Exposing influence campaigns in the age of LLMs: a behavioral-based AI approach to detecting state-sponsored trolls.

The detection of state-sponsored trolls operating in influence campaigns on social media is a critical and unsolved challenge for the research community, which has significant implications beyond the online realm. To address this challenge, we propose a new AI-based solution that identifies troll accounts solely through behavioral cues associated with their sequences of sharing activity, encompassing both their actions and the feedback they receive from others. Our approach does not incorporate any textual content shared and consists of two steps: First, we leverage an LSTM-based classifier to determine whether account sequences belong to a state-sponsored troll or an organic, legitimate user. Second, we employ the classified sequences to calculate a metric named the "Troll Score", quantifying the degree to which an account exhibits troll-like behavior. To assess the effectiveness of our method, we examine its performance in the context of the 2016 Russian interference campaign during the U.S. Presidential election. Our experiments yield compelling results, demonstrating that our approach can identify account sequences with an AUC close to 99% and accurately differentiate between Russian trolls and organic users with an AUC of 91%. Notably, our behavioral-based approach holds a significant advantage in the ever-evolving landscape, where textual and linguistic properties can be easily mimicked by Large Language Models (LLMs): In contrast to existing language-based techniques, it relies on more challenging-to-replicate behavioral cues, ensuring greater resilience in identifying influence campaigns, especially given the potential increase in the usage of LLMs for generating inauthentic content. Finally, we assessed the generalizability of our solution to various entities driving different information operations and found promising results that will guide future research.

How does Twitter account moderation work? Dynamics of account creation and suspension on Twitter during major geopolitical events.

Social media moderation policies are often at the center of public debate, and their implementation and enactment are sometimes surrounded by a veil of mystery. Unsurprisingly, due to limited platform transparency and data access, relatively little research has been devoted to characterizing moderation dynamics, especially in the context of controversial events and the platform activity associated with them. Here, we study the dynamics of account creation and suspension on Twitter during two global political events: Russia's invasion of Ukraine and the 2022 French Presidential election. Leveraging a large-scale dataset of 270M tweets shared by 16M users in multiple languages over several months, we identify peaks of suspicious account creation and suspension, and we characterize behaviors that more frequently lead to account suspension. We show how large numbers of accounts get suspended within days of their creation. Suspended accounts tend to mostly interact with legitimate users, as opposed to other suspicious accounts, making unwarranted and excessive use of reply and mention features, and sharing large amounts of spam and harmful content. While we are only able to speculate about the specific causes leading to a given account suspension, our findings contribute to shedding light on patterns of platform abuse and subsequent moderation during major events.

Exploring the Behavior of Users With Attention-Deficit/Hyperactivity Disorder on Twitter: Comparative Analysis of Tweet Content and User Interactions.

BACKGROUND: With the widespread use of social media, people share their real-time thoughts and feelings via interactions on these platforms, including those revolving around mental health problems. This can provide a new opportunity for researchers to collect health-related data to study and analyze mental disorders. However, as one of the most common mental disorders, there are few studies regarding the manifestations of attention-deficit/hyperactivity disorder (ADHD) on social media. OBJECTIVE: This study aims to examine and identify the different behavioral patterns and interactions of users with ADHD on Twitter through the text content and metadata of their posted tweets. METHODS: First, we built 2 data sets: an ADHD user data set containing 3135 users who explicitly reported having ADHD on Twitter and a control data set made up of 3223 randomly selected Twitter users without ADHD. All historical tweets of users in both data sets were collected. We applied mixed methods in this study. We performed Top2Vec topic modeling to extract topics frequently mentioned by users with ADHD and those without ADHD and used thematic analysis to further compare the differences in contents that were discussed by the 2 groups under these topics. We used a distillBERT sentiment analysis model to calculate the sentiment scores for the emotion categories and compared the sentiment intensity and frequency. Finally, we extracted users' posting time, tweet categories, and the number of followers and followings from the metadata of tweets and compared the statistical distribution of these features between ADHD and non-ADHD groups. RESULTS: In contrast to the control group of the non-ADHD data set, users with ADHD tweeted about the inability to concentrate and manage time, sleep disturbance, and drug abuse. Users with ADHD felt confusion and annoyance more frequently, while they felt less excitement, caring, and curiosity (all P<.001). Users with ADHD were more sensitive to emotions and felt more intense feelings of nervousness, sadness, confusion, anger, and amusement (all P<.001). As for the posting characteristics, compared with controls, users with ADHD were more active in posting tweets (P=.04), especially at night between midnight and 6 AM (P<.001); posting more tweets with original content (P<.001); and following fewer people on Twitter (P<.001). CONCLUSIONS: This study revealed how users with ADHD behave and interact differently on Twitter compared with those without ADHD. On the basis of these differences, researchers, psychiatrists, and clinicians can use Twitter as a potentially powerful platform to monitor and study people with ADHD, provide additional health care support to them, improve the diagnostic criteria of ADHD, and design complementary tools for automatic ADHD detection.

#JusticeforGeorgeFloyd: How Instagram facilitated the 2020 Black Lives Matter protests.

We present and analyze a database of 1.13 million public Instagram posts during the Black Lives Matter protests of 2020, which erupted in response to George Floyd's public murder by police on May 25. Our aim is to understand the growing role of visual media, focusing on a) the emergent opinion leaders and b) the subsequent press concerns regarding frames of legitimacy. We perform a comprehensive view of the spatial (where) and temporal (when) dynamics, the visual and textual content (what), and the user communities (who) that drove the social movement on Instagram. Results reveal the emergence of non-institutional opinion leaders such as meme groups, independent journalists, and fashion magazines, which contrasts with the institutionally reinforcing nature of Twitter. Visual analysis of 1.69 million photos show symbols of injustice are the most viral coverage, and moreover, actual protest coverage is framed positively, in contrast with combatant frames traditionally found from legacy media. Together, these factors helped facilitate the online movement through three phases, culminating with online international solidarity in #BlackOutTuesday. Through this case study, we demonstrate the precarious nature of protest journalism, and how content creators, journalists, and everyday users co-evolved with social media to shape one of America's largest-ever human rights movements.

TILES-2019: A longitudinal physiologic and behavioral data set of medical residents in an intensive care unit.

The TILES-2019 data set consists of behavioral and physiological data gathered from 57 medical residents (i.e., trainees) working in an intensive care unit (ICU) in the United States. The data set allows for the exploration of longitudinal changes in well-being, teamwork, and job performance in a demanding environment, as residents worked in the ICU for three weeks. Residents wore a Fitbit, a Bluetooth-based proximity sensor, and an audio-feature recorder. They completed daily surveys and interviews at the beginning and end of their rotation. In addition, we collected data from environmental sensors (i.e., Internet-of-Things Bluetooth data hubs) and obtained hospital records (e.g., patient census) and residents' job evaluations. This data set may be may be of interest to researchers interested in workplace stress, group dynamics, social support, the physical and psychological effects of witnessing patient deaths, predicting survey data from sensors, and privacy-aware and privacy-preserving machine learning. Notably, a small subset of the data was collected during the first wave of the COVID-19 pandemic.

Botometer 101: social bot practicum for computational social scientists.

Social bots have become an important component of online social media. Deceptive bots, in particular, can manipulate online discussions of important issues ranging from elections to public health, threatening the constructive exchange of information. Their ubiquity makes them an interesting research subject and requires researchers to properly handle them when conducting studies using social media data. Therefore, it is important for researchers to gain access to bot detection tools that are reliable and easy to use. This paper aims to provide an introductory tutorial of Botometer, a public tool for bot detection on Twitter, for readers who are new to this topic and may not be familiar with programming and machine learning. We introduce how Botometer works, the different ways users can access it, and present a case study as a demonstration. Readers can use the case study code as a template for their own research. We also discuss recommended practice for using Botometer.

Comparative analysis of social bots and humans during the COVID-19 pandemic.

Using more than 4 billion tweets and labels on more than 5 million users, this paper compares the behavior of humans and bots politically and semantically during the pandemic. Results reveal liberal bots are more central than humans in general, but less important than institutional humans as the elite circle grows smaller. Conservative bots are surprisingly absent when compared to prior work on political discourse, but are better than liberal bots at eliciting replies from humans, which suggest they may be perceived as human more frequently. In terms of topic and framing, conservative humans and bots disproportionately tweet about the Bill Gates and bio-weapons conspiracy, whereas the 5G conspiracy is bipartisan. Conservative humans selectively ignore mask-wearing and we observe prevalent out-group tweeting when discussing policy. We discuss and contrast how humans appear more centralized in health-related discourse as compared to political events, which suggests the importance of credibility and authenticity for public health in online information diffusion.

Charting the Information and Misinformation Landscape to Characterize Misinfodemics on Social Media: COVID-19 Infodemiology Study at a Planetary Scale.

BACKGROUND: The novel coronavirus, also known as SARS-CoV-2, has come to define much of our lives since the beginning of 2020. During this time, countries around the world imposed lockdowns and social distancing measures. The physical movements of people ground to a halt, while their online interactions increased as they turned to engaging with each other virtually. As the means of communication shifted online, information consumption also shifted online. Governing authorities and health agencies have intentionally shifted their focus to use social media and online platforms to spread factual and timely information. However, this has also opened the gate for misinformation, contributing to and accelerating the phenomenon of misinfodemics. OBJECTIVE: We carried out an analysis of Twitter discourse on over 1 billion tweets related to COVID-19 over a year to identify and investigate prevalent misinformation narratives and trends. We also aimed to describe the Twitter audience that is more susceptible to health-related misinformation and the network mechanisms driving misinfodemics. METHODS: We leveraged a data set that we collected and made public, which contained over 1 billion tweets related to COVID-19 between January 2020 and April 2021. We created a subset of this larger data set by isolating tweets that included URLs with domains that had been identified by Media Bias/Fact Check as being prone to questionable and misinformation content. By leveraging clustering and topic modeling techniques, we identified major narratives, including health misinformation and conspiracies, which were present within this subset of tweets. RESULTS: Our focus was on a subset of 12,689,165 tweets that we determined were representative of COVID-19 misinformation narratives in our full data set. When analyzing tweets that shared content from domains known to be questionable or that promoted misinformation, we found that a few key misinformation narratives emerged about hydroxychloroquine and alternative medicines, US officials and governing agencies, and COVID-19 prevention measures. We further analyzed the misinformation retweet network and found that users who shared both questionable and conspiracy-related content were clustered more closely in the network than others, supporting the hypothesis that echo chambers can contribute to the spread of health misinfodemics. CONCLUSIONS: We presented a summary and analysis of the major misinformation discourse surrounding COVID-19 and those who promoted and engaged with it. While misinformation is not limited to social media platforms, we hope that our insights, particularly pertaining to health-related emergencies, will help pave the way for computational infodemiology to inform health surveillance and interventions.

#Election2020: the first public Twitter dataset on the 2020 US Presidential election.

Credible evidence-based political discourse is a critical pillar of democracy and is at the core of guaranteeing free and fair elections. The study of online chatter is paramount, especially in the wake of important voting events like the recent November 3, 2020 U.S. Presidential election and the inauguration on January 21, 2021. Limited access to social media data is often the primary obstacle that limits our abilities to study and understand online political discourse. To mitigate this impediment and empower the Computational Social Science research community, we are publicly releasing a massive-scale, longitudinal dataset of U.S. politics- and election-related tweets. This multilingual dataset encompasses over 1.2 billion tweets and tracks all salient U.S. political trends, actors, and events from 2019 to the time of this writing. It predates and spans the entire period of the Republican and Democratic primaries, with real-time tracking of all presidential contenders on both sides of the aisle. The dataset also focuses on presidential and vice-presidential candidates, the presidential elections and the transition from the Trump administration to the Biden administration. Our dataset release is curated, documented, and will continue to track relevant events. We hope that the academic community, computational journalists, and research practitioners alike will all take advantage of our dataset to study relevant scientific and social issues, including problems like misinformation, information manipulation, conspiracies, and the distortion of online political discourse that has been prevalent in the context of recent election events in the United States. Our dataset is available at: https://github.com/echen102/us-pres-elections-2020.

COVID-19 Vaccine Hesitancy on Social Media: Building a Public Twitter Data Set of Antivaccine Content, Vaccine Misinformation, and Conspiracies.

BACKGROUND: False claims about COVID-19 vaccines can undermine public trust in ongoing vaccination campaigns, posing a threat to global public health. Misinformation originating from various sources has been spreading on the web since the beginning of the COVID-19 pandemic. Antivaccine activists have also begun to use platforms such as Twitter to promote their views. To properly understand the phenomenon of vaccine hesitancy through the lens of social media, it is of great importance to gather the relevant data. OBJECTIVE: In this paper, we describe a data set of Twitter posts and Twitter accounts that publicly exhibit a strong antivaccine stance. The data set is made available to the research community via our AvaxTweets data set GitHub repository. We characterize the collected accounts in terms of prominent hashtags, shared news sources, and most likely political leaning. METHODS: We started the ongoing data collection on October 18, 2020, leveraging the Twitter streaming application programming interface (API) to follow a set of specific antivaccine-related keywords. Then, we collected the historical tweets of the set of accounts that engaged in spreading antivaccination narratives between October 2020 and December 2020, leveraging the Academic Track Twitter API. The political leaning of the accounts was estimated by measuring the political bias of the media outlets they shared. RESULTS: We gathered two curated Twitter data collections and made them publicly available: (1) a streaming keyword-centered data collection with more than 1.8 million tweets, and (2) a historical account-level data collection with more than 135 million tweets. The accounts engaged in the antivaccination narratives lean to the right (conservative) direction of the political spectrum. The vaccine hesitancy is fueled by misinformation originating from websites with already questionable credibility. CONCLUSIONS: The vaccine-related misinformation on social media may exacerbate the levels of vaccine hesitancy, hampering progress toward vaccine-induced herd immunity, and could potentially increase the number of infections related to new COVID-19 variants. For these reasons, understanding vaccine hesitancy through the lens of social media is of paramount importance. Because data access is the first obstacle to attain this goal, we published a data set that can be used in studying antivaccine misinformation on social media and enable a better understanding of vaccine hesitancy.

The impact of peer review on the contribution potential of scientific papers.

The peer-reviewing process has long been regarded as an indispensable tool in ensuring the quality of a scientific publication. While previous studies have tried to understand the process as a whole, not much effort has been devoted to investigating the determinants and impacts of the content of the peer review itself. This study leverages open data from nearly 5,000 PeerJ publications that were eventually accepted. Using sentiment analysis, Latent Dirichlet Allocation (LDA) topic modeling, mixed linear regression models, and logit regression models, we examine how the peer-reviewing process influences the acceptance timeline and contribution potential of manuscripts, and what modifications were typically made to manuscripts prior to publication. In an open review paradigm, our findings indicate that peer reviewers' choice to reveal their names in lieu of remaining anonymous may be associated with more positive sentiment in their review, implying possible social pressure from name association. We also conduct a taxonomy of the manuscript modifications during a revision, studying the words added in response to peer reviewer feedback. This study provides insights into the content of peer reviews and the subsequent modifications authors make to their manuscripts.

Social Media Polarization and Echo Chambers in the Context of COVID-19: Case Study.

BACKGROUND: Social media chatter in 2020 has been largely dominated by the COVID-19 pandemic. Existing research shows that COVID-19 discourse is highly politicized, with political preferences linked to beliefs and disbeliefs about the virus. As it happens with topics that become politicized, people may fall into echo chambers, which is the idea that one is only presented with information they already agree with, thereby reinforcing one's confirmation bias. Understanding the relationship between information dissemination and political preference is crucial for effective public health communication. OBJECTIVE: We aimed to study the extent of polarization and examine the structure of echo chambers related to COVID-19 discourse on Twitter in the United States. METHODS: First, we presented Retweet-BERT, a scalable and highly accurate model for estimating user polarity by leveraging language features and network structures. Then, by analyzing the user polarity predicted by Retweet-BERT, we provided new insights into the characterization of partisan users. RESULTS: We observed that right-leaning users were noticeably more vocal and active in the production and consumption of COVID-19 information. We also found that most of the highly influential users were partisan, which may contribute to further polarization. Importantly, while echo chambers exist in both the right- and left-leaning communities, the right-leaning community was by far more densely connected within their echo chamber and isolated from the rest. CONCLUSIONS: We provided empirical evidence that political echo chambers are prevalent, especially in the right-leaning community, which can exacerbate the exposure to information in line with pre-existing users' views. Our findings have broader implications in developing effective public health campaigns and promoting the circulation of factual information online.

Topics of Nicotine-Related Discussions on Twitter: Infoveillance Study.

BACKGROUND: Cultural trends in the United States, the nicotine consumer marketplace, and tobacco policies are changing. OBJECTIVE: The goal of this study was to identify and describe nicotine-related topics of conversation authored by the public and social bots on Twitter, including any misinformation or misconceptions that health education campaigns could potentially correct. METHODS: Twitter posts containing the term "nicotine" were obtained from September 30, 2018 to October 1, 2019. Methods were used to distinguish between posts from social bots and nonbots. Text classifiers were used to identify topics in posts (n=300,360). RESULTS: Prevalent topics of posts included vaping, smoking, addiction, withdrawal, nicotine health risks, and quit nicotine, with mentions of going "cold turkey" and needing help in quitting. Cessation was a common topic, with mentions of quitting and stopping smoking. Social bots discussed unsubstantiated health claims including how hypnotherapy, acupuncture, magnets worn on the ears, and time spent in the sauna can help in smoking cessation. CONCLUSIONS: Health education efforts are needed to correct unsubstantiated health claims on Twitter and ultimately direct individuals who want to quit smoking to evidence-based cessation strategies. Future interventions could be designed to follow these topics of discussions on Twitter and engage with members of the public about evidence-based cessation methods in near real time when people are contemplating cessation.

Political Partisanship and Antiscience Attitudes in Online Discussions About COVID-19: Twitter Content Analysis.

BACKGROUND: The novel coronavirus pandemic continues to ravage communities across the United States. Opinion surveys identified the importance of political ideology in shaping perceptions of the pandemic and compliance with preventive measures. OBJECTIVE: The aim of this study was to measure political partisanship and antiscience attitudes in the discussions about the pandemic on social media, as well as their geographic and temporal distributions. METHODS: We analyzed a large set of tweets from Twitter related to the pandemic, collected between January and May 2020, and developed methods to classify the ideological alignment of users along the moderacy (hardline vs moderate), political (liberal vs conservative), and science (antiscience vs proscience) dimensions. RESULTS: We found a significant correlation in polarized views along the science and political dimensions. Moreover, politically moderate users were more aligned with proscience views, while hardline users were more aligned with antiscience views. Contrary to expectations, we did not find that polarization grew over time; instead, we saw increasing activity by moderate proscience users. We also show that antiscience conservatives in the United States tended to tweet from the southern and northwestern states, while antiscience moderates tended to tweet from the western states. The proportion of antiscience conservatives was found to correlate with COVID-19 cases. CONCLUSIONS: Our findings shed light on the multidimensional nature of polarization and the feasibility of tracking polarized opinions about the pandemic across time and space through social media data.

Gender Disparity in the Authorship of Biomedical Research Publications During the COVID-19 Pandemic: Retrospective Observational Study.

BACKGROUND: Gender imbalances in academia have been evident historically and persist today. For the past 60 years, we have witnessed the increase of participation of women in biomedical disciplines, showing that the gender gap is shrinking. However, preliminary evidence suggests that women, including female researchers, are disproportionately affected by the COVID-19 pandemic in terms of unequal distribution of childcare, elderly care, and other kinds of domestic and emotional labor. Sudden lockdowns and abrupt shifts in daily routines have had disproportionate consequences on their productivity, which is reflected by a sudden drop in research output in biomedical research, consequently affecting the number of female authors of scientific publications. OBJECTIVE: The objective of this study is to test the hypothesis that the COVID-19 pandemic has had a disproportionate adverse effect on the productivity of female researchers in the biomedical field in terms of authorship of scientific publications. METHODS: This is a retrospective observational bibliometric study. We investigated the proportion of male and female researchers who published scientific papers during the COVID-19 pandemic, using bibliometric data from biomedical preprint servers and selected Springer-Nature journals. We used the ordinary least squares regression model to estimate the expected proportions over time by correcting for temporal trends. We also used a set of statistical methods, such as the Kolmogorov-Smirnov test and regression discontinuity design, to test the validity of the results. RESULTS: A total of 78,950 papers from the bioRxiv and medRxiv repositories and from 62 selected Springer-Nature journals by 346,354 unique authors were analyzed. The acquired data set consisted of papers that were published between January 1, 2019, and August 2, 2020. The proportion of female first authors publishing in the biomedical field during the pandemic dropped by 9.1%, on average, across disciplines (expected arithmetic mean yest=0.39; observed arithmetic mean y=0.35; standard error of the estimate, Sest=0.007; standard error of the observation, σx=0.004). The impact was particularly pronounced for papers related to COVID-19 research, where the proportion of female scientists in the first author position dropped by 28% (yest=0.39; y=0.28; Sest=0.007; σx=0.007). When looking at the last authors, the proportion of women dropped by 7.9%, on average (yest=0.25; y=0.23; Sest=0.005; σx=0.003), while the proportion of women writing about COVID-19 as the last author decreased by 18.8% (yest=0.25; y=0.21; Sest=0.005; σx=0.007). Further, by geocoding authors' affiliations, we showed that the gender disparities became even more apparent when disaggregated by country, up to 35% in some cases. CONCLUSIONS: Our findings document a decrease in the number of publications by female authors in the biomedical field during the global pandemic. This effect was particularly pronounced for papers related to COVID-19, indicating that women are producing fewer publications related to COVID-19 research. This sudden increase in the gender gap was persistent across the 10 countries with the highest number of researchers. These results should be used to inform the scientific community of this worrying trend in COVID-19 research and the disproportionate effect that the pandemic has had on female academics.

Disrupting the COVID-19 Misinfodemic With Network Interventions: Network Solutions for Network Problems.

Amid the COVID-19 global pandemic, a highly troublesome influx of viral misinformation threatens to exacerbate the crisis through its deleterious effects on public health outcomes and health behavior decisions.This "misinfodemic" has ignited a surge of ongoing research aimed at characterizing its content, identifying its sources, and documenting its effects. Noticeably absent as of yet is a cogent strategy to disrupt misinformation.We start with the premise that the diffusion and persistence of COVID-19 misinformation are networked phenomena that require network interventions. To this end, we propose five classes of social network intervention to provide a roadmap of opportunities for disrupting misinformation dynamics during a global health crisis. Collectively, these strategies identify five distinct yet interdependent features of information environments that present viable opportunities for interventions.

Network modularity controls the speed of information diffusion.

The rapid diffusion of information and the adoption of social behaviors are of critical importance in situations as diverse as collective actions, pandemic prevention, or advertising and marketing. Although the dynamics of large cascades have been extensively studied in various contexts, few have systematically examined the impact of network topology on the efficiency of information diffusion. Here, by employing the linear threshold model on networks with communities, we demonstrate that a prominent network feature-the modular structure-strongly affects the speed of information diffusion in complex contagion. Our simulations show that there always exists an optimal network modularity for the most efficient spreading process. Beyond this critical value, either a stronger or a weaker modular structure actually hinders the diffusion speed. These results are confirmed by an analytical approximation. We further demonstrate that the optimal modularity varies with both the seed size and the target cascade size and is ultimately dependent on the network under investigation. We underscore the importance of our findings in applications from marketing to epidemiology, from neuroscience to engineering, where the understanding of the structural design of complex systems focuses on the efficiency of information propagation.

Misinformation, manipulation, and abuse on social media in the era of COVID-19.

The COVID-19 pandemic represented an unprecedented setting for the spread of online misinformation, manipulation, and abuse, with the potential to cause dramatic real-world consequences. The aim of this special issue was to collect contributions investigating issues such as the emergence of infodemics, misinformation, conspiracy theories, automation, and online harassment on the onset of the coronavirus outbreak. Articles in this collection adopt a diverse range of methods and techniques, and focus on the study of the narratives that fueled conspiracy theories, on the diffusion patterns of COVID-19 misinformation, on the global news sentiment, on hate speech and social bot interference, and on multimodal Chinese propaganda. The diversity of the methodological and scientific approaches undertaken in the aforementioned articles demonstrates the interdisciplinarity of these issues. In turn, these crucial endeavors might anticipate a growing trend of studies where diverse theories, models, and techniques will be combined to tackle the different aspects of online misinformation, manipulation, and abuse.

Predictability limit of partially observed systems.

Applications from finance to epidemiology and cyber-security require accurate forecasts of dynamic phenomena, which are often only partially observed. We demonstrate that a system's predictability degrades as a function of temporal sampling, regardless of the adopted forecasting model. We quantify the loss of predictability due to sampling, and show that it cannot be recovered by using external signals. We validate the generality of our theoretical findings in real-world partially observed systems representing infectious disease outbreaks, online discussions, and software development projects. On a variety of prediction tasks-forecasting new infections, the popularity of topics in online discussions, or interest in cryptocurrency projects-predictability irrecoverably decays as a function of sampling, unveiling predictability limits in partially observed systems.

TILES-2018, a longitudinal physiologic and behavioral data set of hospital workers.

We present a novel longitudinal multimodal corpus of physiological and behavioral data collected from direct clinical providers in a hospital workplace. We designed the study to investigate the use of off-the-shelf wearable and environmental sensors to understand individual-specific constructs such as job performance, interpersonal interaction, and well-being of hospital workers over time in their natural day-to-day job settings. We collected behavioral and physiological data from n = 212 participants through Internet-of-Things Bluetooth data hubs, wearable sensors (including a wristband, a biometrics-tracking garment, a smartphone, and an audio-feature recorder), together with a battery of surveys to assess personality traits, behavioral states, job performance, and well-being over time. Besides the default use of the data set, we envision several novel research opportunities and potential applications, including multi-modal and multi-task behavioral modeling, authentication through biometrics, and privacy-aware and privacy-preserving machine learning.