In the classical mathematical study of democracy, researchers often rely on a simplifying assumption called "impartial culture." This model treats every voter as a perfectly independent coin flip—statistical islands unaffected by their neighbors, the news, or shared cultural values. Under this assumption, elegant theorems like Condorcet’s Jury Theorem promise that as a group grows, it will almost certainly arrive at the "correct" decision.
However, real-world voters do not live in vacuums. They share social networks, consume the same media, and belong to ideological blocs. Recent research has begun to dismantle the "spherical cow" of independent voting, revealing that correlation doesn't just change the math—it fundamentally shifts the boundaries of what is possible in social choice.
The Fragility of Collective Wisdom
The "wisdom of the crowd" relies on the idea that individual errors are random and will cancel each other out. Olle Häggström and Gil Kalai demonstrated in A Law of Large Numbers for Weighted Majority that this property is surprisingly robust for weighted majorities, provided no single voter is too dominant. But as soon as voters become highly correlated, the "wisdom" can evaporate.
When individuals influence one another through social networks, their errors become correlated. Elchanan Mossel, Joe Neeman, and Omer Tamuz explored this in Majority Dynamics and Aggregation of Information in Social Networks, showing that talking to neighbors can actually hinder the crowd's accuracy by creating feedback loops of misinformation. In these settings, the group's final choice may reflect the bias of a few influential "gurus" rather than the aggregate knowledge of the population.
Strategic Voting and the Bliss of Consensus
One of the most famous "black pills" in social choice is the Gibbard-Satterthwaite theorem, which states that any fair voting system is vulnerable to strategic manipulation. If you know how others will vote, you can often get a better outcome by lying about your own preferences.
However, correlation might actually be the cure for this manipulation. Debmalya Mandal and David C. Parkes found in Correlated Voting that when voters believe their peers have similar preferences, the incentive to lie vanishes surprisingly quickly as the group size grows. Intuitively, if I believe most people agree with me, my "honest" vote is already part of a powerful bloc, making a strategic lie both risky and unnecessary.
This theme is echoed by Daniel Halpern and Ariel D. Procaccia in Strategyproof Voting under Correlated Beliefs. They argue that "top-choice correlated beliefs"—the assumption that if I like a candidate, others probably do too—makes simple rules like Plurality remarkably robust against gaming.
Phase Transitions in Impossibility
Perhaps the most striking finding in this research area is that correlation can "break" the famous impossibility theorems that have long haunted social choice. Arrow’s Theorem famously proves that "fair" voting systems are doomed to produce irrational, circular outcomes (like the group preferring A over B, B over C, and C over A).
But Frederic Koehler and Elchanan Mossel showed in A Phase Transition in Arrow's Theorem that these paradoxes are largely artifacts of the independence assumption. By modeling voter interactions as a Markov Random Field, they discovered a sharp phase transition: once social influence reaches a certain threshold, the mathematical "impossibility" of a rational group preference effectively disappears. Social consensus acts as a stabilizing force that irons out the logical kinks of individual preferences.
The Cost of Correlation in Selection
While correlation can stabilize large elections, it can be devastating for "impartial selection"—scenarios where the voters are also the candidates (such as a committee choosing its own chair).
Mechanisms designed to be impartial usually rely on the statistical independence of votes to guarantee accuracy. Ioannis Caragiannis and George Christodoulou have shown in Impartial selection with prior information that when voters form coordinated blocs or exhibit dependent preferences, these mechanisms collapse. The error rate jumps from a manageable logarithmic scale to a linear one, meaning the mechanism becomes no better than picking a name at random.
Why It Works: The Geometry of Influence
The underlying reason these effects occur is geometric. In an independent model, voter preferences are scattered uniformly across the space of all possible rankings. In a correlated model, preferences cluster.
As Elchanan Mossel, Miklós Z. Rácz, and Ariel D. Procaccia demonstrated in A Smooth Transition from Powerlessness to Absolute Power, the ability of a small group to manipulate an election depends on how close the election is to a "boundary" where the winner changes. Correlation moves the entire electorate closer to or further from these boundaries in predictable ways, creating "smooth" transitions in power and manipulability rather than the erratic jumps predicted by worst-case theory.
By moving beyond the "impartial culture," researchers are building a social choice theory that finally looks like the society it intends to model: a web of interconnected individuals whose shared beliefs are both the greatest strength and the most significant challenge of collective decision-making.