Evaluating generative AI responses to real-world drug-related questions.

Generative Artificial Intelligence (AI) systems such as OpenAI's ChatGPT, capable of an unprecedented ability to generate human-like text and converse in real time, hold potential for large-scale deployment in clinical settings such as substance use treatment. Treatment for substance use disorders (SUDs) is particularly high stakes, requiring evidence-based clinical treatment, mental health expertise, and peer support. Thus, promises of AI systems addressing deficient healthcare resources and structural bias are relevant within this domain, especially in an anonymous setting. This study explores the effectiveness of generative AI in answering real-world substance use and recovery questions. We collect questions from online recovery forums, use ChatGPT and Meta's LLaMA-2 for responses, and have SUD clinicians rate these AI responses. While clinicians rated the AI-generated responses as high quality, we discovered instances of dangerous disinformation, including disregard for suicidal ideation, incorrect emergency helplines, and endorsement of home detox. Moreover, the AI systems produced inconsistent advice depending on question phrasing. These findings indicate a risky mix of seemingly high-quality, accurate responses upon initial inspection that contain inaccurate and potentially deadly medical advice. Consequently, while generative AI shows promise, its real-world application in sensitive healthcare domains necessitates further safeguards and clinical validation.

The role of negative affect in shaping populist support: Converging field evidence from across the globe.

Support for populism has grown substantially during the past 2 decades, a development that has coincided with a marked increase in the experience of negative affect around the world. We use a multimodal, multimethod empirical approach, with data from a diverse set of geographical and political contexts, to investigate the extent to which the rising electoral demand for populism can be explained by negative affect. We demonstrate that negative affect-measured via (a) self-reported emotions in surveys as well as (b) automated text analyses of Twitter data-predicts individual-level populist attitudes in two global surveys (Studies 1a and 1b), longitudinal changes in populist party vote shares at general elections in Europe (Study 2), district-level Brexit voting in the 2016 U.K. referendum (Study 3), and county-level vote shares for Donald Trump in the 2016 and 2020 U.S. presidential elections (Studies 4a and 4b). We find that negative emotions-such as fear and anger as well as more often overlooked low-arousal negative emotions like depression and sadness-are predictive of populist beliefs as well as voting and election results at scale. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Robust language-based mental health assessments in time and space through social media.

In the most comprehensive population surveys, mental health is only broadly captured through questionnaires asking about "mentally unhealthy days" or feelings of "sadness." Further, population mental health estimates are predominantly consolidated to yearly estimates at the state level, which is considerably coarser than the best estimates of physical health. Through the large-scale analysis of social media, robust estimation of population mental health is feasible at finer resolutions. In this study, we created a pipeline that used ~1 billion Tweets from 2 million geo-located users to estimate mental health levels and changes for depression and anxiety, the two leading mental health conditions. Language-based mental health assessments (LBMHAs) had substantially higher levels of reliability across space and time than available survey measures. This work presents reliable assessments of depression and anxiety down to the county-weeks level. Where surveys were available, we found moderate to strong associations between the LBMHAs and survey scores for multiple levels of granularity, from the national level down to weekly county measurements (fixed effects ß = 0.34 to 1.82; p < 0.001). LBMHAs demonstrated temporal validity, showing clear absolute increases after a list of major societal events (+23% absolute change for depression assessments). LBMHAs showed improved external validity, evidenced by stronger correlations with measures of health and socioeconomic status than population surveys. This study shows that the careful aggregation of social media data yields spatiotemporal estimates of population mental health that exceed the granularity achievable by existing population surveys, and does so with generally greater reliability and validity.

Key language markers of depression on social media depend on race.

Depression has robust natural language correlates and can increasingly be measured in language using predictive models. However, despite evidence that language use varies as a function of individual demographic features (e.g., age, gender), previous work has not systematically examined whether and how depression's association with language varies by race. We examine how race moderates the relationship between language features (i.e., first-person pronouns and negative emotions) from social media posts and self-reported depression, in a matched sample of Black and White English speakers in the United States. Our findings reveal moderating effects of race: While depression severity predicts I-usage in White individuals, it does not in Black individuals. White individuals use more belongingness and self-deprecation-related negative emotions. Machine learning models trained on similar amounts of data to predict depression severity performed poorly when tested on Black individuals, even when they were trained exclusively using the language of Black individuals. In contrast, analogous models tested on White individuals performed relatively well. Our study reveals surprising race-based differences in the expression of depression in natural language and highlights the need to understand these effects better, especially before language-based models for detecting psychological phenomena are integrated into clinical practice.

Characterizing empathy and compassion using computational linguistic analysis.

Many scholars have proposed that feeling what we believe others are feeling-often known as "empathy"-is essential for other-regarding sentiments and plays an important role in our moral lives. Caring for and about others (without necessarily sharing their feelings)-often known as "compassion"-is also frequently discussed as a relevant force for prosocial motivation and action. Here, we explore the relationship between empathy and compassion using the methods of computational linguistics. Analyses of 2,356,916 Facebook posts suggest that individuals (N = 2,781) high in empathy use different language than those high in compassion, after accounting for shared variance between these constructs. Empathic people, controlling for compassion, often use self-focused language and write about negative feelings, social isolation, and feeling overwhelmed. Compassionate people, controlling for empathy, often use other-focused language and write about positive feelings and social connections. In addition, high empathy without compassion is related to negative health outcomes, while high compassion without empathy is related to positive health outcomes, positive lifestyle choices, and charitable giving. Such findings favor an approach to moral motivation that is grounded in compassion rather than empathy. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

A linguistic analysis of dehumanization toward substance use across three decades of news articles.

Introduction: Substances and the people who use them have been dehumanized for decades. As a result, lawmakers and healthcare providers have implemented policies that subjected millions to criminalization, incarceration, and inadequate resources to support health and wellbeing. While there have been recent shifts in public opinion on issues such as legalization, in the case of marijuana in the U.S., or addiction as a disease, dehumanization and stigma are still leading barriers for individuals seeking treatment. Integral to the narrative of "substance users" as thoughtless zombies or violent criminals is their portrayal in popular media, such as films and news. Methods: This study attempts to quantify the dehumanization of people who use substances (PWUS) across time using a large corpus of over 3 million news articles. We apply a computational linguistic framework for measuring dehumanization across three decades of New York Times articles. Results: We show that (1) levels of dehumanization remain high and (2) while marijuana has become less dehumanized over time, attitudes toward other substances such as heroin and cocaine remain stable. Discussion: This work highlights the importance of a holistic view of substance use that places all substances within the context of addiction as a disease, prioritizes the humanization of PWUS, and centers around harm reduction.

Smartphone sensor data estimate alcohol craving in a cohort of patients with alcohol-associated liver disease and alcohol use disorder.

BACKGROUND: Sensors within smartphones, such as accelerometer and location, can describe longitudinal markers of behavior as represented through devices in a method called digital phenotyping. This study aimed to assess the feasibility of digital phenotyping for patients with alcohol-associated liver disease and alcohol use disorder, determine correlations between smartphone data and alcohol craving, and establish power assessment for future studies to prognosticate clinical outcomes. METHODS: A total of 24 individuals with alcohol-associated liver disease and alcohol use disorder were instructed to download the AWARE application to collect continuous sensor data and complete daily ecological momentary assessments on alcohol craving and mood for up to 30 days. Data from sensor streams were processed into features like accelerometer magnitude, number of calls, and location entropy, which were used for statistical analysis. We used repeated measures correlation for longitudinal data to evaluate associations between sensors and ecological momentary assessments and standard Pearson correlation to evaluate within-individual relationships between sensors and craving. RESULTS: Alcohol craving significantly correlated with mood obtained from ecological momentary assessments. Across all sensors, features associated with craving were also significantly correlated with all moods (eg, loneliness and stress) except boredom. Individual-level analysis revealed significant relationships between craving and features of location entropy and average accelerometer magnitude. CONCLUSIONS: Smartphone sensors may serve as markers for alcohol craving and mood in alcohol-associated liver disease and alcohol use disorder. Findings suggest that location-based and accelerometer-based features may be associated with alcohol craving. However, data missingness and low participant retention remain challenges. Future studies are needed for further digital phenotyping of relapse risk and progression of liver disease.

Filling in the white space: Spatial interpolation with Gaussian processes and social media data.

Full national coverage below the state level is difficult to attain through survey-based data collection. Even the largest survey-based data collections, such as the CDC's Behavioral Risk Factor Surveillance System or the Gallup-Healthways Well-being Index (both with more than 300,000 responses p.a.) only allow for the estimation of annual averages for about 260 out of roughly U.S. 3,000 counties when a threshold of 300 responses per county is used. Using a relatively high threshold of 300 responses gives substantially higher convergent validity-higher correlations with health variables-than lower thresholds but covers a reduced and biased sample of the population. We present principled methods to interpolate spatial estimates and show that including large-scale geotagged social media data can increase interpolation accuracy. In this work, we focus on Gallup-reported life satisfaction, a widely-used measure of subjective well-being. We use Gaussian Processes (GP), a formal Bayesian model, to interpolate life satisfaction, which we optimally combine with estimates from low-count data. We interpolate over several spaces (geographic and socioeconomic) and extend these evaluations to the space created by variables encoding language frequencies of approximately 6 million geotagged Twitter users. We find that Twitter language use can serve as a rough aggregate measure of socioeconomic and cultural similarity, and improves upon estimates derived from a wide variety of socioeconomic, demographic, and geographic similarity measures. We show that applying Gaussian Processes to the limited Gallup data allows us to generate estimates for a much larger number of counties while maintaining the same level of convergent validity with external criteria (i.e., N = 1,133 vs. 2,954 counties). This work suggests that spatial coverage of psychological variables can be reliably extended through Bayesian techniques while maintaining out-of-sample prediction accuracy and that Twitter language adds important information about cultural similarity over and above traditional socio-demographic and geographic similarity measures. Finally, to facilitate the adoption of these methods, we have also open-sourced an online tool that researchers can freely use to interpolate their data across geographies.

Extended impact of the COVID-19 pandemic: Trajectories of mental health and substance use among U.S. adults, September 2020-August 2021.

Background: Americans reported significant increases in mental health and substance use problems after the COVID-19 pandemic outbreak. This can be a product of the pandemic disruptions in everyday life, with some populations being more impacted than others. Objectives: To assess the ongoing impact of the COVID-19 pandemic on mental health and substance use in U.S. adults from September 2020 to August 2021. Methods: Participants included 1056 adults (68.5% women) who participated in a national longitudinal online survey assessing the perceived impact of COVID-19 on daily life, stress, depression and anxiety symptoms, and alcohol and cannabis use at 3-time points from September 2020 to August 2021. Results: Individuals with lower self-reported social status reported the highest perceived impact. Participants' perceived impact of the COVID-19 pandemic on daily life, stress, anxiety, and alcohol use risk significantly decreased over time but remained high. However, there was no change in depressive symptoms and cannabis use. Higher levels of perceived impact of the pandemic significantly predicted both more baseline mental health concerns and lower decreases over time. Lower self-report social status predicted more baseline mental health concerns and smaller decreases in those concerns. Black adults reported significantly higher cannabis use rates than non-Hispanic White adults. Conclusion: The impact of COVID-19 on daily life continued to be a risk factor for mental health during the second wave of the pandemic. In addition to infection prevention, public health policies should focus on pandemic-related social factors such as economic concerns and caretaking that continue to affect mental health.

Predicting U.S. county opioid poisoning mortality from multi-modal social media and psychological self-report data.

Opioid poisoning mortality is a substantial public health crisis in the United States, with opioids involved in approximately 75% of the nearly 1 million drug related deaths since 1999. Research suggests that the epidemic is driven by both over-prescribing and social and psychological determinants such as economic stability, hopelessness, and isolation. Hindering this research is a lack of measurements of these social and psychological constructs at fine-grained spatial and temporal resolutions. To address this issue, we use a multi-modal data set consisting of natural language from Twitter, psychometric self-reports of depression and well-being, and traditional area-based measures of socio-demographics and health-related risk factors. Unlike previous work using social media data, we do not rely on opioid or substance related keywords to track community poisonings. Instead, we leverage a large, open vocabulary of thousands of words in order to fully characterize communities suffering from opioid poisoning, using a sample of 1.5 billion tweets from 6 million U.S. county mapped Twitter users. Results show that Twitter language predicted opioid poisoning mortality better than factors relating to socio-demographics, access to healthcare, physical pain, and psychological well-being. Additionally, risk factors revealed by the Twitter language analysis included negative emotions, discussions of long work hours, and boredom, whereas protective factors included resilience, travel/leisure, and positive emotions, dovetailing with results from the psychometric self-report data. The results show that natural language from public social media can be used as a surveillance tool for both predicting community opioid poisonings and understanding the dynamic social and psychological nature of the epidemic.

The text-package: An R-package for analyzing and visualizing human language using natural language processing and transformers.

The language that individuals use for expressing themselves contains rich psychological information. Recent significant advances in Natural Language Processing (NLP) and Deep Learning (DL), namely transformers, have resulted in large performance gains in tasks related to understanding natural language. However, these state-of-the-art methods have not yet been made easily accessible for psychology researchers, nor designed to be optimal for human-level analyses. This tutorial introduces text (https://r-text.org/), a new R-package for analyzing and visualizing human language using transformers, the latest techniques from NLP and DL. The text-package is both a modular solution for accessing state-of-the-art language models and an end-to-end solution catered for human-level analyses. Hence, text provides user-friendly functions tailored to test hypotheses in social sciences for both relatively small and large data sets. The tutorial describes methods for analyzing text, providing functions with reliable defaults that can be used off-the-shelf as well as providing a framework for the advanced users to build on for novel pipelines. The reader learns about three core methods: (1) textEmbed(): to transform text to modern transformer-based word embeddings; (2) textTrain() and textPredict(): to train predictive models with embeddings as input, and use the models to predict from; (3) textSimilarity() and textDistance(): to compute semantic similarity/distance scores between texts. The reader also learns about two extended methods: (1) textProjection()/textProjectionPlot() and (2) textCentrality()/textCentralityPlot(): to examine and visualize text within the embedding space. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Pandemic distress associated with segregation and social stressors.

Background: Racial/ethnic minorities are disproportionately impacted by the COVID-19 pandemic, as they are more likely to experience structural and interpersonal racial discrimination, and thus social marginalization. Based on this, we tested for associations between pandemic distress outcomes and four exposures: racial segregation, coronavirus-related racial bias, social status, and social support. Methods: Data were collected as part of a larger longitudinal national study on mental health during the pandemic (n = 1,309). We tested if county-level segregation and individual-level social status, social support, and coronavirus racial bias were associated with pandemic distress using cumulative ordinal regression models, both unadjusted and adjusted for covariates (gender, age, education, and income). Results: Both the segregation index (PR = 1.19; 95% CI 1.03, 1.36) and the coronavirus racial bias scale (PR = 1.17; 95% CI 1.06, 1.29) were significantly associated with pandemic distress. Estimates were similar, after adjusting for covariates, for both segregation (aPR = 1.15; 95% CI 1.01, 1.31) and coronavirus racial bias (PR = 1.12; 95% CI 1.02, 1.24). Higher social status (aPR = 0.74; 95% CI 0.64, 0.86) and social support (aPR = 0.81; 95% CI 0.73, 0.90) were associated with lower pandemic distress after adjustment. Conclusion: Segregation and coronavirus racial bias are relevant pandemic stressors, and thus have implications for minority health. Future research exploring potential mechanisms of this relationship, including specific forms of racial discrimination related to pandemic distress and implications for social justice efforts, are recommended.

AI-based analysis of social media language predicts addiction treatment dropout at 90 days.

The reoccurrence of use (relapse) and treatment dropout is frequently observed in substance use disorder (SUD) treatment. In the current paper, we evaluated the predictive capability of an AI-based digital phenotype using the social media language of patients receiving treatment for substance use disorders (N = 269). We found that language phenotypes outperformed a standard intake psychometric assessment scale when predicting patients' 90-day treatment outcomes. We also use a modern deep learning-based AI model, Bidirectional Encoder Representations from Transformers (BERT) to generate risk scores using pre-treatment digital phenotype and intake clinic data to predict dropout probabilities. Nearly all individuals labeled as low-risk remained in treatment while those identified as high-risk dropped out (risk score for dropout AUC = 0.81; p < 0.001). The current study suggests the possibility of utilizing social media digital phenotypes as a new tool for intake risk assessment to identify individuals most at risk of treatment dropout and relapse.

Differences in mental health and alcohol use across profiles of COVID-19 disruptions.

This study aimed to examine differences in mental health and alcohol use outcomes across distinct patterns of work, home, and social life disruptions associated with the COVID-19 pandemic. Data from 2093 adult participants were collected from September 2020 to April 2021 as a part of a larger study examining the impacts of the COVID-19 pandemic on substance use. Participants provided data on COVID-19 pandemic experiences, mental health outcomes, media consumption, and alcohol use at baseline. Alcohol use difficulties, including problems related to the use, desire to use alcohol, failure to cut down on alcohol use, and family/friend concern with alcohol use, were measured at 60-day follow-up. Factor mixture modeling followed by group comparisons, multiple linear regressions, and multiple logistic regressions was conducted. A four-profile model was selected. Results indicated that profile membership predicted differences in mental health and alcohol use outcomes above and beyond demographics. Individuals experiencing the most disruption reported the strongest daily impact of COVID-19 and significantly high levels of depression, anxiety, loneliness, overwhelm, alcohol use at baseline, and alcohol use difficulties measured at 60-day follow-up. The findings highlight the need for integrated mental health and/or alcohol services and social services targeting work, home, and social life during public health emergencies in order to respond effectively and comprehensively to the needs of those requiring different types of support.

Measuring disadvantage: A systematic comparison of United States small-area disadvantage indices.

Extensive evidence demonstrates the effects of area-based disadvantage on a variety of life outcomes, such as increased mortality and low economic mobility. Despite these well-established patterns, disadvantage, often measured using composite indices, is inconsistently operationalized across studies. To address this issue, we systematically compared 5 U.S. disadvantage indices at the county-level on their relationships to 24 diverse life outcomes related to mortality, physical health, mental health, subjective well-being, and social capital from heterogeneous data sources. We further examined which domains of disadvantage are most important when creating these indices. Of the five indices examined, the Area Deprivation Index (ADI) and Child Opportunity Index 2.0 (COI) were most related to a diverse set of life outcomes, particularly physical health. Within each index, variables from the domains of education and employment were most important in relationships with life outcomes. Disadvantage indices are being used in real-world policy and resource allocation decisions; an index's generalizability across diverse life outcomes, and the domains of disadvantage which constitute the index, should be considered when guiding such decisions.

Opioid death projections with AI-based forecasts using social media language.

Targeting of location-specific aid for the U.S. opioid epidemic is difficult due to our inability to accurately predict changes in opioid mortality across heterogeneous communities. AI-based language analyses, having recently shown promise in cross-sectional (between-community) well-being assessments, may offer a way to more accurately longitudinally predict community-level overdose mortality. Here, we develop and evaluate, TROP (Transformer for Opiod Prediction), a model for community-specific trend projection that uses community-specific social media language along with past opioid-related mortality data to predict future changes in opioid-related deaths. TOP builds on recent advances in sequence modeling, namely transformer networks, to use changes in yearly language on Twitter and past mortality to project the following year's mortality rates by county. Trained over five years and evaluated over the next two years TROP demonstrated state-of-the-art accuracy in predicting future county-specific opioid trends. A model built using linear auto-regression and traditional socioeconomic data gave 7% error (MAPE) or within 2.93 deaths per 100,000 people on average; our proposed architecture was able to forecast yearly death rates with less than half that error: 3% MAPE and within 1.15 per 100,000 people.