Exploring “Angry” and “Like” Reactions on Uncivil Facebook Comments That Correct Misinformation in the News

Abstract

This study examines the effect that uncivil news engagement comments that correct misinformation on a news story posted on Facebook have on newsreaders’ perceptions and attitudes. This study also considers whether social reactions, such as a “like” or an “angry” posted on correcting comments, will influence newsreaders’ perceptions and attitudes. Results from a survey experiment (N = 873) in the United States show that if comments that correct misinformation are uncivil, they led newsreaders to perceive the correcting comments and the commenters who posted them less favourably, although no effects on perceptions of the credibility of the news story itself were found. In addition, we found that uncivil correcting comments increased dislike for the political out-group in some instances unless they were marked with an “angry” social reaction. The “angry” reaction reversed the effect of the incivility, leading to decreased dislike for the out-group. Our findings advance digital journalism studies related to news engagement by showing the effects of incivility in corrections of misinformation and explaining how social reactions may alter the effect of these corrections.

Keywords:

• Incivility
• news
• credibility
• misinformation

Disclosure Statement

No potential conflict of interest was reported by the authors.

Notes

1 The Facebook post generator was accessed at http://www.prankmenot.com/. The Facebook generator only provided an option to add the “like” reaction to comments or stories, so we used a screen capture of the “angry” reaction and Photo Shopped it onto comments for those conditions.

2 See Medicaid eligibility requirements at: https://www.medicaid.gov/medicaid/eligibility/index.html

3 We conducted three pretests (n = 50, n = 100, and n = 49) on Amazon Mechanical Turk (MTurk) to determine which comments most accurately depicted either civil or uncivil messages. In each, participants rated a series of comments on a 1(strongly disagree) to 7 (strongly agree) scale, with a higher number meaning more civil. However, ratings from the first two pretests did not show significant differences between civil and uncivil comments. So the comments were modified to better depict both civil and uncivil messages. Participants in the third pretest were 37 years old on average (SD = 17.58), and most (63.3%) were female. The majority (67.3%) of the sample was white, followed by 10.2% black/African-American, 8.2% Hispanic/Latino, 6.1% prefer not to respond, 4.1% Asian, and 4.1% multiracial or other. Based on ratings from the third pretest, we chose the two comments for each topic (DACA and health care) with the highest means for civility for the civil condition and the lowest means for civility for the uncivil condition.

4 PROCESS Model 1, which tests simple moderation, was used.

5 It should be noted that we were concerned that participants’ political beliefs or their pre-existing attitudes toward the two contentious topics used in the experiment might influence results. Roughly a third of our sample (34.2%) identified as democrats, 29% as Republicans, and 36.8% as not affiliated with either major part. We measured attitudes toward the topics on two 1(strongly oppose) to 5 (strongly support) scales, where participants rated how they felt about Trump’s health care plan (M = 3.31, SE = 1.13) and the DACA program (M = 2.78, SD = 1.25). To decide whether to add these variables as covariates in our hypothesis tests, we examined whether these variables moderated the effects of either experimental manipulations (the tone of comments or the reactions posted on the comments), using a series of repeated measures analyses of variances (ANOVAs). None of these interactions were significant, suggesting pre-existing attitudes toward the two topics and political beliefs should not be included in hypothesis tests. We also conducted all our analyses both including and excluding political beliefs and pre-existing attitudes toward DACA and the health plan, and significant results held up whether these covariates were included or not. We report results without these covariates in the interest of parsimony.

6 Facebook introduced the seventh reaction, a heart being hugged to symbolize caring in April 2020, well after data for this study were collected.

Additional information

Funding

This work was supported by Texas A&M International University, USA (University Research Grant 2017–2018).