Keywords: Social media, big data, sentiment analysis, and emerging technologies

As networked and mobile communication technologies permeate civil society around the globe, practitioners and researchers alike are increasingly looking to online behavioral markers and trace data as proxy measures for popular and public opinion. Likes, shares, retweets, and comments are the common currency for market researchers and news companies. They also influence the selection criteria for the algorithms that determine which stories filter to the top of the news feeds across media platforms. Thus, news engagement metrics represent the ever-present “pulse” of news saliency that ultimately directs public attention.

Despite the relative importance of these metrics, very little has been written about the type of content these measures select for. Some researchers have proposed that social media, and in particular social media text, can be treated as public opinion (Murphy et al., 2014). At least one study argues that so-called “found data” in social media text might approximate traditional survey responses (Schober et al., 2016). Yet, predicting what makes a news article popular is a difficult task (Arapakis, Cambazoglu, & Lalmas, 2017) and less is known about the audience preferences and behaviors that drive newsworthiness in the digital realm. Drawing on methodologies of sentiment scoring and topic modeling, the following research question is proposed:

RQ. How is public opinion reflected in the news headlines that garner the most engagement on Facebook in terms of likes, shares and comments?

Methodology:
Based on this general research gap, the current study will explore the latent public opinion embedded in the news engagement statistics of the top 15 news websites (including US and UK news organizations) on Facebook from 2012-2016 (N > 600,000). To parse the universe of content, headlines scoring in the top quartile per week for likes, shares, and comments will be isolated. Headlines will then be analyzed for sentiment and topic prevalence. Facebook also includes options for users to explicitly express discreet emotions (e.g. love, anger, sadness). These attributes of opinion will be treated separately.

Implications:
Engagement statistics play an increasingly outsized role in what people pay attention to online. The current study has implications for extending existing methodologies for emerging and alternative forms of opinion mining (as opposed to traditional surveys). In addition, and perhaps most importantly, results have implications for how mass opinion is directed on a daily basis.

References:
Arapakis, I., Cambazoglu, B. B., & Lalmas, M. (2017). On the feasibility of predicting popular news at cold start. Journal of the Association for Information Science and Technology, 68(5), 1149-1164.

Murphy, J., Link, M. W., Childs, J. H., Tesfaye, C. L., Dean, E., Stern, M., ... & Harwood, P. (2014). Social media in public opinion research: executive summary of the Aapor task force on emerging technologies in public opinion research. Public Opinion Quarterly, 78(4), 788-794.

Schober, M. F., Pasek, J., Guggenheim, L….. (2016). Social media analyses for social measurement. Public opinion quarterly, 80(1), 180-211.