Understanding hostility in online political discussions: Non-hostile people opt out of the discussions, individuals prone to hostility could be more likely to participate in online than offline discussions

Why so angry? Understanding hostility in online political discussions. Alexander Bor &Michael Bang Petersen. Online disinformation: an integrated view conference 2019. Aarhus University. https://nordis.research.it.uu.se/wp-content/uploads/2019/04/Online-Disinformation-Conference-Abstracts.pdf

Most US citizens consider online political discussions to be uncivil, aggressive and hostile. Although political discussions can get heated by their very nature, the available data suggests that online discussions are seen as much worse than offline discussions. Yet, while a range of studies has documented the existence of widespread perceptions of online political hostility, our understanding of the causes of these perceptions are limited. The aim of the present paper is to provide the first comprehensive review and empirical test of the potential mechanisms that could cause widespread online political hostility.

As our theoretical starting point, we distinguish between two broad factors: the messages sent and the message received. Two potential explanations relate to the first factor. The first explanation proposes negative behavioral changes. This suggests that people are more likely to send hostile messages online than offline. Anonymity is a frequently blamed culprit of online hostility, but online environments have several other unique features too; for example, people are often distracted or tired while crafting their messages. Such characteristics may undermine people’s emotion regulation mechanisms and may lead to increased hostility on online platforms. The second explanation proposes that online environments attract particular individuals and, thereby, change the composition of people participating in political discussions. Through such sorting, individuals prone to hostility could be more likely to participate in online than offline discussions. It is possible that whereas social defense mechanisms effectively guard offline discussions against hostile intruders, it is more difficult to exclude hostile parties from online discussions.

It is also important to consider whether the perceived hostility is exacerbated on the receivers’ end too. A third explanation suggests that the density of communication networks in online environments may also contribute to higher perceived hostility, even if the same messages are being sent by the same people. Whereas offline political discussions typically involve only a handful of people and rarely more than a few dozen, in online discussions hundreds can participate. In other words, any hostile message is likely to be received by many more people inan online discussion. Computer algorithms prioritizing comments with attention grabbing or controversial content likely intensify this effect. Finally, the perception effect proposes that even if none of the explanations above were true, it is possible that the same messages are perceived as more hostile online than in face-to-face interactions. People may have a harder time judging the hostility of messages they receive and may make more false-positive mistakes online, where they lack non-verbal cues, lack reputational information about the sender, and often see only fragments of a conversation thread. We test observable implications for these explanations relying on an original online survey of US citizens (N = 1500), conducted by YouGov on an approximately representative sample. The survey includes measures of political participation and hostility online and offline, perceptions of these environments as well as a wide array of personality measures.

Against common predictions, we find little evidence for behavior being corrupted by online environments or for sorting effects revealing the entry of new, hostile parties to online discussions. Our data shows very high correlations between online and offline behaviors including hostility (rs> 0.8). Moreover, personality measures, which indicate a general tendency for hostile political behavior(such as need for chaos, trait aggression, status driven risk-seeking and difficulties in emotion regulation), correlate highly with both online and offline hostility (0.4 < rs <0.6). We do find firm evidence for sorting caused by hostile individuals being more active onlinethan non-hostile individuals. According to our data, people who self-report sending hostile messages 1) spend more time on social media (but less time on other parts of the internet) and 2) discuss politics more than non-hostile individuals. These differences are particularly large when it comes to discussing politics with strangers.

We find suggestive evidence that perception and network effects also increase perceived hostility. Discussions with strangers correlate with negative impressions of online discussions. Importantly, this is not true for offline discussions: the more people discuss politics with strangers face-to-face, the more positive their evaluations get. Furthermore, due to the density of online communication networks, discussions with strangers are more frequent both in absolute and relative terms online than offline. The paper concludes with discussing the implications of these findings.