Synopsis

Modeling a French Election

Physics 18, s76
A fine-grained database of Twitter posts gathered during the 2017 French presidential election validates a popular network model of opinion dynamics.
Rama/CC BY-SA 3.0/Wikimedia Commons

The French presidential election of 2017 was unusually fluid. Two new parties entered the fray, including that of the eventual victor, Emmanuel Macron. Despite a financial scandal, the original front-runner, conservative François Fillon, remained in the race. Spotting an opportunity, a team of researchers harvested election-based Twitter activity to create a database reflecting voters’ shifting opinions [1]. Now Antoine Vendeville of Sciences Po in Paris has used the database to evaluate the ability of a popular network model, called the voter model, to predict the election’s outcome [2]. The model succeeded, demonstrating its relevance to fine-grained, highly heterogeneous settings.

The voter model features a network of interconnected individuals, some of whom are zealots—that is, people who will not change their voting intentions. The rest of the individuals are free to change their minds on the basis of the opinions of their neighbors in the network. Initially, individuals are assigned voting intentions at random. But after each iteration of the model, depending on certain weights, the voting intentions change and eventually converge on an equilibrium state, the election result.

The database that Vendeville used consists of 22,853 Twitter profiles and their election-related posts and reposts. The database’s creators went to the trouble of identifying the political affiliations of all those profiles. In Vendeville’s voter model the zealots―political organizations, not individual users―were assigned the appropriate affiliations, but the rest of the profiles were given affiliations at random. Nevertheless, the model converged to a state in which most users’ final opinions matched the affiliations inferred by the database’s creators. The model was also able to efficiently distinguish friends from foes through their probability of disagreeing.

–Charles Day

Charles Day is a Senior Editor for Physics Magazine.

References

  1. O. Fraisier et al., “#Élysée2017fr: The 2017 French presidential campaign on Twitter,” Proc. Intl. AAAI Conf. Web and Social Media 12, 1 (2018).
  2. A. Vendeville, “Voter model can accurately predict individual opinions in online populations,” Phys. Rev. E 111, 064310 (2025).

Subject Areas

Statistical PhysicsComplex Systems

Related Articles

Faster Rankings for Group Contests
Interdisciplinary Physics

Faster Rankings for Group Contests

A new algorithm efficiently ranks competitors in settings ranging from sports tournaments to food-preference surveys. Read More »

Topological Twist for Phase Transitions
Statistical Physics

Topological Twist for Phase Transitions

Contrary to conventional wisdom, so-called order parameters that distinguish symmetry-governed phases of matter can have topological structure. Read More »

Cells Show “Personality” When They Move
Statistical Physics

Cells Show “Personality” When They Move

A study of cellular motion demonstrates a method for isolating individual cellular behavior from environmentally induced random motion. Read More »

More Articles