Negative affect variability differs between anxiety and depression on social media.

OBJECTIVE: Negative affect variability is associated with increased symptoms of internalizing psychopathology (i.e., depression, anxiety). The Contrast Avoidance Model (CAM) suggests that individuals with anxiety avoid negative emotional shifts by maintaining pathological worry. Recent evidence also suggests that the CAM can be applied to major depression and social phobia, both characterized by negative affect changes. Here, we compare negative affect variability between individuals with a variety of anxiety and depression diagnoses by measuring the levels and degree of change in the sentiment of their online communications. METHOD: Participants were 1,853 individuals on Twitter who reported that they had been clinically diagnosed with an anxiety disorder (A cohort, n = 896) or a depressive disorder (D cohort, n = 957). Mean negative affect (NA) and negative affect variability were calculated using the Valence Aware Dictionary for Sentiment Reasoning (VADER), an accurate sentiment analysis tool that scores text in terms of its negative affect content. RESULTS: Findings showed differences in negative affect variability between the D and A cohort, with higher levels of NA variability in the D cohort than the A cohort, U = 367210, p < .001, r = 0.14, d = 0.25. Furthermore, we found that A and D cohorts had different average NA, with the D cohort showing higher NA overall, U = 377368, p < .001, r = 0.12, d = 0.21. LIMITATIONS: Our sample is limited to individuals who disclosed their diagnoses online, which may involve bias due to self-selection and stigma. Our sentiment analysis of online text may not completely capture all nuances of individual affect. CONCLUSIONS: Individuals with depression diagnoses showed a higher degree of negative affect variability compared to individuals with anxiety disorders. Our findings support the idea that negative affect variability can be measured using computational approaches on large-scale social media data and that social media data can be used to study naturally occurring mental health effects at scale.

Misinformation and Public Health Messaging in the Early Stages of the Mpox Outbreak: Mapping the Twitter Narrative With Deep Learning.

BACKGROUND: Shortly after the worst of the COVID-19 pandemic, an outbreak of mpox introduced another critical public health emergency. Like the COVID-19 pandemic, the mpox outbreak was characterized by a rising prevalence of public health misinformation on social media, through which many US adults receive and engage with news. Digital misinformation continues to challenge the efforts of public health officials in providing accurate and timely information to the public. We examine the evolving topic distributions of social media narratives during the mpox outbreak to map the tension between rapidly diffusing misinformation and public health communication. OBJECTIVE: This study aims to observe topical themes occurring in a large-scale collection of tweets about mpox using deep learning. METHODS: We leveraged a data set comprised of all mpox-related tweets that were posted between May 7, 2022, and July 23, 2022. We then applied Sentence Bidirectional Encoder Representations From Transformers (S-BERT) to the content of each tweet to generate a representation of its content in high-dimensional vector space, where semantically similar tweets will be located closely together. We projected the set of tweet embeddings to a 2D map by applying principal component analysis and Uniform Manifold Approximation Projection (UMAP). Finally, we group these data points into 7 topical clusters using k-means clustering and analyze each cluster to determine its dominant topics. We analyze the prevalence of each cluster over time to evaluate longitudinal thematic changes. RESULTS: Our deep-learning pipeline revealed 7 distinct clusters of content: (1) cynicism, (2) exasperation, (3) COVID-19, (4) men who have sex with men, (5) case reports, (6) vaccination, and (7) World Health Organization (WHO). Clusters that largely communicated erroneous or irrelevant information began earlier and grew faster, reaching a wider audience than later communications by official instances and health officials. CONCLUSIONS: Within a few weeks of the first reported mpox cases, an avalanche of mostly false, misleading, irrelevant, or damaging information started to circulate on social media. Official institutions, including the WHO, acted promptly, providing case reports and accurate information within weeks, but were overshadowed by rapidly spreading social media chatter. Our results point to the need for real-time monitoring of social media content to optimize responses to public health emergencies.

"I Haven't Been Diagnosed, but I Should Be"-Insight Into Self-diagnoses of Common Mental Health Disorders: Cross-sectional Study.

BACKGROUND: In recent years, social media has become a rich source of mental health data. However, there is a lack of web-based research on the accuracy and validity of self-reported diagnostic information available on the web. OBJECTIVE: An analysis of the degree of correspondence between self-reported diagnoses and clinical indicators will afford researchers and clinicians higher levels of trust in social media analyses. We hypothesized that self-reported diagnoses would correspond to validated disorder-specific severity questionnaires across 2 large web-based samples. METHODS: The participants of study 1 were 1123 adults from a national Qualtrics panel (mean age 34.65, SD 12.56 years; n=635, 56.65% female participants,). The participants of study 2 were 2237 college students from a large university in the Midwest (mean age 19.08, SD 2.75 years; n=1761, 75.35% female participants). All participants completed a web-based survey on their mental health, social media use, and demographic information. Additionally, the participants reported whether they had ever been diagnosed with a series of disorders, with the option of selecting "Yes"; "No, but I should be"; "I don't know"; or "No" for each condition. We conducted a series of ANOVA tests to determine whether there were differences among the 4 diagnostic groups and used post hoc Tukey tests to examine the nature of the differences. RESULTS: In study 1, for self-reported mania (F3,1097=2.75; P=.04), somatic symptom disorder (F3,1060=26.75; P<.001), and alcohol use disorder (F3,1097=77.73; P<.001), the pattern of mean differences did not suggest that the individuals were accurate in their self-diagnoses. In study 2, for all disorders but bipolar disorder (F3,659=1.43; P=.23), ANOVA results were consistent with our expectations. Across both studies and for most conditions assessed, the individuals who said that they had been diagnosed with a disorder had the highest severity scores on self-report questionnaires, but this was closely followed by individuals who had not been diagnosed but believed that they should be diagnosed. This was especially true for depression, generalized anxiety, and insomnia. For mania and bipolar disorder, the questionnaire scores did not differentiate individuals who had been diagnosed from those who had not. CONCLUSIONS: In general, if an individual believes that they should be diagnosed with an internalizing disorder, they are experiencing a degree of psychopathology similar to those who have already been diagnosed. Self-reported diagnoses correspond well with symptom severity on a continuum and can be trusted as clinical indicators, especially in common internalizing disorders such as depression and generalized anxiety disorder. Researchers can put more faith into patient self-reports, including those in web-based experiments such as social media posts, when individuals report diagnoses of depression and anxiety disorders. However, replication and further study are recommended.

Quantifying changes in societal optimism from online sentiment.

Individuals can hold contrasting views about distinct times: for example, dread over tomorrow's appointment and excitement about next summer's vacation. Yet, psychological measures of optimism often assess only one time point or ask participants to generalize about their future. Here, we address these limitations by developing the optimism curve, a measure of societal optimism that compares positivity toward different future times that was inspired by the Treasury bond yield curve. By performing sentiment analysis on over 3.5 million tweets that reference 23 future time points (2 days to 30 years), we measured how positivity differs across short-, medium-, and longer-term future references. We found a consistent negative association between positivity and the distance into the future referenced: From August 2017 to February 2020, the long-term future was discussed less positively than the short-term future. During the COVID-19 pandemic, this relationship inverted, indicating declining near-future- but stable distant-future-optimism. Our results demonstrate that individuals hold differentiated attitudes toward the near and distant future that shift in aggregate over time in response to external events. The optimism curve uniquely captures these shifting attitudes and may serve as a useful tool that can expand existing psychometric measures of optimism.

Deep learning for topical trend discovery in online discourse about Pre-Exposure Prophylaxis (PrEP).

Pre-Exposure Prophylaxis (PrEP) interventions are increasingly prevalent on social media. These data can be mined for insights about PrEP that may not be as apparent in surveys including personal musings about PrEP and barriers/facilitators to PrEP uptake. This study explores online discourse about PrEP using an interdisciplinary public health and computational informatics approach. We collected (N = 4,020) tweets using Twitter's Application Programming Interface (API). These data underwent a three-step neural network/deep learning process to identify clusters within these tweets and relative similarity/dissimilarity between clusters. We identified 25 distinct clusters from our original collection of tweets. These clusters represent general information about PrEP, how PrEP is communicated among diverse groups, and potential pockets of misinformation and disinformation regarding PrEP. Specific clusters of interest include discussions of medication side effects, social perception of PrEP usage, and concerns with costs and barriers to access of PrEP interventions. Our approach revealed diverse ways PrEP is contextualized online. Importantly this information can be leveraged to identify points of possible intervention for disinformation and misinformation about PrEP.

Sociodemographics and Transdiagnostic Mental Health Symptoms in SOCIAL (Studies of Online Cohorts for Internalizing Symptoms and Language) I and II: Cross-sectional Survey and Botometer Analysis.

BACKGROUND: Internalizing, externalizing, and somatoform disorders are the most common and disabling forms of psychopathology. Our understanding of these clinical problems is limited by a reliance on self-report along with research using small samples. Social media has emerged as an exciting channel for collecting a large sample of longitudinal data from individuals to study psychopathology. OBJECTIVE: This study reported the results of 2 large ongoing studies in which we collected data from Twitter and self-reported clinical screening scales, the Studies of Online Cohorts for Internalizing Symptoms and Language (SOCIAL) I and II. METHODS: The participants were a sample of Twitter-using adults (SOCIAL I: N=1123) targeted to be nationally representative in terms of age, sex assigned at birth, race, and ethnicity, as well as a sample of college students in the Midwest (SOCIAL II: N=1988), of which 61.78% (1228/1988) were Twitter users. For all participants who were Twitter users, we asked for access to their Twitter handle, which we analyzed using Botometer, which rates the likelihood of an account belonging to a bot. We divided participants into 4 groups: Twitter users who did not give us their handle or gave us invalid handles (invalid), those who denied being Twitter users (no Twitter, only available for SOCIAL II), Twitter users who gave their handles but whose accounts had high bot scores (bot-like), and Twitter users who provided their handles and had low bot scores (valid). We explored whether there were significant differences among these groups in terms of their sociodemographic features, clinical symptoms, and aspects of social media use (ie, platforms used and time). RESULTS: In SOCIAL I, most individuals were classified as valid (580/1123, 51.65%), and a few were deemed bot-like (190/1123, 16.91%). A total of 31.43% (353/1123) gave no handle or gave an invalid handle (eg, entered "N/A"). In SOCIAL II, many individuals were not Twitter users (760/1988, 38.23%). Of the Twitter users in SOCIAL II (1228/1988, 61.78%), most were classified as either invalid (515/1228, 41.94%) or valid (484/1228, 39.41%), with a smaller fraction deemed bot-like (229/1228, 18.65%). Participants reported high rates of mental health diagnoses as well as high levels of symptoms, especially in SOCIAL II. In general, the differences between individuals who provided or did not provide their social media handles were small and not statistically significant. CONCLUSIONS: Triangulating passively acquired social media data and self-reported questionnaires offers new possibilities for large-scale assessment and evaluation of vulnerability to mental disorders. The propensity of participants to share social media handles is likely not a source of sample bias in subsequent social media analytics.

Quantifying collective identity online from self-defining hashtags.

Mass communication over social media can drive rapid changes in our sense of collective identity. Hashtags in particular have acted as powerful social coordinators, playing a key role in organizing social movements like the Gezi park protests, Occupy Wall Street, #metoo, and #blacklivesmatter. Here we quantify collective identity from the use of hashtags as self-labels in over 85,000 actively-maintained Twitter user profiles spanning 2017-2019. Collective identities emerge from a graph model of individuals' overlapping self-labels, producing a hierarchy of graph clusters. Each cluster is bound together and characterized semantically by specific hashtags key to its formation. We define and apply two information-theoretic measures to quantify the strength of identities in the hierarchy. First we measure collective identity coherence to determine how integrated any identity is from local to global scales. Second, we consider the conspicuousness of any identity given its vocabulary versus the global identity map. Our work reveals a rich landscape of online identity emerging from the hierarchical alignment of uncoordinated self-labeling actions.

Quantifying societal emotional resilience to natural disasters from geo-located social media content.

Natural disasters can have devastating and long-lasting effects on a community's emotional well-being. These effects may be distributed unequally, affecting some communities more profoundly and possibly over longer time periods than others. Here, we analyze the effects of four major US hurricanes, namely, Irma, Harvey, Florence, and Dorian on the emotional well-being of the affected communities and regions. We show that a community's emotional response to a hurricane event can be measured from the content of social media that its population posted before, during, and after the hurricane. For each hurricane making landfall in the US, we observe a significant decrease in sentiment in the affected areas before and during the hurricane followed by a rapid return to pre-hurricane baseline, often within 1-2 weeks. However, some communities exhibit markedly different rates of decline and return to previous equilibrium levels. This points towards the possibility of measuring the emotional resilience of communities from the dynamics of their online emotional response.

Assessing rigid modes of thinking in self-declared abortion ideology: natural language processing insights from an online pilot qualitative study on abortion attitudes.

INTRODUCTION: Although much work has been done on US abortion ideology, less is known relative to the psychological processes that distinguish personal abortion beliefs or how those beliefs are communicated to others. As part of a forthcoming probability-based sampling designed study on US abortion climate, we piloted a study with a controlled sample to determine whether psychological indicators guiding abortion beliefs can be meaningfully extracted from qualitative interviews using natural language processing (NLP) substring matching. Of particular interest to this study is the presence of cognitive distortions-markers of rigid thinking-spoken during interviews and how cognitive distortion frequency may be tied to rigid, or firm, abortion beliefs. METHODS: We ran qualitative interview transcripts against two lexicons. The first lexicon, the cognitive distortion schemata (CDS), was applied to identify cognitive distortion n-grams (a series of words) embedded within the qualitative interviews. The second lexicon, the Linguistic Inquiry Word Count (LIWC), was applied to extract other psychological indicators, including the degrees of (1) analytic thinking, (2) emotional reasoning, (3) authenticity, and (4) clout. RESULTS: People with polarized abortion views (i.e., strongly supportive of or opposed to abortion) had the highest observed usage of CDS n-grams, scored highest on authenticity, and lowest on analytic thinking. By contrast, people with moderate or uncertain abortion views (i.e., people holding more complex or nuanced views of abortion) spoke with the least CDS n-grams and scored slightly higher on analytic thinking. DISCUSSION AND CONCLUSION: Our findings suggest people communicate about abortion differently depending on their personal abortion ideology. Those with strong abortion views may be more likely to communicate with authoritative words and patterns of words indicative of cognitive distortions-or limited complexity in belief systems. Those with moderate views are more likely to speak in conflicting terms and patterns of words that are flexible and open to change-or high complexity in belief systems. These findings suggest it is possible to extract psychological indicators with NLP from qualitative interviews about abortion. Findings from this study will help refine our protocol ahead of full-study launch.

The rise and fall of rationality in language.

The surge of post-truth political argumentation suggests that we are living in a special historical period when it comes to the balance between emotion and reasoning. To explore if this is indeed the case, we analyze language in millions of books covering the period from 1850 to 2019 represented in Google nGram data. We show that the use of words associated with rationality, such as "determine" and "conclusion," rose systematically after 1850, while words related to human experience such as "feel" and "believe" declined. This pattern reversed over the past decades, paralleled by a shift from a collectivistic to an individualistic focus as reflected, among other things, by the ratio of singular to plural pronouns such as "I"/"we" and "he"/"they." Interpreting this synchronous sea change in book language remains challenging. However, as we show, the nature of this reversal occurs in fiction as well as nonfiction. Moreover, the pattern of change in the ratio between sentiment and rationality flag words since 1850 also occurs in New York Times articles, suggesting that it is not an artifact of the book corpora we analyzed. Finally, we show that word trends in books parallel trends in corresponding Google search terms, supporting the idea that changes in book language do in part reflect changes in interest. All in all, our results suggest that over the past decades, there has been a marked shift in public interest from the collective to the individual, and from rationality toward emotion.

Historical language records reveal a surge of cognitive distortions in recent decades.

Individuals with depression are prone to maladaptive patterns of thinking, known as cognitive distortions, whereby they think about themselves, the world, and the future in overly negative and inaccurate ways. These distortions are associated with marked changes in an individual's mood, behavior, and language. We hypothesize that societies can undergo similar changes in their collective psychology that are reflected in historical records of language use. Here, we investigate the prevalence of textual markers of cognitive distortions in over 14 million books for the past 125 y and observe a surge of their prevalence since the 1980s, to levels exceeding those of the Great Depression and both World Wars. This pattern does not seem to be driven by changes in word meaning, publishing and writing standards, or the Google Books sample. Our results suggest a recent societal shift toward language associated with cognitive distortions and internalizing disorders.

Declining well-being during the COVID-19 pandemic reveals US social inequities.

BACKGROUND: The COVID-19 pandemic led to mental health fallout in the US; yet research about mental health and COVID-19 primarily rely on samples that may overlook variance in regional mental health. Indeed, between-city comparisons of mental health decline in the US may provide further insight into how the pandemic is disproportionately affecting at-risk groups. PURPOSE: This study leverages social media and COVID-19-city infection data to measure the longitudinal (January 22- July 31, 2020) mental health effects of the COVID-19 pandemic in 20 metropolitan areas. METHODS: We used longitudinal VADER sentiment analysis of Twitter timelines (January-July 2020) for cohorts in 20 metropolitan areas to examine mood changes over time. We then conducted simple and multivariate Ordinary Least Squares (OLS) regressions to examine the relationship between COVID-19 infection city data, population, population density, and city demographics on sentiment across those 20 cities. RESULTS: Longitudinal sentiment tracking showed mood declines over time. The univariate OLS regression highlighted a negative linear relationship between COVID-19 city data and online sentiment (ß = -.017). Residing in predominantly white cities had a protective effect against COVID-19 driven negative mood (ß = .0629, p < .001). DISCUSSION: Our results reveal that metropolitan areas with larger communities of color experienced a greater subjective well-being decline than predominantly white cities, which we attribute to clinical and socioeconomic correlates that place communities of color at greater risk of COVID-19. CONCLUSION: The COVID-19 pandemic is a driver of declining US mood in 20 metropolitan cities. Other factors, including social unrest and local demographics, may compound and exacerbate mental health outlook in racially diverse cities.

Individuals with depression express more distorted thinking on social media.

Depression is a leading cause of disability worldwide, but is often underdiagnosed and undertreated. Cognitive behavioural therapy holds that individuals with depression exhibit distorted modes of thinking, that is, cognitive distortions, that can negatively affect their emotions and motivation. Here, we show that the language of individuals with a self-reported diagnosis of depression on social media is characterized by higher levels of distorted thinking compared with a random sample. This effect is specific to the distorted nature of the expression and cannot be explained by the presence of specific topics, sentiment or first-person pronouns. This study identifies online language patterns that are indicative of depression-related distorted thinking. We caution that any future applications of this research should carefully consider ethical and data privacy issues.

Social Media Insights Into US Mental Health During the COVID-19 Pandemic: Longitudinal Analysis of Twitter Data.

BACKGROUND: The COVID-19 pandemic led to unprecedented mitigation efforts that disrupted the daily lives of millions. Beyond the general health repercussions of the pandemic itself, these measures also present a challenge to the world's mental health and health care systems. Considering that traditional survey methods are time-consuming and expensive, we need timely and proactive data sources to respond to the rapidly evolving effects of health policy on our population's mental health. Many people in the United States now use social media platforms such as Twitter to express the most minute details of their daily lives and social relations. This behavior is expected to increase during the COVID-19 pandemic, rendering social media data a rich field to understand personal well-being. OBJECTIVE: This study aims to answer three research questions: (1) What themes emerge from a corpus of US tweets about COVID-19? (2) To what extent did social media use increase during the onset of the COVID-19 pandemic? and (3) Does sentiment change in response to the COVID-19 pandemic? METHODS: We analyzed 86,581,237 public domain English language US tweets collected from an open-access public repository in three steps. First, we characterized the evolution of hashtags over time using latent Dirichlet allocation (LDA) topic modeling. Second, we increased the granularity of this analysis by downloading Twitter timelines of a large cohort of individuals (n=354,738) in 20 major US cities to assess changes in social media use. Finally, using this timeline data, we examined collective shifts in public mood in relation to evolving pandemic news cycles by analyzing the average daily sentiment of all timeline tweets with the Valence Aware Dictionary and Sentiment Reasoner (VADER) tool. RESULTS: LDA topics generated in the early months of the data set corresponded to major COVID-19-specific events. However, as state and municipal governments began issuing stay-at-home orders, latent themes shifted toward US-related lifestyle changes rather than global pandemic-related events. Social media volume also increased significantly, peaking during stay-at-home mandates. Finally, VADER sentiment analysis scores of user timelines were initially high and stable but decreased significantly, and continuously, by late March. CONCLUSIONS: Our findings underscore the negative effects of the pandemic on overall population sentiment. Increased use rates suggest that, for some, social media may be a coping mechanism to combat feelings of isolation related to long-term social distancing. However, in light of the documented negative effect of heavy social media use on mental health, social media may further exacerbate negative feelings in the long-term for many individuals. Thus, considering the overburdened US mental health care structure, these findings have important implications for ongoing mitigation efforts.

Depression alters the circadian pattern of online activity.

Human sleep/wake cycles follow a stable circadian rhythm associated with hormonal, emotional, and cognitive changes. Changes of this cycle are implicated in many mental health concerns. In fact, the bidirectional relation between major depressive disorder and sleep has been well-documented. Despite a clear link between sleep disturbances and subsequent disturbances in mood, it is difficult to determine from self-reported data which specific changes of the sleep/wake cycle play the most important role in this association. Here we observe marked changes of activity cycles in millions of twitter posts of 688 subjects who explicitly stated in unequivocal terms that they had received a (clinical) diagnosis of depression as compared to the activity cycles of a large control group (n = 8791). Rather than a phase-shift, as reported in other work, we find significant changes of activity levels in the evening and before dawn. Compared to the control group, depressed subjects were significantly more active from 7 PM to midnight and less active from 3 to 6 AM. Content analysis of tweets revealed a steady rise in rumination and emotional content from midnight to dawn among depressed individuals. These results suggest that diagnosis and treatment of depression may focus on modifying the timing of activity, reducing rumination, and decreasing social media use at specific hours of the day.

Mining Social Media Data for Biomedical Signals and Health-Related Behavior.

Social media data have been increasingly used to study biomedical and health-related phenomena. From cohort-level discussions of a condition to population-level analyses of sentiment, social media have provided scientists with unprecedented amounts of data to study human behavior associated with a variety of health conditions and medical treatments. Here we review recent work in mining social media for biomedical, epidemiological, and social phenomena information relevant to the multilevel complexity of human health. We pay particular attention to topics where social media data analysis has shown the most progress, including pharmacovigilance and sentiment analysis, especially for mental health. We also discuss a variety of innovative uses of social media data for health-related applications as well as important limitations of social media data access and use.