Social signals predict contemporary art prices better than visual features, particularly in emerging markets.

What determines the price of an artwork? This article leverages a comprehensive and novel dataset on art auctions of contemporary artists to examine the impact of social and visual features on the valuation of artworks across global markets. Our findings indicate that social signals allow us to predict the price of artwork exceptionally well, even approaching the professionals' prediction accuracy, while the visual features play a marginal role. This pattern is especially pronounced in emerging markets, supporting the idea that social signals become more critical when it is more difficult to assess the quality. These results strongly support that the value of artwork is largely shaped by social factors, particularly in emerging markets where a stronger preference for "buying an artist" than "buying an artwork." Additionally, our study shows that it is possible to boost experts' performance, highlighting the potential benefits of human-machine models in uncertain or rapidly changing markets, where expert knowledge is limited.

Emergence of simple and complex contagion dynamics from weighted belief networks.

Social contagion is a ubiquitous and fundamental process that drives individual and social changes. Although social contagion arises as a result of cognitive processes and biases, the integration of cognitive mechanisms with the theory of social contagion remains an open challenge. In particular, studies on social phenomena usually assume contagion dynamics to be either simple or complex, rather than allowing it to emerge from cognitive mechanisms, despite empirical evidence indicating that a social system can exhibit a spectrum of contagion dynamics-from simple to complex-simultaneously. Here, we propose a model of interacting beliefs, from which both simple and complex contagion dynamics can organically arise. Our model also elucidates how a fundamental mechanism of complex contagion-resistance-can come about from cognitive mechanisms.

Unsupervised embedding of trajectories captures the latent structure of scientific migration.

Human migration and mobility drives major societal phenomena including epidemics, economies, innovation, and the diffusion of ideas. Although human mobility and migration have been heavily constrained by geographic distance throughout the history, advances, and globalization are making other factors such as language and culture increasingly more important. Advances in neural embedding models, originally designed for natural language, provide an opportunity to tame this complexity and open new avenues for the study of migration. Here, we demonstrate the ability of the model word2vec to encode nuanced relationships between discrete locations from migration trajectories, producing an accurate, dense, continuous, and meaningful vector-space representation. The resulting representation provides a functional distance between locations, as well as a "digital double" that can be distributed, re-used, and itself interrogated to understand the many dimensions of migration. We show that the unique power of word2vec to encode migration patterns stems from its mathematical equivalence with the gravity model of mobility. Focusing on the case of scientific migration, we apply word2vec to a database of three million migration trajectories of scientists derived from the affiliations listed on their publication records. Using techniques that leverage its semantic structure, we demonstrate that embeddings can learn the rich structure that underpins scientific migration, such as cultural, linguistic, and prestige relationships at multiple levels of granularity. Our results provide a theoretical foundation and methodological framework for using neural embeddings to represent and understand migration both within and beyond science.

The Social Contagion Potential of Pro-Vaccine Messages on Black Twitter.

Black Americans in the US not only suffered from disproportionately high hospitalization and death rates throughout the pandemic but also from the consequences of low COVID-19 vaccination rates. This pattern of disparity is linked to distrust of public health systems that originates from a history of medical atrocities committed against Black people. For that reason, mitigation of race-based inequity in COVID-19 impacts might find more success in grassroots information contagion than official public health campaigns. While Black Twitter is well-positioned as a conduit for such information contagion, little is known about message characteristics that would afford it. Here, we tested the impact of four different message frames (personalization, interactive, fear appeal, neutral) on the social contagion potential of bi-modal social media messages promoting COVID-19 vaccinations and finding personalized messages to be the most shareable. Wary of recommending personalization as the blueprint for setting a social contagion health campaign in motion, we probed further to understand the influence of individual-level variables on the communicability of personalized messages. Subsequently, regression models and focus group data were consulted, revealing that thinking styles, vaccine confidence levels, and attitudes toward social media were significant factors of influence on the contagion potential of personalized messages. We discussed the implications of these results for health campaigns.

Development and validation of a model for measuring alcohol consumption from transdermal alcohol content data among college students.

BACKGROUND AND AIMS: Transdermal alcohol content (TAC) data collected by wearable alcohol monitors could potentially contribute to alcohol research, but raw data from the devices are challenging to interpret. We aimed to develop and validate a model using TAC data to detect alcohol drinking. DESIGN: We used a model development and validation study design. SETTING: Indiana, USA PARTICIPANTS: In March to April 2021, we enrolled 84 college students who reported drinking at least once a week (median age = 20 years, 73% white, 70% female). We observed participants' alcohol drinking behavior for 1 week. MEASUREMENTS: Participants wore BACtrack Skyn monitors (TAC data), provided self-reported drinking start times in real time (smartphone app) and completed daily surveys about their prior day of drinking. We developed a model using signal filtering, peak detection algorithm, regression and hyperparameter optimization. The input was TAC and outputs were alcohol drinking frequency, start time and magnitude. We validated the model using daily surveys (internal validation) and data collected from college students in 2019 (external validation). FINDINGS: Participants (N = 84) self-reported 213 drinking events. Monitors collected 10 915 hours of TAC. In internal validation, the model had a sensitivity of 70.9% (95% CI = 64.1%-77.0%) and a specificity of 73.9% (68.9%-78.5%) in detecting drinking events. The median absolute time difference between self-reported and model-detected drinking start times was 59 min. Mean absolute error (MAE) for the reported and detected number of drinks was 2.8 drinks. In an exploratory external validation among five participants, number of drinking events, sensitivity, specificity, median time difference and MAE were 15%, 67%, 100%, 45 minutes and 0.9 drinks, respectively. Our model's output was correlated with breath alcohol concentration data (Spearman's correlation [95% CI] = 0.88 [0.77, 0.94]). CONCLUSION: This study, the largest of its kind to date, developed and validated a model for detecting alcohol drinking using transdermal alcohol content data collected with a new generation of alcohol monitors. The model and its source code are available as Supporting Information (https://osf.io/xngbk).

Comparing measures of centrality in bipartite patient-prescriber networks: A study of drug seeking for opioid analgesics.

Visiting multiple prescribers is a common method for obtaining prescription opioids for nonmedical use and has played an important role in fueling the United States opioid epidemic, leading to increased drug use disorder and overdose. Recent studies show that centrality of the bipartite network formed by prescription ties between patients and prescribers of opioids is a promising indicator for drug seeking. However, node prominence in bipartite networks is typically estimated with methods that do not fully account for the two-mode topology of the underlying network. Although several algorithms have been proposed recently to address this challenge, it is unclear how these algorithms perform on real-world networks. Here, we compare their performance in the context of identifying opioid drug seeking behaviors by applying them to massive bipartite networks of patients and providers extracted from insurance claims data. We find that two variants of bipartite centrality are significantly better predictors of subsequent opioid overdose than traditional centrality estimates. Moreover, we show that incorporating non-network attributes such as the potency of the opioid prescriptions into the measures can further improve their performance. These findings can be reproduced on different datasets. Our results demonstrate the potential of bipartiteness-aware indices for identifying patterns of high-risk behavior.

The latent structure of global scientific development.

Science is essential to innovation and economic prosperity. Although studies have shown that national scientific development is affected by geographic, historic and economic factors, it remains unclear whether there are universal structures and trajectories of national scientific development that can inform forecasting and policy-making. Here, by examining the scientific 'exports'-publications that are indexed in international databases-of countries, we reveal a three-cluster structure in the relatedness network of disciplines that underpin national scientific development and the organization of global science. Tracing the evolution of national research portfolios reveals that while nations are proceeding to more diverse research profiles individually, scientific production is increasingly specialized in global science over the past decades. By uncovering the underlying structure of scientific development and connecting it with economic development, our results may offer a new perspective on the evolution of global science.

Misinformation, believability, and vaccine acceptance over 40 countries: Takeaways from the initial phase of the COVID-19 infodemic.

The COVID-19 pandemic has been damaging to the lives of people all around the world. Accompanied by the pandemic is an infodemic, an abundant and uncontrolled spread of potentially harmful misinformation. The infodemic may severely change the pandemic's course by interfering with public health interventions such as wearing masks, social distancing, and vaccination. In particular, the impact of the infodemic on vaccination is critical because it holds the key to reverting to pre-pandemic normalcy. This paper presents findings from a global survey on the extent of worldwide exposure to the COVID-19 infodemic, assesses different populations' susceptibility to false claims, and analyzes its association with vaccine acceptance. Based on responses gathered from over 18,400 individuals from 40 countries, we find a strong association between perceived believability of COVID-19 misinformation and vaccination hesitancy. Our study shows that only half of the online users exposed to rumors might have seen corresponding fact-checked information. Moreover, depending on the country, between 6% and 37% of individuals considered these rumors believable. A key finding of this research is that poorer regions were more susceptible to encountering and believing COVID-19 misinformation; countries with lower gross domestic product (GDP) per capita showed a substantially higher prevalence of misinformation. We discuss implications of our findings to public campaigns that proactively spread accurate information to countries that are more susceptible to the infodemic. We also defend that fact-checking platforms should prioritize claims that not only have wide exposure but are also perceived to be believable. Our findings give insights into how to successfully handle risk communication during the initial phase of a future pandemic.

New means, new measures: assessing prescription drug-seeking indicators over 10 years of the opioid epidemic.

BACKGROUND AND AIMS: Prescription drug-seeking (PDS) from multiple prescribers is a primary means of obtaining prescription opioids; however, PDS behavior has probably evolved in response to policy shifts, and there is little agreement about how to operationalize it. We systematically compared the performance of traditional and novel PDS indicators. DESIGN: Longitudinal study using a de-identified commercial claims database. SETTING: United States, 2009-18. PARTICIPANTS: A total of 318 million provider visits from 21.5 million opioid-prescribed patients. MEASUREMENTS: We applied binary classification and generalized linear models to compare predictive accuracy and average marginal effect size predicting future opioid use disorder (OUD), overdose and high morphine milligram equivalents (MME). We compared traditional indicators of PDS to a network centrality measure, PageRank, that reflects the prominence of patients in a co-prescribing network. Analyses used the same data and adjusted for patient demographics, region, SES, diagnoses and health services. FINDINGS: The predictive accuracy of a widely used traditional measure (N + unique doctors and N + unique pharmacies in 90 days) on OUD, overdose and MME decreased between 2009 and 2018, and performed no better than chance (50% accuracy) after 2015. Binarized PageRank measures however exhibited higher predictive accuracy than the traditional binary measures throughout 2009-2018. Continuous indicators of PDS performed better than binary thresholds, with days of Rx performing best overall with 77-93% predictive accuracy. For example, days of Rx had the highest average marginal effects on overdose and OUD: a 1 standard deviation increase in days of Rx was associated with a 6-8% [confidence intervals (CIs) = 0.058-0.061 and 0.078-0.082] increase in the probability of overdose and a 4-5% (CIs = 0.038-0.043 and 0.047-0.053) increase in the probability of OUD. PageRank performed nearly as well or better than traditional indicators of PDS, with predictive performance increasing after 2016. CONCLUSIONS: In the United States, network-based measures appear to have increasing promise for identifying prescription opioid drug-seeking behavior, while indicators based on quantity of providers or pharmacies appear to have decreasing utility.

Substitution of Nonpharmacologic Therapy With Opioid Prescribing for Pain During the COVID-19 Pandemic.

Importance: During the pandemic, access to medical care unrelated to COVID-19 was limited because of concerns about viral spread and corresponding policies. It is critical to assess how these conditions affected modes of pain treatment, given the addiction risks of prescription opioids. Objective: To assess the trends in opioid prescription and nonpharmacologic therapy (ie, physical therapy and complementary medicine) for pain management during the COVID-19 pandemic in 2020 compared with the patterns in 2019. Design, Setting, and Participants: This retrospective, cross-sectional study used weekly claims data from 24 million US patients in a nationwide commercial insurance database (Optum's deidentified Clinformatics Data Mart Database) from January 1, 2019, to September 31, 2020. Among patients with diagnoses of limb, extremity, or joint pain, back pain, and neck pain for each week, patterns of treatment use were identified and evaluated. Data analysis was performed from April 1, 2021, to September 31, 2021. Main Outcomes and Measures: The main outcomes of interest were weekly rates of opioid prescriptions, the strength and duration of related opioid prescriptions, and the use of nonpharmacologic therapy. Transition rates between different treatment options before the outbreak and during the early months of the pandemic were also assessed. Results: A total of 21 430 339 patients (mean [SD] age, 48.6 [24.0] years; 10 960 507 [51.1%] female; 909 061 [4.2%] Asian, 1 688 690 [7.9%] Black, 2 276 075 [10.6%] Hispanic, 11 192 789 [52.2%] White, and 5 363 724 [25.0%] unknown) were enrolled during the first 3 quarters in 2019 and 20 759 788 (mean [SD] age, 47.0 [23.8] years; 10 695 690 [51.5%] female; 798 037 [3.8%] Asian; 1 508 023 [7.3%] Black, 1 976 248 [9.5%] Hispanic, 10 059 597 [48.5%] White, and 6 417 883 [30.9%] unknown) in the first 3 quarters of 2020. During the COVID-19 pandemic, the proportion of patients receiving a pain diagnosis was smaller than that for the same period in 2019 (mean difference, -15.9%; 95% CI, -16.1% to -15.8%). Patients with pain were more likely to receive opioids (mean difference, 3.5%; 95% CI, 3.3%-3.7%) and less likely to receive nonpharmacologic therapy (mean difference, -6.0%; 95% CI, -6.3% to -5.8%), and opioid prescriptions were longer and more potent during the early pandemic in 2020 relative to 2019 (mean difference, 1.07 days; 95% CI, 1.02-1.17 days; mean difference, 0.96 morphine milligram equivalents; 95% CI, 0.76-1.20). Analysis of individuals' transitions between treatment options for pain found that patients were more likely to transition out of nonpharmacologic therapy, replacing it with opioid prescriptions for pain management during the COVID-19 pandemic than in the year before. Conclusions and Relevance: Nonpharmacologic therapy is a benign treatment for pain often recommended instead of opioid therapy. The decrease in nonpharmacologic therapy and increase in opioid prescription during the COVID-19 pandemic found in this cross-sectional study, especially given longer days of prescription and more potent doses, may exacerbate the US opioid epidemic. These findings suggest that it is imperative to investigate the implications of limited medical access on treatment substitution, which may increase patient risk, and implement policies and guidelines to prevent those substitutions.

Characterizing partisan political narrative frameworks about COVID-19 on Twitter.

The COVID-19 pandemic is a global crisis that has been testing every society and exposing the critical role of local politics in crisis response. In the United States, there has been a strong partisan divide between the Democratic and Republican party's narratives about the pandemic which resulted in polarization of individual behaviors and divergent policy adoption across regions. As shown in this case, as well as in most major social issues, strongly polarized narrative frameworks facilitate such narratives. To understand polarization and other social chasms, it is critical to dissect these diverging narratives. Here, taking the Democratic and Republican political social media posts about the pandemic as a case study, we demonstrate that a combination of computational methods can provide useful insights into the different contexts, framing, and characters and relationships that construct their narrative frameworks which individual posts source from. Leveraging a dataset of tweets from the politicians in the U.S., including the ex-president, members of Congress, and state governors, we found that the Democrats' narrative tends to be more concerned with the pandemic as well as financial and social support, while the Republicans discuss more about other political entities such as China. We then perform an automatic framing analysis to characterize the ways in which they frame their narratives, where we found that the Democrats emphasize the government's role in responding to the pandemic, and the Republicans emphasize the roles of individuals and support for small businesses. Finally, we present a semantic role analysis that uncovers the important characters and relationships in their narratives as well as how they facilitate a membership categorization process. Our findings concretely expose the gaps in the "elusive consensus" between the two parties. Our methodologies may be applied to computationally study narratives in various domains. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1140/epjds/s13688-021-00308-4.

FrameAxis: characterizing microframe bias and intensity with word embedding.

Framing is a process of emphasizing a certain aspect of an issue over the others, nudging readers or listeners towards different positions on the issue even without making a biased argument. Here, we propose FrameAxis, a method for characterizing documents by identifying the most relevant semantic axes ("microframes") that are overrepresented in the text using word embedding. Our unsupervised approach can be readily applied to large datasets because it does not require manual annotations. It can also provide nuanced insights by considering a rich set of semantic axes. FrameAxis is designed to quantitatively tease out two important dimensions of how microframes are used in the text. Microframe bias captures how biased the text is on a certain microframe, and microframe intensity shows how prominently a certain microframe is used. Together, they offer a detailed characterization of the text. We demonstrate that microframes with the highest bias and intensity align well with sentiment, topic, and partisan spectrum by applying FrameAxis to multiple datasets from restaurant reviews to political news. The existing domain knowledge can be incorporated into FrameAxis by using custom microframes and by using FrameAxis as an iterative exploratory analysis instrument. Additionally, we propose methods for explaining the results of FrameAxis at the level of individual words and documents. Our method may accelerate scalable and sophisticated computational analyses of framing across disciplines.

The effectiveness of backward contact tracing in networks.

Effective control of an epidemic relies on the rapid discovery and isolation of infected individuals. Because many infectious diseases spread through interaction, contact tracing is widely used to facilitate case discovery and control. However, what determines the efficacy of contact tracing has not been fully understood. Here we reveal that, compared with 'forward' tracing (tracing to whom disease spreads), 'backward' tracing (tracing from whom disease spreads) is profoundly more effective. The effectiveness of backward tracing is due to simple but overlooked biases arising from the heterogeneity in contacts. We argue that, even if the directionality of infection is unknown, it is possible to perform backward-aiming contact tracing. Using simulations on both synthetic and high-resolution empirical contact datasets, we show that strategically executed contact tracing can prevent a substantial fraction of transmissions with a higher efficiency-in terms of prevented cases per isolation-than case isolation alone. Our results call for a revision of current contact-tracing strategies so that they leverage all forms of bias. It is particularly crucial that we incorporate backward and deep tracing in a digital context while adhering to the privacy-preserving requirements of these new platforms.

Principled approach to the selection of the embedding dimension of networks.

Network embedding is a general-purpose machine learning technique that encodes network structure in vector spaces with tunable dimension. Choosing an appropriate embedding dimension - small enough to be efficient and large enough to be effective - is challenging but necessary to generate embeddings applicable to a multitude of tasks. Existing strategies for the selection of the embedding dimension rely on performance maximization in downstream tasks. Here, we propose a principled method such that all structural information of a network is parsimoniously encoded. The method is validated on various embedding algorithms and a large corpus of real-world networks. The embedding dimension selected by our method in real-world networks suggest that efficient encoding in low-dimensional spaces is usually possible.

Neural embeddings of scholarly periodicals reveal complex disciplinary organizations.

Understanding the structure of knowledge domains is one of the foundational challenges in the science of science. Here, we propose a neural embedding technique that leverages the information contained in the citation network to obtain continuous vector representations of scientific periodicals. We demonstrate that our periodical embeddings encode nuanced relationships between periodicals and the complex disciplinary and interdisciplinary structure of science, allowing us to make cross-disciplinary analogies between periodicals. Furthermore, we show that the embeddings capture meaningful "axes" that encompass knowledge domains, such as an axis from "soft" to "hard" sciences or from "social" to "biological" sciences, which allow us to quantitatively ground periodicals on a given dimension. By offering novel quantification in the science of science, our framework may, in turn, facilitate the study of how knowledge is created and organized.

Persona2vec: a flexible multi-role representations learning framework for graphs.

Graph embedding techniques, which learn low-dimensional representations of a graph, are achieving state-of-the-art performance in many graph mining tasks. Most existing embedding algorithms assign a single vector to each node, implicitly assuming that a single representation is enough to capture all characteristics of the node. However, across many domains, it is common to observe pervasively overlapping community structure, where most nodes belong to multiple communities, playing different roles depending on the contexts. Here, we propose persona2vec, a graph embedding framework that efficiently learns multiple representations of nodes based on their structural contexts. Using link prediction-based evaluation, we show that our framework is significantly faster than the existing state-of-the-art model while achieving better performance.

Systematic Evaluation of State Policy Interventions Targeting the US Opioid Epidemic, 2007-2018.

Importance: In response to the increase in opioid overdose deaths in the United States, many states recently have implemented supply-controlling and harm-reduction policy measures. To date, an updated policy evaluation that considers the full policy landscape has not been conducted. Objective: To evaluate 6 US state-level drug policies to ascertain whether they are associated with a reduction in indicators of prescription opioid abuse, the prevalence of opioid use disorder and overdose, the prescription of medication-assisted treatment (MAT), and drug overdose deaths. Design, Setting, and Participants: This cross-sectional study used drug overdose mortality data from 50 states obtained from the National Vital Statistics System and claims data from 23 million commercially insured patients in the US between 2007 and 2018. Difference-in-differences analysis using panel matching was conducted to evaluate the prevalence of indicators of prescription opioid abuse, opioid use disorder and overdose diagnosis, the prescription of MAT, and drug overdose deaths before and after implementation of 6 state-level policies targeting the opioid epidemic. A random-effects meta-analysis model was used to summarize associations over time for each policy and outcome pair. The data analysis was conducted July 12, 2020. Exposures: State-level drug policy changes to address the increase of opioid-related overdose deaths included prescription drug monitoring program (PDMP) access, mandatory PDMPs, pain clinic laws, prescription limit laws, naloxone access laws, and Good Samaritan laws. Main Outcomes and Measures: The outcomes of interests were quarterly state-level mortality from drug overdoses, known indicators for prescription opioid abuse and doctor shopping, MAT, and prevalence of drug overdose and opioid use disorder. Results: This cross-sectional study of drug overdose mortality data and insurance claims data from 23 million commercially insured patients (12 582 378 female patients [55.1%]; mean [SD] age, 45.9 [19.9] years) in the US between 2007 and 2018 found that mandatory PDMPs were associated with decreases in the proportion of patients taking opioids (-0.729%; 95% CI, -1.011% to -0.447%), with overlapping opioid claims (-0.027%; 95% CI, -0.038% to -0.017%), with daily morphine milligram equivalent greater than 90 (-0.095%; 95% CI, -0.150% to -0.041%), and who engaged in drug seeking (-0.002%; 95% CI, -0.003% to -0.001%). The proportion of patients receiving MAT increased after the enactment of mandatory PDMPs (0.015%; 95% CI, 0.002% to 0.028%), pain clinic laws (0.013%, 95% CI, 0.005%-0.021%), and prescription limit laws (0.034%, 95% CI, 0.020% to 0.049%). Mandatory PDMPs were associated with a decrease in the number of overdose deaths due to natural opioids (-518.5 [95% CI, -728.5 to -308.5] per 300 million people) and methadone (-122.7 [95% CI, -207.5 to -37.8] per 300 million people). Prescription drug monitoring program access policies showed similar results, although these policies were also associated with increases in overdose deaths due to synthetic opioids (380.3 [95% CI, 149.6-610.8] per 300 million people) and cocaine (103.7 [95% CI, 28.0-179.5] per 300 million people). Except for the negative association between prescription limit laws and synthetic opioid deaths (-723.9 [95% CI, -1419.7 to -28.1] per 300 million people), other policies were associated with increasing overdose deaths, especially those attributed to non-prescription opioids such as synthetic opioids and heroin. This includes a positive association between naloxone access laws and the number of deaths attributed to synthetic opioids (1338.2 [95% CI, 662.5 to 2014.0] per 300 million people). Conclusions and Relevance: Although this study found that existing state policies were associated with reduced misuse of prescription opioids, they may have the unintended consequence of motivating those with opioid use disorders to access the illicit drug market, potentially increasing overdose mortality. This finding suggests that there is no easy policy solution to reverse the epidemic of opioid dependence and mortality in the US.

Prevalence of Misinformation and Factchecks on the COVID-19 Pandemic in 35 Countries: Observational Infodemiology Study.

BACKGROUND: The COVID-19 pandemic has been accompanied by an infodemic, in which a plethora of false information has been rapidly disseminated online, leading to serious harm worldwide. OBJECTIVE: This study aims to analyze the prevalence of common misinformation related to the COVID-19 pandemic. METHODS: We conducted an online survey via social media platforms and a survey company to determine whether respondents have been exposed to a broad set of false claims and fact-checked information on the disease. RESULTS: We obtained more than 41,000 responses from 1257 participants in 85 countries, but for our analysis, we only included responses from 35 countries that had at least 15 respondents. We identified a strong negative correlation between a country's Gross Domestic Product per-capita and the prevalence of misinformation, with poorer countries having a higher prevalence of misinformation (Spearman ρ=-0.72; P<.001). We also found that fact checks spread to a lesser degree than their respective false claims, following a sublinear trend (ß=.64). CONCLUSIONS: Our results imply that the potential harm of misinformation could be more substantial for low-income countries than high-income countries. Countries with poor infrastructures might have to combat not only the spreading pandemic but also the COVID-19 infodemic, which can derail efforts in saving lives.

Evidence from internet search data shows information-seeking responses to news of local COVID-19 cases.

The COVID-19 outbreak is a global pandemic with community circulation in many countries, including the United States, with confirmed cases in all states. The course of this pandemic will be shaped by how governments enact timely policies and disseminate information and by how the public reacts to policies and information. Here, we examine information-seeking responses to the first COVID-19 case public announcement in a state. Using an event study framework for all US states, we show that such news increases collective attention to the crisis right away. However, the elevated level of attention is short-lived, even though the initial announcements are followed by increasingly strong policy measures. Specifically, searches for "coronavirus" increased by about 36% (95% CI: 27 to 44%) on the day immediately after the first case announcement but decreased back to the baseline level in less than a week or two. We find that people respond to the first report of COVID-19 in their state by immediately seeking information about COVID-19, as measured by searches for coronavirus, coronavirus symptoms, and hand sanitizer. On the other hand, searches for information regarding community-level policies (e.g., quarantine, school closures, testing) or personal health strategies (e.g., masks, grocery delivery, over-the-counter medications) do not appear to be immediately triggered by first reports. These results are representative of the study period being relatively early in the epidemic, and more-elaborate policy responses were not yet part of the public discourse. Further analysis should track evolving patterns of responses to subsequent flows of public information.