Social Media Messages Research Articles

The Gender Gap in Elite-Voter Responsiveness Online

A number of important studies have documented gender gaps in the effectiveness or performance of individual representatives. Yet whether these differences are observable when it comes to responsiveness to public opinion is unclear. In this article, I examine the degree to which representatives use social media to dynamically respond to shifts in issue salience among the electorate. After combining nearly 400 bi-weekly repeated public opinion surveys from YouGov asking voters about their issue priorities, I trained a large language model to classify the universe of elected U.S. and UK representatives’ social media messages on Twitter to the same issues. Findings reveal that women representatives demonstrate greater responsiveness than their male counterparts to shifts in issue salience according to both women and men constituents. Despite an overall bias toward male constituents, female representatives play a crucial role in narrowing the gender gap by consistently aligning their attention with the issues prioritized by female constituents. These findings not only contribute to our understanding of elite-voter responsiveness but also underscore the substantive benefits that women representatives provide for all constituents.

From recovery to preparedness: an examination of resources, individual disaster communication, and anticipatory resilience capacity during the 2023 U.S. tornado season

ABSTRACT Multi-phase research is needed to understand how disaster-impacted residents maintain and adapt individual disaster communication (IDC; Spialek & Houston, 2018; Spialek et al., 2021). Residents engage in IDC when they search for, create, or share interpersonal or social media messages intended to help themselves and other individuals cope across disaster phases. This study examined how recovery-focused IDC and anticipatory resilience capacity explain the relationship between resources and preparedness focused IDC. We surveyed 417 residents who lived in 13 counties impacted by a spring 2023 tornado outbreak. Results revealed that individuals who have (a) more physical and social resources and (b) engage in more recovery-focused IDC following a disaster report higher levels of anticipatory resilience capacity in those recovery-focused messages, which in turn, leads individuals to participate in more preparedness-focused communication about potential future disasters. The findings support the development of interventions and campaigns to increase preparedness behaviors.

Culprits or Victims? Social Media Use, Mediated Ethnic Blame, and Asian American Mental Health During COVID-19

ABSTRACT The present study aims to understand whether and how social media use negatively affected Asian Americans’ mental health during the COVID-19 pandemic. We build on previous research on media framing, stigma, and stereotyping to understand whether social media messages about Asians as culprits to blame for the pandemic, direct experiences of racism, or media coverage of victims of anti-Asian hate crimes led to mental health outcomes for this racial group. We use an online survey of Asian American participants (N = 527) conducted in 2020 to test the role of (a) mediated ethnic blame (MEB), (b) direct stigma, and (c) indirect stigma in explaining the relationship between social media use and mental health among Asian Americans during the COVID-19 pandemic. Our results demonstrate that mediated ethnic blame, but not direct or indirect stigma experiences, is associated with Asian Americans’ mental health. Implications for media framing, health communication, and journalism are discussed.

Breast milk sharing practices in emergency settings

Abstract Background In emergencies, breast milk sharing becomes essential to sustain infant nutrition when access to maternal breastfeeding is disrupted. This study examines donor motivations and assesses the impact of such initiatives on infant health outcomes during wartime. Methods This qualitative study investigated a community of breast milk donors on a Facebook platform for the month following the start of the war on October 7, 2023. We investigated two main groups participating in breast milk sharing through reflective thematic analysis of social media messages. Results A count of posts one month before and after October 7 revealed a doubling in post frequency post-7/10, emphasizing the surge in war-related posts. Two main themes emerged concerning the war: 1. Infant needs: Posts addressed various infant needs, including general necessities and those specific to war-related situations. Requests for milk sharing emerged from the second day of the war, particularly for infants whose mothers were missing or injured. Government information from the National Human Milk Bank was shared within these Facebook groups, encouraging donations for these infants. 2. Donor Motivation: Posts reflected both non-war-related and war-related motivations, with a notable emphasis on the latter. Authors, including both women and men, actively promoted the donation of human milk to conflict-affected areas. Suggestions included pumping milk specifically for war-affected infants or offering frozen human milk. Some posts also offered to collect and deliver human milk to needy infants within the affected regions. Conclusions Post frequency doubled in both Facebook groups compared to the previous month, mainly focusing on war-related topics, indicating an increased drive to share human milk with those affected by the conflict. Key messages • Breast milk sharing, crucial in emergencies, analyzed via social media, sheds light on donor motivations and impacts on infant health during conflicts. • This qualitative study investigated a community of breast milk donors on a Facebook platform for the month following the start of the war on October 7, 2023.

From online hate speech to offline hate crime: the role of inflammatory language in forecasting violence against migrant and LGBT communities

Social media messages often provide insights into offline behaviors. Although hate speech proliferates rapidly across social media platforms, it is rarely recognized as a cybercrime, even when it may be linked to offline hate crimes that typically involve physical violence. This paper aims to anticipate violent acts by analyzing online hate speech (hatred, toxicity, and sentiment) and comparing it to offline hate crime. The dataset for this preregistered study included social media posts from X (previously called Twitter) and Facebook and internal police records of hate crimes reported in Spain between 2016 and 2018. After conducting preliminary data analysis to check the moderate temporal correlation, we used time series analysis to develop computational models (VAR, GLMNet, and XGBTree) to predict four time periods of these rare events on a daily and weekly basis. Forty-eight models were run to forecast two types of offline hate crimes, those against migrants and those against the LGBT community. The best model for migrant crime achieved an R2 of 64%, while that for LGBT crime reached 53%. According to the best ML models, the weekly aggregations outperformed the daily aggregations, the national models outperformed those geolocated in Madrid, and those about migration were more effective than those about LGBT people. Moreover, toxic language outperformed hatred and sentiment analysis, Facebook posts were better predictors than tweets, and in most cases, speech temporally preceded crime. Although we do not make any claims about causation, we conclude that online inflammatory language could be a leading indicator for detecting potential hate crimes acts and that these models can have practical applications for preventing these crimes.

“Should you really chat while reading?” effects of on-screen multitasking and text disfluency on integrated understanding

Students' multitasking during lectures or academic assignments is a common activity that may have negative effects on performance. It is therefore relevant to investigate under what conditions such negative effects can be reduced. The current study brought together two separate lines of research to examine the effects of on-screen multitasking and text disfluency. Multitasking required students to read and respond to social media messages during reading. Text disfluency was perceptual in the form of harder-to-read versus easy-to-read text. We included 208 university students, who were randomly assigned to one of the four conditions resulting from a 2 x 2 between-subjects design. Participants read two dual-positional texts on sun exposure and health. Dependent on the assigned condition, they could read perceptually fluent or disfluent texts, and they could intermittently receive, or not receive, on-screen social media messages while reading. As outcome variables, we measured participants’ perception of cognitive load, integrated text understanding, and metacognitive calibration of text understanding. We also controlled the possible contributions of the individual differences of prior knowledge, reading comprehension skills, working memory, and cognitive flexibility. Results showed a negative effect of multitasking on the integrated understanding of the dual positional texts with non-multitaskers outperforming multitaskers. Perception of cognitive load and metacognitive calibration were not affected by multitasking. No impact of disfluency was observed for any of the outcome variables, and the interactive effect of multitasking and disfluency was also not statistically significant. Limitations and directions for future research are discussed.

Talk to Me: A social media campaign about clinical trials tailored to the needs of young adults with cancer.

67 Background: Among adolescents and young adults (AYAs; 15-39 years) with cancer, clinical trial enrollment (CTE) rates range from 3% to 14%. Patient-directed interventions to improve CTE are rare. Internet-based interventions, while feasible and acceptable to AYAs with cancer, often lack patient input and engagement that are critical to success. AYAs are digital natives and report that the internet and social media are the two most important sources of health-related information. We leveraged AYA engagement with social media and designed a mixed methods research study, "Talk to Me," to create targeted social media messages to increase awareness about CTE for AYAs with cancer. Methods: Our interdisciplinary study team includes clinicians, researchers, health educators, and AYA patient advocates. We conducted 5 focus groups among a demographically, clinically, and geographically diverse group of AYAs with cancer (n=35) to identify priority topics to 1) inform the development of social media message libraries to address informational gaps and dispel myths surrounding CTE and 2) guide implementation strategies of these messages. Design mock-ups and color palettes were co-designed and co-developed with focus group participants and patient advocates through iterative feedback. Results: Guided by qualitative findings, social media posts were categorized into "introduction", "knowledge dissemination", and "addressing social, cultural, emotional, and financial issues" related to CTE. Instagram was selected to host our messages as the demographics of users most closely matched that of our target audience. Hashtags were carefully chosen based on existing posts of related material on Instagram. Links to trusted sources for additional information were embedded in the post descriptions and housed in a linktr.ee bio-landing page. The campaign was officially launched on Instagram in April 2024. We posted 3 times a week in both English and Spanish and monitored metrics biweekly to assess engagement and reach. Over nearly 2 months, 42 messages were posted, 116 users followed the account, 656 accounts were reached, 108 accounts were engaged, and 432 visited our profile page. Engagement peaked in Week 2 with 238 accounts reached and in Week 6 with 119. Overall, there were 173 likes (English=128, Spanish=43). Conclusions: An interdisciplinary, patient-informed, iterative approach to developing a social media knowledge intervention demonstrates a robust, methodologic approach to developing acceptable information and messages for AYAs who may consider participating in clinical trials.

Expressions of uncertainty in online science communication hinder information diffusion.

Despite the importance of transparent communication of uncertainty surrounding scientific findings, there are concerns that communicating uncertainty might damage the public perception and dissemination of science. Yet, a lack of empirical research on the potential impact of uncertainty communication on the diffusion of scientific findings poses challenges in assessing such claims. We studied the effect of uncertainty in a field study and a controlled experiment. In Study 1, a natural language processing analysis of over 2 million social media (Twitter/X) messages about scientific findings revealed that more uncertain messages were shared less often. Study 2 replicated this pattern using an experimental design where participants were presented with large-language-model (LLM)-generated high- and low-uncertainty messages. These results underscore the role of uncertainty in the dissemination of scientific findings and inform the ongoing debates regarding the benefits and the risks of uncertainty in science communication.

Authenticity and high performance: Nonmarket social, public and media strategies of foreign SMEs during a black swan event

AbstractNonmarket (NM) strategy is an emerging concept, similar to the notion of NM social, public and media strategy (SPMS). Previous literature has predominantly focused on larger corporations, often overlooking smaller enterprises. Additionally, much of the existing research has been centred on developed markets, leaving a gap in qualitative research. The current analysis seeks to elucidate how foreign small and medium‐sized enterprises (FSMEs) in transitional markets, such as Shanghai, adopt specific NM SPMS to mitigate the liability of foreignness (LOF) during a pandemic. The study utilizes a CGT approach, employing semi‐structured interviews with 27 participants, including foreign entrepreneurs and digital marketing specialists. Additionally, qualitative data was analysed comprehensively using Atlas.ti 9 software. The qualitative methodology implementation resulted in the creation of the authenticity and high performance framework from where specific strategies and general implications are presented. The study highlights two principal themes: (1) The need for increased flexibility and speed in social media message communication and (2) Customized FSME strategies through digital marketing—a never‐going‐back market digitalization. The second theme is subdivided into (a) Tailored public relations for FSMEs. (b) An increased need for internal cohesion—fostering a stable mindset and positive rapport among staff. (c) An increased need for trialling new target customers. (d) Creating a relatable story—showing a particular vulnerability can be viral and (e) Maintaining a strong reputation for operational performance.

Large Language Models Outperform Expert Coders and Supervised Classifiers at Annotating Political Social Media Messages

Instruction-tuned Large Language Models (LLMs) have recently emerged as a powerful new tool for text analysis. As these models are capable of zero-shot annotation based on instructions written in natural language, they obviate the need of large sets of training data—and thus bring potential paradigm-shifting implications for using text as data. While the models show substantial promise, their relative performance compared to human coders and supervised models remains poorly understood and subject to significant academic debate. This paper assesses the strengths and weaknesses of popular fine-tuned AI models compared to both conventional supervised classifiers and manual annotation by experts and crowd workers. The task used is to identify the political affiliation of politicians based on a single X/Twitter message, focusing on data from 11 different countries. The paper finds that GPT-4 achieves higher accuracy than both supervised models and human coders across all languages and country contexts. In the US context, it achieves an accuracy of 0.934 and an inter-coder reliability of 0.982. Examining the cases where the models fail, the paper finds that the LLM—unlike the supervised models—correctly annotates messages that require interpretation of implicit or unspoken references, or reasoning on the basis of contextual knowledge—capacities that have traditionally been understood to be distinctly human. The paper thus contributes to our understanding of the revolutionary implications of LLMs for text analysis within the social sciences.

The meaning of obstetric violence experiences: A qualitative content analysis of the Break the Silence Campaign

BackgroundObstetric violence involves the mistreatment, disrespect, and abuse of birthing people and the problem has been recognized in healthcare systems worldwide. Obstetric violence is a gendered, sex-specific form of violence against women that is a public health problem and a violation of human rights. There are an unknowable number of online posts and social media messages that describe obstetric violence experiences. There are no known studies about self-published experiences of obstetric violence in the US maternity care system. ObjectiveTo understand the meaning of obstetric violence experiences in the US maternity care system from a naturally occurring, purposive sample. DesignA secondary analysis of the textual data from the original Break the Silence social media campaign using qualitative content analysis. The theory of social justice in nursing provided a theoretical framework. A healthcare systems approach was used for a wide-angle view of the multidirectional structure, processes, and outcome of obstetric violence. SettingThe study setting is the public Facebook page where the Break the Silence social media campaign can be seen. Break the Silence was an online activism response to the problem of obstetric violence in the US maternity care system with signboard messages posted from 31 known US states. The setting is bounded by the digital page where the campaign is published. ParticipantsThere were 139 participants, and 11 of them posted more than one signboard message. Most participants were birthing people (n = 125) followed by doulas (n = 10). MethodsKrippendorff's methodology for qualitative content analysis was applied to 156 signboard messages posted on Break the Silence from 2014 to 2016. Qualitative content analysis was supported by Atlas.ti 23. ResultsFour themes illustrated the meaning of obstetric violence in US maternity care: 1) pregnancy and birth as a battle with healthcare providers and the healthcare system, 2) sacrifice of the maternal body normalized and assumed as a gender stereotype, 3) disrupted rites of passage from childbirth, and 4) abuse of fiduciary power by healthcare providers. ConclusionsThis study demonstrated thematic meanings for the experience of obstetric violence in US maternity care with a healthcare systems approach that included structural and organizational considerations to increase understanding. Categories and forms of obstetric violence from the existing literature were expanded and strengthened by findings from this study. Themes were validated in principle with consistency in findings across the international evidence base on obstetric violence.Tweetable abstract: The meaning of #obstetricviolence experiences is interpersonal and structural with thematic consistency across international studies.

Augmenting surveys with social media discourse on the workings of democracy from a cross-national perspective

This paper introduces an unsupervised framework that illustrates how insights gathered from opinion surveys regarding the functionality of democracy can be connected to social media messages of politicians on an international scale. By concentrating on the influence of social media messages from elected officials, the study adopts a “top-down” theoretical approach that links citizens’ attitudes towards democracy with the viewpoints about democracy expressed by politicians within social media discourses. Using a word embedding classification strategy, democracy-related themes are extracted from politicians’ messages. The research is conducted across 11 European countries, namely, Austria, Finland, France, Germany, Greece, Italy, the Netherlands, Spain, Sweden, and the United Kingdom. Switzerland is also included in the study because of its direct democracy system. The 10th round of the European Social Survey serves as the basis for assessing citizens’ democratic attitudes. The study encompasses two main analytical segments. First, aggregated analyses conducted at the country level reveal the degree of alignment between prominent democracy-related themes present in politicians’ discourse and citizens’ perceived significance of these same themes. Second, individual-level connections between social media data and survey respondents are established through their preferred political party (or orientation). Variable importance analysis is subsequently applied to explore which democracy-related themes conveyed by politicians hold significance in predicting public contentment with democracy.

Examining Texting in the Classroom Between Adolescents Without and With Learning Disabilities: Do Classroom Climate and Academic Boredom Matter?

The present study investigated texting in the classroom through written (SMS) and instant (social media) messages comparatively between adolescents without and with LD. Also, applying the Social Cognitive Theory, the role of perceived classroom climate and academic boredom in texting in the classroom was co-examined through an explanatory mediation model. Overall, 439 students (59% girls) without LD and 408 students (47% girls) with LD, who both attended randomly selected Greek junior high schools of Mainstream Education and used a mobile phone, completed self-reported scales about the variables involved with reference to the Modern Greek language course. The results showed that texting in the classroom (through written/instant messages) was above average for both student subgroups, regardless of their gender. However, within a comparative perspective, texting in the classroom through instant (social media) messages was higher for students with LD. Furthermore, in the relationship between perceived classroom climate and texting in the classroom (through written/instant messages) academic boredom proved a full mediator for students without LD and a partial mediator for students with LD. The findings imply the need for psychoeducational interventions in the context of specific school courses, to strengthen differentiated dimensions of psychosocial classroom climate for adolescents without and with LD. Thus, positive academic emotions and minimized disruptive texting in the classroom could be achieved.

Influence of social media on health-related decision-making among adults attending an outpatient department of a tertiary care centre in India: A cross-sectional analytical study.

Background Social media platforms, especially Facebook and WhatsApp, can spread public health information effectively. We aimed to estimate the influence of health-related messages circulated through these social media platforms on health-related decision-making and its associated factors. Methods We did a cross-sectional analytical study among adults (aged >18 years) who visited the outpatient department of a tertiary care hospital in suburban West Bengal, during July-September 2021. A structured questionnaire was used regarding receiving health-related messages on social media and the subsequent effect on health-related decision-making in the past year. Results A total of 673 individuals participated in the study. Their mean (SD) age was 34.4 (10.2) years and 56.8% (382) were men, 50.8% (342) were graduates, 63.6% (428) were from rural areas and 82.9% (558) were active users of more than one social media platform. A total of 474 (70.4%; 95% CI 67.0-73.9) study participants reported health-related decision-making based on social media messages, whereas 44.7% (301) reported checking the authenticity of forwarded messages or posts or updates with healthcare professionals before making a decision. On adjusted analysis, participants who had secondary education (adjusted prevalence ratio [aPR] 1.40; 95% CI 1.01-1.94), used both the media (aPR 1.31; 95% CI 1.09-1.58) and checked the authenticity of the messages with a healthcare professional (aPR 1.52, 95% CI 1.38-1.68) were significantly more influenced by the messages, posts or updates received on social media platforms. Conclusion WhatsApp forwards or updates and Facebook posts or updates influence health-related decision-making among the Indian adult population.

Perspectives of Sexual and Gender Minority Youth on Anti-Vaping Messages in Social Media.

Sexual and gender minority (SGM) youth have higher rates of nicotine vaping than other youth in the United States. While social media can be effective in reaching youth and discouraging vaping, informed cultural tailoring is necessary to ensure effective messaging to SGM youth. This study aimed to understand SGM youth perspectives on anti-vaping social media messages and tailoring approaches. In-depth, qualitative videoconference interviews were conducted from February to July 2022 with 34 SGM youth recruited in the United States via social media ads. The interview guide addressed participants' beliefs about vaping, the context of vaping, perspectives on tailoring messages, and responses to examples of social media anti-vaping messages. Coding and thematic analysis followed a team-based approach. SGM youth perspectives fell into four categories - representation and diversity, facts and evidence, empowering messages, and source credibility. Participants stressed the importance of accurate, genuine representation of SGM youth in messages, but also noted that more overt representation may be seen as tokenizing. Participants recommended partnering with known LGBTQ+ influencers who can promote or share anti-vaping messages on social media platforms. They also recommended using culturally tailored language, including statistics specific to SGM youth, and invoking themes of empowerment to improve the relevance, reach, and effectiveness of anti-vaping campaigns. Findings can inform future efforts to develop anti-vaping messages for SGM youth with effective reach through social media. Nuanced perspectives on SGM representation in messages suggest a careful approach to tailoring. Concerns around inauthenticity may be minimized by ensuring SGM youth are included in message development and dissemination. This study describes the importance of being attentive to the tailoring preferences among the current generation of sexual and gender minority youth. Findings will inform social media-based messaging strategies that discourage nicotine vaping tailored for SGM youth in health campaign material design and evaluation, ensuring that tailored messages are designed in ways that avoid unintended consequences. The study also describes methods for effectively engaging SGM youth in research to improve the relevance of health education materials for this population and increase reach, which in turn can lead to reduction in vaping practices among SGM youth.

“Here’s what you can do to stop the spread”: Frames and mobilizing information in CDC’s social media messages during the COVID-19 pandemic

ABSTRACT Although there is a body of research on message frames and mobilizing information (MI), not enough attention has been paid to whether public health organizations are benefiting from the research in their social media messages during a public health crisis. To fill this research gap, this study aimed to examine how the CDC in the U.S. used gain and loss frames and MI in their Instagram posts about COVID-19 prevention and whether the use of such message strategies changed throughout the pandemic. A mixed-methods content analysis of a sample of 565 posts showed that the CDC overwhelmingly relied on gain frame, interactive MI, and tactical MI as their message strategies. The use of frames and MI was mostly consistent throughout the pandemic, although the target behavior changed. The supplemental analysis also revealed that loss-framed or gain- and loss-framed messages led to more user engagement. The findings highlighted the need for an evidence-based approach in social media messages by public health organizations. Theoretical and practical implications are discussed.

Evaluation of BERT-Based Models on Patient Data from French Social Media.

With the objective of extracting new knowledge about rare diseases from social media messages, we evaluated three models on a Named Entity Recognition (NER) task, consisting of extracting phenotypes and treatments from social media messages. We trained the three models on a dataset with social media messages about Developmental and Epileptic Encephalopathies and more common diseases. This preliminary study revealed that CamemBERT and CamemBERT-bio exhibit similar performance on social media testimonials, slightly outperforming DrBERT. It also highlighted that their performance was lower on this type of data than on structured health datasets. Limitations, including a narrow focus on NER performance and dataset-specific evaluation, call for further research to fully assess model capabilities on larger and more diverse datasets.

Automated Disaster Monitoring From Social Media Posts Using AI-Based Location Intelligence and Sentiment Analysis

Worldwide disasters like bushfires, earthquakes, floods, cyclones, and heatwaves have affected the lives of social media users in an unprecedented manner. They are constantly posting their level of negativity over the disaster situations at their location of interest. Understanding location-oriented sentiments about disaster situation is of prime importance for political leaders, and strategic decision-makers. To this end, we present a new fully automated algorithm based on artificial intelligence (AI) and natural language processing (NLP), for extraction of location-oriented public sentiments on global disaster situation. We designed the proposed system to obtain exhaustive knowledge and insights on social media feeds related to disaster in 110 languages through AI- and NLP-based sentiment analysis, named entity recognition (NER), anomaly detection, regression, and Getis Ord Gi* algorithms. We deployed and tested this algorithm on live Twitter feeds from 28 September to 6 October 2021. Tweets with 67 515 entities in 39 different languages were processed during this period. Our novel algorithm extracted 9727 location entities with greater than 70% confidence from live Twitter feed and displayed the locations of possible disasters with disaster intelligence. The rates of average precision, recall, and F₁-Score were measured to be 0.93, 0.88, and 0.90, respectively. Overall, the fully automated disaster monitoring solution demonstrated 97% accuracy. To the best of our knowledge, this study is the first to report location intelligence with NER, sentiment analysis, regression and anomaly detection on social media messages related to disasters and has covered the largest set of languages.

Hybrid model for sentiment analysis combination of PSO, genetic algorithm and voting classification

As social network services like Weibo and Twitter have grown in popularity, natural language processing (NLP) has seen a great deal of interest in sentiment analysis of social media messages and Information mining. Social media users, whose numbers are always increasing, have the ability to exchange information on their platforms. The study of sentiment, domains and themes are closely related. Manually collecting enough labelled data from the vast array of subjects covered by large-scale social media to train sentiment classifiers across several domains would be extremely difficult. The literature review conducted concludes that models already proposed in the previous researches are not able to achieve good accuracy. This work suggests a unique model that combines of genetic algorithm and particle swarm optimization to effectively extract the features and then the voting technique is applied for the classification. Model proposed is compared with 4 ensemble datasets achieving a consistent accuracy of more than 90% for three different diversified database owing to natural selection of sequences by GA and at the same time achieves a fast convergence with PSO, the model may be employed for highly accurate recommenders demanding precision and accuracy.

Emotion labelling strengthens readers’ emotional responses to online messages

ABSTRACT Emotionally overheated social media discussions cause many harms both to the individuals participating in the discussion and to the wider society. To support emotion regulation in social media discussions, the strategy of affect labelling is explored. Two experiments were run to study the effects of emotion labels presented with social media messages on participants’ emotional responses to the messages. Altogether 84 participants read emotional social media messages presented without and with emotion labels (Experiment 1, N = 39), and without labels and with emotion / control labels (Experiment 2, N = 45) and rated their subjective emotional valence. Results show that the ratings of valence were significantly higher for the positive messages presented with label than for the messages presented as such without emotion label or presented with a control label. In addition, the results of the second experiment show that the negative social media messages were experienced as more negative when the messages were presented with an emotion label than without emotion label or with control label. The results suggest that affect labelling of social media messages is an ineffective strategy to down-regulate emotions but could be used to support and strengthen positive emotions.