisticated information filtering.

There are several avenues for future research:

  • Strategic Voting: In massive populations, voters may vote strategically (considering the probability of being a tie-breaker) rather than sincerely.
  • Competition: The presence of multiple platforms with competing agendas could mitigate the manipulation power of a single monopoly.
  • Correlation Neglect: Real-world voters often suffer from "correlation neglect," failing to realize that their peers are receiving similarly biased information. This would likely make the platform's task even easier than the Bayesian model suggests.

Ultimately, the paper suggests that the mathematical safeguards of democracy—the collective intelligence of the many—can be systematically dismantled by the private, algorithmic nature of modern information consumption.

Summary updated

Abstract

We study the ability of a social media platform with a political agenda to influence voting outcomes. Our benchmark is Condorcet's jury theorem, which states that the likelihood of a correct decision under majority voting increases with the number of voters. We show how information manipulation by a social media platform can overturn the jury theorem, thereby undermining democracy. We also show that sometimes the platform can do so only by providing information that is biased in the opposite direction of its preferred outcome. Finally, we compare manipulation of voting outcomes through social media to manipulation through traditional media.
BibTeX
@misc{gradwohl2022social,
  author = {Ronen Gradwohl and Yuval Heller and Arye Hillman},
  title = {{Social Media and Democracy}},
  year = {2022},
  eprint = {2206.14430},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2206.14430},
}
Figure