Internet users engage more with phatic posts than with health misinformation on Facebook

The study addresses concerns that Facebook facilitates the spread of health misinformation, a domain less examined than political misinformation. Focusing on the highly popular French Facebook page Santé + Mag, the authors ask how much of its content is health misinformation and whether such content particularly drives user engagement. They define health misinformation as information contrary to the epistemic consensus of the scientific community and distinguish it from disinformation by not inferring intent. The paper integrates perspectives from cultural evolution—emphasizing cognitive factors of attraction such as threat, disgust, sexuality, social information, and emotional valence—and media studies, which highlight the social and phatic functions of communication on social networks. The core hypotheses test whether engagement is explained by cognitively attractive content and/or by the phatic function of posts (i.e., content used to maintain social bonds rather than convey information).

Cultural attraction theory posits that psychological and ecological factors make certain content more appealing and memorable, facilitating its cultural spread. Prior work suggests false information can thrive due to psychological attractiveness rather than falsity per se, often leveraging threat, disgust, or social themes. Predicted effects include: negative over positive content; heightened engagement for threat-related, disgust-eliciting, and sex-related content; and an advantage for social information due to human social cognition. Media studies emphasize uses and gratifications and show Facebook primarily serves to maintain existing relationships, with prevalent small talk and phatic communication (e.g., greetings, affirmations of affection). The authors therefore propose that phatic posts may strongly drive engagement. The study formulates hypotheses that: health misinformation would predict interactions only if it includes cognitive factors of attraction; negative content would outperform neutral/positive; threat, disgust, and sexuality would increase interactions; social content would outperform nonsocial; and phatic posts would generate more interactions than nonphatic posts.

Data were collected via the CrowdTangle API for all 500 posts published by the Facebook page Santé + Mag from January 10 to January 31, 2019, along with their interactions: 2,673,972 reactions, 183,730 comments, and 3,626,186 shares, totaling 6,483,888 interactions. The sample’s distribution of post formats (articles, photos, videos) closely matched the page’s yearly activity, suggesting representativeness. Text appearing in images and videos was manually transcribed. Each post was manually coded by the first author for: (1) health misinformation status and potential harmfulness; (2) emotional valence (positive, negative, neutral); (3) presence of cognitive factors of attraction (negative emotion, threat, disgust, sexuality, social relations); (4) specific social topics (family, love, friends, society, antisocial people, pregnancy, mourning, infidelity); and (5) phatic versus nonphatic function. Posts could receive multiple codes (up to two attraction factors and up to three social topics). To validate coding, the second author and three independent coders (blind to hypotheses) each coded 50 randomly selected posts; agreement ranged from 85.55% to 90.91% with Cohen’s Kappa indicating substantial to almost perfect agreement. Statistical analyses (R 3.6.0, RStudio 1.1.419) included log-transforming interactions for normality (Shapiro-Wilk W=0.99, p=0.28) and conducting linear regressions. Holm-Bonferroni corrections were applied for multiple comparisons.

Content composition and prevalence: 28% (140/500) of posts were health misinformation; 21% (105/500) were potentially harmful. Cognitive factors were common: 50.4% social, 27.8% threat, 14.0% sexuality, 2.6% disgust; 82.0% contained at least one factor of attraction. Within social content, family (34.9%) and love (31.8%) predominated. Engagement predictors: Health misinformation with attraction factors negatively predicted interactions (adjusted R²=0.03, F(1,498)=15.24, p<0.001; β=-0.53, t(498)=-3.90, p<0.001). Among all posts, health misinformation was a negative predictor (adjusted R²=0.03, F(1,498)=17.32, p<0.001; β=-0.52, t(498)=-4.16, p<0.001). Among health misinformation only, the presence of attraction factors did not significantly predict interactions (β=0.17, t(138)=0.93, p=0.35); the number of such factors negatively predicted interactions (β=-0.28, t(138)=-2.09, p=0.04). Interaction shares: Only 14.3% of total interactions came from health misinformation; 10.7% from potentially harmful misinformation. Most interactions (85.8%) came from non-health-misinformation posts. Multiple regression with 10 predictors explained 52.8% of variance (adjusted R²≈0.53, F(10,489)=56.9, p<0.001). The strongest positive predictor was phatic posts (β=1.46, t(489)=13.59, p<0.001), followed by positive valence compared to neutral (β=0.40, t(489)=3.28, p=0.009) and compared to negative (β=0.48, t(489)=3.83, p=0.03). Sexual content negatively predicted interactions (β=-1.14, t(489)=-7.69, p<0.001). No significant effects were found for social information, threat, disgust, negative valence vs neutral, health misinformation (within the full model), potential harmfulness, or absence of attraction factors. Exploratory comment analysis of five highly commented harmful misinformation posts (n=4,737 comments) suggested many interactions were jokes or social tagging rather than endorsements (e.g., only 3% approval on one sample; high rates of tagging and humor).

Findings indicate that on Santé + Mag, engagement is driven primarily by the phatic function of posts and positive emotional valence rather than by cognitively attractive misinformation content. Despite the prevalence of cognitive factors among posts labeled as health misinformation, such content did not garner more interactions; indeed, health misinformation negatively predicted engagement and accounted for a minority of interactions. This aligns with broader evidence that misinformation is consumed and shared by a small minority of users and that its overall impact may be overstated. The predominance of phatic and positive content supports the view that Facebook functions chiefly as a platform for maintaining social relations and bonding, with users engaging to connect with friends and family more than to propagate misinformation. The lack of advantage for negative, threat, disgust, or sexual content challenges assumptions in cultural evolution about the primacy of negativity and hazard cues in transmission in this context. Exploratory comment analyses further suggest that interactions with misinformation posts often serve social or humorous purposes rather than signaling credulity. Overall, the results underscore the importance of examining the social context and function of communication, not only content features, when assessing cultural dynamics and misinformation on social media.

The study contributes a fine-grained, real-world analysis of Facebook content showing that phatic communication and positive valence are key drivers of engagement, whereas health misinformation—despite containing cognitively attractive features—receives comparatively less interaction and negatively predicts engagement. These results reinforce the characterization of Facebook as a platform primarily used to foster social ties rather than to disseminate misinformation. The authors advocate for interdisciplinary approaches combining cultural evolution and media studies, and for more detailed content analyses of social media to understand what truly drives interactions. Future research should further investigate the phatic dimension and positive content, explore diverse communication contexts on social platforms (e.g., public vs private spaces potentially affecting sharing of sexual content), and extend analyses to other pages, languages, and time periods to assess generalizability.

The analysis focuses on a single Facebook page (Santé + Mag) over a 21-day period, which may limit generalizability across platforms, regions, and times. Coding was primarily conducted by one author (though validated with additional coders), and observational interaction metrics may not capture endorsement or belief. The exploratory comment dataset for five posts is not publicly available due to sharing restrictions. Public–private dynamics on Facebook may modulate sharing of certain content (e.g., sexual themes), potentially affecting observed engagement patterns.