Abstract:
Adversarial interactions against politicians on social media such as Twitter have significant impact on society, and in particular both discourage people from seeking office and disrupt substantive political discussions online. In this study, we measure the adversarial interactions towards candidates during the run-up to the 2018 US general election. We gather a new dataset consisting of 1.7 million tweets involving candidates, one of the largest corpora focusing on political discourse. We then develop new techniques for detecting tweets with toxic content and the target of its hostility, which allows us to quantify adversarial interactions towards political candidates at scale. We go on to design a new algorithm to induce candidate-specific adversarial terms to capture more nuanced adversarial interactions that are in most other contexts not considered toxic. Together our techniques enable us to categorize the breadth of adversarial interactions seen in the election, including offensive name-calling, threats of violence, posting discrediting information, attacks on identity, and adversarial message repetition.
