The processing and evaluation of news content on social media is influenced by peer-user commentary

The study examines whether social cues on social media—specifically Likes and peer-user comments—create a bandwagon effect that influences how readers process and evaluate news content. Grounded in conformity research (e.g., Asch) and bandwagon/consensus heuristics, the authors hypothesize that Likes and comment sentiment will shape readers’ attitudes, sharing intentions, agreement with content (ideological congruence), perceptions of public opinion, and perceived credibility. They test five hypotheses: H1 that Likes improve evaluations; H2 that positive comments improve and negative comments impair evaluations; H3 that positive cues have stronger effects when readers initially disagree with content; H4 that negative comments have greater influence than positive comments (negativity bias); and H5 that effects are stronger for ideologically neutral content where prior dispositions are weak. The study also considers ecological validity by using an interactive, Facebook-like interface and varies article slant (congruent, incongruent, neutral) to examine interactions with reader predispositions.

Prior work shows social media cues act as bandwagon signals influencing attitudes and behavior. Likes are paralinguistic digital affordances used for engagement, social support, and impression management. Studies have reported that higher Likes can increase likeability or mimicry of Liking behavior, yet evidence for effects on evaluation is mixed. Peer comments can alter acceptance, attitudes, and perceived credibility, often interpreted via bandwagon or consensus heuristics. Comments invoking scientific evidence can be more persuasive than anecdotal ones, though anecdotal rhetoric can still sway certain audiences (e.g., low need for cognition). Winter et al. (2015) found no Like effects but stronger influence of negative comments, possibly due to ceiling effects and negativity bias. Negativity bias literature suggests negative information exerts stronger effects than positive. The current work builds on these findings, manipulating presence/omission of Likes and subjective comment sentiment (positive, negative, mixed, none) across articles differing in ideological slant to test bandwagon and negativity bias mechanisms.

Preregistration: Hypotheses, design, sample size, analyses, exclusion and inference criteria were preregistered on OSF (https://osf.io/d37f5), with sequential analyses and O’Brien-Fleming alpha correction (alpha = 0.011). Ethics approval was obtained (Erasmus University Rotterdam, DPECS 19-043); informed consent was collected. Participants: 560 psychology undergraduates were recruited; exclusions applied for reading time under two minutes, incomplete questionnaires, and use of mobile OS. Final N = 412 (330 female; Mean age = 20.52, SD = 2.85). Design and materials: A custom interactive website resembling Facebook presented three Guardian-adapted news articles: (1) ideologically congruent (Greta Thunberg meeting Justin Trudeau), (2) ideologically incongruent (violent video games and aggression; slightly altered to more strongly convey the relation), and (3) ideologically neutral (tropical storm near Ireland). Social cues were manipulated between-subjects in an 2 x 4 design: Likes (presented vs omitted) and comments (positive, negative, mixed [two positive + two negative], or no comments). Likes appeared under the article and under each comment. Comments were subjective/emotive/fallacious (no objective arguments). Content type (congruent, incongruent, neutral) was a within-subjects factor; each participant viewed all three articles. Eight between-subject conditions were implemented (presence/omission of Likes crossed with four comment conditions). Procedure: Participants were randomly assigned to one of eight conditions, accessed the site with a code, consented, and read three articles in block-randomized order (starting with congruent or incongruent, then neutral, then the remaining congruent/incongruent). After each article, embedded surveys captured outcome measures. A demographics/social media usage questionnaire followed, and a debrief disclosed the altered aspects of the video game article. Measures: Five seven-point scale outcomes per article with established internal consistency: (a) Personal attitude (Cronbach’s alpha = 0.81), (b) Share intent (0.86), (c) Ideological congruence—agreement with article’s conveyed ideas, measured separately per article content (alphas ~0.77), (d) Perceived public attitude (0.85), (e) Credibility (0.81). Items included statements like “I feel positive about the article,” “I would share this article,” topic-specific agreement items, public attitude perceptions, and credibility descriptors (accurate, authentic, believable). Analysis: Mixed factorial ANOVAs (2 Likes: omitted vs presented; 4 comment sentiments: positive, negative, mixed, none; 3 content types: congruent, incongruent, neutral) were run for each outcome. Mauchly’s test checked sphericity; Greenhouse-Geisser corrections applied; Bonferroni-adjusted pairwise comparisons; alpha = 0.011 via O’Brien-Fleming. Significant tests and generalized eta-squared are reported.

- Likes: Across all five outcomes, the presence versus omission of Likes had no significant effects: attitude F(1,404) = 0.83, p = 0.362; share intent F(1,404) = 0.09, p = 0.761; ideological congruence F(1,404) < 0.01, p = 0.995; perceived public attitude F(1,404) = 0.21, p = 0.644; credibility F(1,404) = 1.41, p = 0.236. - Comments—Attitude: Significant main effect of comment sentiment, F(3,404) = 4.19, p = 0.006, generalized η² = 0.013; negative and mixed comments yielded more negative attitudes than positive comments (positive vs mixed p = 0.002; positive vs negative p = 0.008). Interaction with content type marginal (p = 0.028 > alpha). Main effect of content type significant. - Share intent: Significant main effect of comments, F(3,404) = 4.52, p = 0.006, generalized η² = 0.004; positive > mixed (p = 0.002) and positive > negative (p = 0.003). Differences vs no comments for mixed (p = 0.013) and negative (p = 0.019) did not meet alpha. Interaction with content type marginal (p = 0.045 > alpha); overall presence of comments reduced sharing for congruent article. Main effect of content type significant. - Ideological congruence: Significant main effect of comments, F(3,404) = 12.15, p < 0.001, generalized η² = 0.039; negative and mixed < positive and no comments; positive vs no comments not different; negative vs mixed not different. Main effect of content type significant; no interaction with content; Likes null. - Perceived public attitude: Strong main effect of comments, F(3,404) = 87.22, p < 0.001, generalized η² = 0.248; all pairwise differences significant (p ≤ 0.007): positive highest, negative lowest, mixed intermediate. Significant interaction with content type, F(6.10,821.93) = 3.07, p = 0.006; effects larger for neutral content. Main effect of content type significant; Likes null. - Credibility: Significant main effect of comments, F(3,404) = 8.68, p < 0.001, generalized η² = 0.030; negative and mixed decreased credibility relative to positive and no comments; positive vs no comments not different; negative vs mixed not different. Main effect of content type significant; no interaction with content; Likes null. - Hypotheses: H1 (Likes improve evaluations) not supported; H2 (comment sentiment affects evaluations) supported primarily via negative/mixed comments lowering evaluations; H3 (stronger positive-cue effects under initial disagreement) not supported; H4 (negativity bias) supported—negative comments exerted stronger influence than positive, including in mixed condition; H5 (stronger effects for neutral content) not supported for negative comments, though perceived public attitude showed larger shifts for neutral content. - Mixed-comment condition: Perceived public attitude was intermediate between positive and negative, but personal evaluations (attitude, share intent, ideological congruence, credibility) aligned with negative comments, indicating a negativity bias rather than a simple bandwagon majority effect.

Findings show peer-user comments, especially negative ones composed of subjective/emotive rhetoric, reliably shift readers’ personal evaluations of news content—reducing attitudes, share intent, agreement with conveyed ideas, perceived public support, and credibility. Likes did not produce a bandwagon effect across outcomes. The pattern supports a negativity bias: negative comments dominated personal judgments even when balanced by positive comments (mixed condition), while perceived public attitude tracked the visible balance of comments, indicating distinct processing of personal evaluations versus perceptions of public opinion. Positive comments had limited effects and occasionally coincided with reduced willingness to share congruent content, suggesting the presence of subjective comment sections can dampen enthusiasm to share even agreeable articles. The null effects of Likes may stem from their ambiguity as aggregated numeric cues (paralinguistic digital affordances), aligning with exemplification theory: concrete exemplars (comments) are more persuasive than abstract metrics. Dual-process accounts suggest numbers may require analytic inference (system 2) whereas comment rhetoric is more readily processed (system 1). The results highlight how subjective, even fallacious, peer commentary can distort evaluations of ostensibly objective reporting and underscore the potential for social heuristics to shape online cognition.

Negative peer-user comments on social media can meaningfully alter how readers perceive and intend to act upon news content, while Likes do not reliably influence such evaluations. The study extends prior work by using an interactive platform, manipulating presence versus omission of Likes, and introducing a mixed-comment condition, revealing a clear negativity bias in personal evaluations. Practical implications include reconsideration of comment sections on news platforms seeking to inform objectively and user awareness of susceptibility to persuasive, critical comments. Future research should test whether older adults—who may exhibit a positivity bias—show different susceptibility patterns; examine whether perceived authority of critical commenters mediates persuasiveness; and investigate cognitive mechanisms by which Likes might gain meaning in familiar social contexts (e.g., from friends) or under different presentation formats.

- Assumed reader predispositions were weaker than expected: participants were less negative toward the ideologically incongruent article than planned, potentially limiting detection of interactions between negative predispositions and positive comments. - Sample characteristics: young adults (mean age ~20.5) who may exhibit stronger negativity bias than older adults, which may limit generalizability. - Comments were exclusively subjective/emotive; results may differ with objective/argumentative comments or scientific evidence. - Likes reflected aggregated feedback from unknown users; effects might differ in contexts where Likes come from known social ties (not tested here).