Bots don't Vote, but They Surely Bother! A Study of Anomalous Accounts in a National Referendum

TL;DR

This study analyzes Twitter discussions during the 2020 Chilean referendum, identifying anomalous accounts, including bots, that influence the spread of false information, using machine learning for profile-based detection.

Contribution

It introduces a profile-oriented machine learning approach to characterize and detect anomalous accounts involved in political discourse on social media.

Findings

Anomalous accounts correlate with vote turnout patterns.

Bots and automated accounts significantly spread false information.

Profile analysis effectively isolates suspicious activity.

Abstract

The Web contains several social media platforms for discussion, exchange of ideas, and content publishing. These platforms are used by people, but also by distributed agents known as bots. Although bots have existed for decades, with many of them being benevolent, their influence in propagating and generating deceptive information in the last years has increased. Here we present a characterization of the discussion on Twitter about the 2020 Chilean constitutional referendum. The characterization uses a profile-oriented analysis that enables the isolation of anomalous content using machine learning. As result, we obtain a characterization that matches national vote turnout, and we measure how anomalous accounts (some of which are automated bots) produce content and interact promoting (false) information.