4 million tweets on two controversial topics were examined via sentiment analysis; negative sentiment of a person’s own tweets increased polarization; negative sentiment of tweets from a person’s friends reduced polarization

Does Negatively Toned Language Use on Social Media Lead to Attitude Polarization? Jürgen Buder et al. Computers in Human Behavior, December 11 2020, 106663. https://doi.org/10.1016/j.chb.2020.106663

Rolf Degen's take: https://twitter.com/DegenRolf/status/1337281131696283650

Highlights

• Research on attitude polarization has typically focused on content or congeniality.

• This paper looks at language sentiment as a precursor of polarization.

• 4 million tweets on two controversial topics were examined via sentiment analysis.

• Negative sentiment of a person’s own tweets increased polarization.

• Negative sentiment of tweets from a person’s friends reduced polarization.

Abstract: Prior research has indicated that both attitudinal homogeneity of communication networks (“echo chambers”) and attitudinal heterogeneity of communication networks (“adversarial debates”) can lead to attitude polarization. The present paper argues that communication in both echo chambers and adversarial debates is dominated by network negativity, a negative valence in the tone of discussions which might be associated with attitude polarization. Combining methods from sentiment analysis and social network analysis, more than 4 million tweets on two controversial topics (Brexit, Trump) were analyzed to investigate the occurrence of network negativity and its association with two proxies of attitude polarization (extremity and ambivalence). Results indicate that negativity in users’ own tweets was most strongly related to polarization, whereas negativity among users’ friends, or consonance of sentiments between users and friends had less impact on polarization. The findings are related to literatures on negativity bias, optimal distinctiveness theory, and intergroup contact theory.

Keywords: echo chamberattitudesattitude strengthsocial mediasentiment analysisnegativity bias