Measuring Hate: Does a Definition Affect Self-Reported Levels of Perpetration and Exposure to Online Hate in Surveys?

Abstract

The purpose of this research is to test the validity of commonly used measures of exposure to and production of online extremism. Specifically, we investigate if a definition of hate influences survey responses about the production of and exposure to online hate. To explore the effects of a definition, we used a split experimental design on a sample of 18 to 25-year-old Americans where half of the respondents were exposed to the European Union’s definition of hate speech and the other half were not. Then, all respondents completed a survey with commonly used items measuring exposure to and perpetration of online hate. The results reveal that providing a definition affects self-reported levels of exposure and perpetration, but the effects are dependent on race. The findings provide evidence that survey responses about online hate may be conditioned by social desirability and framing biases. The findings that group differences exist in how questions about hate are interpreted when definitions of it are not provided mean we must be careful when using measures that try to capture exposure to and the production of hate. While more research is needed, we recommend providing a clear, unambiguous definition when using surveys to measure online hate.

How Algorithms Promote Self-Radicalization: Audit of TikTok’s Algorithm Using a Reverse Engineering Method

“…However, it is now easier to access such a group, no matter where it is in the world. The radicalizing process can now be initiated online, where the radicalized message is shared, regenerated, amplified, and spread rapidly at an unparalleled speed (Hawdon et al, 2023;Shin et al, 2024).…”

“…Through the FYP, TikTok’s algorithms increasingly present users with far-right accounts to follow. This often sends users spiraling down a rabbit hole of progressively further aggressive, hateful, and violent content (Cotter et al, 2022; Hawdon et al, 2023).…”

“…Having conducted interviews with Islamic extremists, Baugut and Neumann (2020) found that users began adopting such ideologies despite having only an initial basic interest in them, but they followed the platform’s recommendations when they were shown radical propaganda, which pushed them to engage in violence. Hawdon et al (2023) found that Facebook and X recommend radicalized hate stories for users to follow after an individual engages with extreme content, arguing that platforms unintentionally forge a network that connects extremists.…”

“…Algorithms that personalize user content have raised significant social concerns because they have the potential to intensify and reinforce spirals that can result in users’ adoption of radical and extremist beliefs (Baugut & Neumann, 2020). Social media platforms are embedded with algorithms that feed users with content that stimulates extremism and promotes radicalism (Hawdon et al, 2023). Specifically, social media users are algorithmically categorized into similar peer groups based on their interactions with platforms, driving them into steadily more insular online environments and creating what is known as filter bubbles.…”

“…People who join online communities are funneled toward more extreme and insular communities and ultimately into invite-only online groups that require a checking process to make users more parochial and insular. For example, Hawdon et al (2023) demonstrated that algorithmic personalization is higher for populations with extreme ideological beliefs; thus, there is an urgent need to identify factors that make one vulnerable to radicalization online (Hollewell & Longpré, 2022). The cognitive process by which users embrace radical beliefs is called self-radicalization (hereafter radicalization).…”