Cross-platform Analysis of Twitter and Parler during the 2020 U.S. Presidential Election

In the recent 2020 Presidential Election, President Trump and his campaign alleged that mail-in ballots were likely to be fraudulent and this claim stood against Twitter’s efforts to curb spreading of misinformation (Lima, 2020). This claim resulted in suspending those who participated in voter fraud misinformation (Twitter, 2021), including Trump himself. In response to Twitter’s action, Trump and those who supported Trump left Twitter, seeking an alternative social media. This migration was a strong collective action by users who felt more than simply constrained (Kiene, Monroy-Hernández & Hill, 2016) by a loss of belonging to the community when users faced increased censorship. Those who left Twitter found Parler as an alternative social networking service, which proclaims that they allow a user to “speak freely and express yourself openly, without fear of being deplatformed for your views” as an asylum. Parler has gained attention from conservatives who are looking for alternative social media, which supposedly accepts them for who they are. Based on this unique case, this study seeks to understand the impact of echo chambers on people’s expressed opinions on social media. Past research efforts on echo chambers, selective exposure, and network homogeneity (Stewart, Arif & Starbird, 2018; Jacobson, Myung & Johnson, 2016) mostly focused on a handful of popular social media, mostly either Facebook or Twitter, while neglecting the unique roles of other niche social media platforms in building online communities (Zannettou et al., 2018). We will address this critical gap by leveraging data from two social media platforms: Parler and Twitter as examples that represent distinctive user bases in terms of political ideology. We identify users who have the same account names on both platforms and examine the role of political homogeneity in the online opinion expression and sharing of information. We rely on the Social Identity Deindividuation Effects (SIDE) model to understand political behaviors of the users who used both Twitter and Parler. The SIDE model explains that deindividualization occurs when group norms are more salient and have a greater effect on individual behaviors than individual processes (Lea & Spears, 1992). The SIDE models focus on anonymity and explicit and implicit norms of online spaces, and supports that anonymity enhances the social influence processes and collective behavior (Spears, 2017). By applying this theoretical model, we are aiming to reveal how Parler’s homogeneous political climate – more conservative than Twitter – helped users to feel more anonymous than Twitter by providing a safe place for them to speak hatred. There are two research questions we wanted to answer. Our focus of interest is the people who used both Twitter and Parler and hereafter, they are called cross-platform users. RQ 1. Can we make use of the machine learning technique to identify the pattern of increasing or decreasing use of toxic language by cross-platform users in Twitter? RQ 2. Can we make use of the machine learning technique to identify the pattern of increasing or decreasing use of toxic language by cross-platform users in Parler?Ope