ABSTRACT. This work is focused on studying patterns of human behaviour in museums. Using machine learning and computer vision methods, we are studying human behaviour in the State Hermitage, Saint Petersburg, Russia, based on Instagram data. Analysis of Instagram data revealed the heterogeneous nature of public attention to different museum halls.

ABSTRACT. In this study, we quantify the scholars’ aggregated flow among interacting subfields in Phys and compare the predicted results between the gravity model and the radiation model. we found scholars generally prefer short-range migrations rather than long-range migration in the knowledge space, reflecting the insurmountable knowledge barriers between different subfields.

ABSTRACT. We focus on the public diplomacy efforts made by China and the United States on Twitter under the COVID-19 pandemic. We investigate the divide between the US world and China world by “echo chamber” environment. Such divisions can be observed most prominently in political events.

ABSTRACT. Across both residential and fully online degree programs, including students' protected attributes (gender, ethnicity, first-gen student, financial need) in college dropout prediction models does not affect the overall performance and only marginally improves the fairness of predictions across different protected groups.

ABSTRACT. We contrast two methods for identifying non-users of online news. One compares participants’ tracked Internet consumption with a list of news sources to identify news at the domain level. The second uses collected text data of each visited website and automated text analysis to identify news at the article level.

ABSTRACT. We present Quotebank, a public repository of 178 million speaker-attributed quotations that we extracted from over a decade of news articles, alongside Quobert, the extraction framework that we used to generate this dataset.

ABSTRACT. What makes a tweet sexist? Sexism is a multifaceted construct that can be expressed in subtle as well as overt ways. These characteristics pose challenges to the validity and reliability of computational models aiming to detect sexism in social media text. In this paper we explore solutions to these challenges.

ABSTRACT. Many attempts have been made to gauge the efficacy of digital traces, especially social media, in measuring public opinion, with mixed results. As a new data source, social media present several challenges. We evaluate different methods that have been used to measure public opinion from social media.

ABSTRACT. We propose a point process model describing the spread of fake news on Twitter, which is superior to other methods in predicting the spread of fake news, and its ability to extract a compact representation of the spreading pattern could be useful in the detection and mitigation of fake news.

ABSTRACT. We audit Twitter's shadow ban algorithms for ideological bias, by testing the profile and content visibility for a representative sample of US users. Controlling for incivility, bot-like behavior, and network size, there is little to suggest that right wing accounts are indeed more likely to be shadow banned.

ABSTRACT. To explore the agenda interaction between NGOs and media (two main actors in transnational advocacy networks) on Twitter, this study used Network Agenda Setting theory to analyze 88584 climate-related tweets posted by 22 countries' NGOs and media around COP25 and found an agenda setting relationship from NGOs to media.

ABSTRACT. We propose an agent-based model that explores the effects of small instances of gender bias and discrimination on gender inequality in corporate settings.

ABSTRACT. Utilizing social network analysis this paper examines how gender dictates interaction on the programming forum Stack Overflow. I found that the relationship between answer effort and the score users receive is mediated by gender, with feminine users receiving lower scores. These findings hold across a non-binary categorization of gender.

ABSTRACT. Structural factors cannot completely explain the gender pay gap making it important to understand the potential role of gender biases. We use data from an online platform that facilitates finding private tutors, and by analyzing texts of reviews are able to get insight into the potential role of gender stereotypes.

ABSTRACT. We use automatically collected Strava data on outdoor cycling for 70 cities across the USA, the UK, Italy and Benelux, to investigate the determinants of gender gap in cycling. The results support the hypothesis of gender-specific preferences, with women displaying a stronger preference for dedicated cycling infrastructure.

ABSTRACT. We study the structure of the French media ecosystem. We both apply SBM analysis to the hyperlink network between information sources and measure the ideological distribution of news stories using Twitter data. Both analyses suggest that the French media landscape is not as polarized as in the US.

ABSTRACT. We use network science and computational linguistics to map the psychometric landscape of climate activists and deniers on Twitter between 2015 and 2020. We focus on how these two camps use agency in relation to other psychometric features, and how these trends correlate between the activists an deniers over time.

ABSTRACT. Using data from Reddit and applying statistical methods and NLP, we find that affective political polarization on Reddit decreased over the last decade, which contradicts observations from survey-based studies about affective polarization and social media as a driver of polarization.

ABSTRACT. We perform a cross-platform analysis in which we study how does linking YouTube content on Reddit conspiracy forum impact language used in user comments on YouTube. Our findings show a slight change in user language in that it becomes more similar to language used on Reddit.

ABSTRACT. We combine an experiment on biased argument processing with argument-based models of opinion formation. Calibration shows that the incorporation of biased processing increases the micro validity of the models and has a strong impact on their macro-level predictions. Weak biases increase group efficiency, stronger biases lead to persistent polarization.

ABSTRACT. Most Network analyses implicitly assume that considered data is error-free and reliable. We model how systematic uncertainty on edges of an attributed network can impact network analysis. Thereby we systematically discuss how erroneous edge observations can be driven by external node attributes and the relative edge positions in the network.

ABSTRACT. We provide evidence that anti-intervention protests in the US are temporally associated with COVID-19 interventions and political geography, while the scale of protests is mainly related to population size. The risk of protest-driven epidemics is estimated by an integrated model of social mobilization and SEIR epidemic model.

ABSTRACT. We find that distribution of neighborhood sizes in cities follows exponential decay and explain this using a computational model of neighborhood dynamics, where agent movement is mediated by wealth. The use of a comparative wealth-based metric by agents combined with affordability thresholds appear necessary conditions for emergence of exponential decay.

ABSTRACT. Previous works have identified across-platform differences in campaigning strategies of political actors as a result of the effect of platform affordances and the discrepancy of user demographics. This paper investigates the effect of the selection of social media platforms on the result of political communication research.

ABSTRACT. We model online misinformation waves that are accompanying COVID-19 pandemic, as observed on Twitter across 40 countries worldwide. We test three models and find that the most suitable is an evolutionary model driven by imitation and perceived risk, which is in turn depending on local and global epidemic levels.

ABSTRACT. We investigated emotional expressions in tweets during five weeks after the COVID-19 outbreak in 18 countries. We observed early strong upsurges of anxiety-related terms in all countries, followed by increases in sadness and decreases in anger in most countries, including some of the most enduring emotional changes observed so far.

ABSTRACT. The paper shows how a machine learning model can be trained to identify reports on government policies against COVID-19 in international media. It illustrates how transfer learning techniques can be applied to a dataset from political scientists to train transformer models for a real-world use case in a different domain.

ABSTRACT. In this work, citizen profiles of a candidate city for the European Green Capital award are investigated using an integrated Knowledge Discovery in Databases process in order to understand whether the citizens are psychologically and behaviorally green.

ABSTRACT. Using a large multilingual corpus, this paper applies a novel approach for a comprehensive, systematic, large-scale, comparative examination of the media coverage of Eastern European migrants in the French, German, and UK press, studying their relative salience and relative framing: across countries, and, notably, compared to other, non-European migrants.

ABSTRACT. We synthesize political communication with NLP to computational analyze framing on Twitter, with a particular focus on immigration. We create a novel dataset of immigration-related tweets and use supervised machine learning to detect frames. We then investigate the relationship between framing and political ideology as well as audience engagement.

ABSTRACT. Applying neural community embeddings, we accurately quantify U.S. partisan affiliation and ideological extremism across communities on Reddit and track its change over time, finding that political polarization increased dramatically in 2016, and that this change was disproportionately driven by new users joining the platform, not change in existing users.

ABSTRACT. In this work, we develop a framework for studying human-AI interaction by learning highly accurate personalized models of human decision-making in chess, and show that they capture signatures of individual style.

ABSTRACT. We address the concern that training datasets for supervised learning could have been poisoned through discrimination by i) defining and modeling discrimination as a dataset poisoning, ii) proposing novel interventional mixtures to inhibit discrimination, iii) and evaluating these and other methods addressing discrimination on synthetic and real-world datasets.

ABSTRACT. This study proposed an automated classification method to disaggregate Twitter users into five categories: news media, politicians/government, environmental advocacy groups, individual users, and other organizations. We built a reliable training dataset that reduces the workload of human labeling. Initial results using one event showed our method produced reliable results.

ABSTRACT. We explore polarization on Twitter in a different context, namely the protest that paralyzed several countries in the South American region in 2019. We explore polarization in two related dimensions: language and news consumption patterns. We show communities speak consistently “different” languages and find evidence of “filter bubbles”.

ABSTRACT. We investigate the relationship between individual differences in cognitive reflection and behavior on the social media platform Twitter N= 1,901 individuals We find that people who score higher on the Cognitive Reflection Test—a widely used measure of reflective thinking—were more discerning in their social media use.

ABSTRACT. We develop a fitness-based model for the spread of information on Twitter and fit it to Covid-related tweet cascades. We use the cascade size distribution to infer the distribution of fitnesses of tweets and find that a Generalized Gamma explains the data well and infer its parameters.

ABSTRACT. From a global survey of 65,031 readers of Wikipedia and their reading logs, we present evidence of gender differences: women are underrepresented among readers, view fewer pages per reading session, visit Wikipedia for similar reasons to men, and exhibit specific topical preferences. Our findings lay the foundation toward knowledge equity.

ABSTRACT. We advance the computational social science of gender by developing a Bayesian framework for name-based gender inference that is more statistically sound, conceptually rigorous, and ethically defensible than existing methods.

ABSTRACT. We study the structure of the gender salary gap in higher education based on computational techniques applied to the data linking Web of Science and University of California faculty salary disclosure. We show that the differential return on faculty’s academic performance is the critical component of persisting gender pay gap.

ABSTRACT. Using naming practice as a proxy to investigate gender equality in Javanese society, we found that Javanese males had significantly longer personal names than Javanese females, but the differences became no longer significant at different years, illuminating the progress that spread from urban, to suburb and rural Java areas.

ABSTRACT. We present computationally feasible methods for partitioning signed networks into internally cohesive and mutually divisive groups. These methods lead to unique insights as demonstrated using US Congress networks, which reveals that the US House is structured by three coalitions, in contrast to the traditional two-coalition or two-party views. Preprint: arxiv.org/pdf/2105.01913

ABSTRACT. We find that estimating homophily is a problem of predicting dyad categories, rather than node categories. Node-level prediction models are biased. We explore sources of bias, and evaluate the efficacy of various modeling strategies for estimating homophily.

ABSTRACT. Our broad aim is to improve bias-adjustments of Facebook advertising data for nowcasting migrant stocks. We examine the systematic bias and variability of Facebook advertising data over two years and identify key challenges to overcome when modeling biases.

ABSTRACT. We ran an online survey to gather daily emotion self-reports and compare them with aggregated results of sentiment analysis of user discussions on the same newspaper platform. Using a combination of supervised and unsupervised methods we show that macro level dynamics of emotions can be tracked with social media text.

ABSTRACT. We introduce (1) a “recommender” model describing a generic item-matching process under a personalized recommender system and (2) an “organic” model describing a counterfactual baseline where users find items on their own. We show through theorems and simulations with real data that the models result in fundamentally different user outcomes.

ABSTRACT. The leading eight social norms prove sustain a high level of cooperation also with local information and evolution. Local evolution is better for cooperation than global evolution, while local reputation does not hinder cooperation compared to global reputation. Stricter norms against defection favour cooperation under local evolution/reputation.

ABSTRACT. We model the cost of corporate tax avoidance to countries using a dynamical model. We show that developing countries suffer losses 10-100 times larger than developed countries. This is due to higher borrowing costs, and lower tax revenue gains from taxing firms' shareholders

ABSTRACT. Many democratic societies have become more politically polarized, with the U.S. being the main example. To provide insight into some of the mechanisms underlying political polarization, we develop a mathematical framework and employ Bayesian MCMC concepts to analyze empirical data on political polarization in the U.S. from 1994 to 2017.

ABSTRACT. Do collaboration networks change under the pandemic? Findings show that the collaboration networks become smaller with denser connections between scientists, and geographic location and personal connection remain significant factors. Most biomedical knowledge is stable, but central and novel concepts expand semantic meanings through scientific collaboration and social learning.

ABSTRACT. The U.S. experienced strong partisan divide regarding the response to the COVID-19 pandemic. We characterize and compare the partisan narratives of the Democratic and Republican politicians in response to the pandemic using novel computational methods, systematically uncovering the contrasting topics, frames, and semantic agents that shape their narratives.

ABSTRACT. The spread of the novel coronavirus is affected by the spread of related infodemic content that makes populations vulnerable through resistance to mitigation efforts. Here we study infodemic spreading on Twitter and Facebook by characterizing cross-platform similarities and differences in popular sources, diffusion patterns, influencers, coordination, and automation.

ABSTRACT. We monitor tens of dark web marketplaces during the covid-19 pandemic, analysing millions of scraped listings, COVID-19 related tweets and wikipedia page views. We reveal how they swiftly reacted to product shortages and public attention by offering COVID-19 related goods like PPE, hydroxychloroquine and vaccines with related temporal price variations.

ABSTRACT. Motivated by the goal of designing interventions for softening polarized opinions on the Web, and building on results from psychology, we hypothesized that people would be moved more easily towards opposing opinions when the latter were voiced by a celebrity they like, rather than by a celebrity they dislike. We tested this hypothesis in a survey-based randomized controlled trial. Unlike hypothesized, no softening of opinions was observed regardless of the respondents’ attitudes towards the celebrity. Instead, we found strong evidence of a hardening of pre-treatment opinions. Our results confirm that naïve strategies at mediation may not yield intended results, and how difficult it is to depolarize—and how easy it is to further polarize or provoke emotional responses. I will also discuss a second study, where we explored the implications of the above findings in the context of the COVID-19 pandemic by measuring which spokespersons are most effective at convincing people to support social distancing measures.

ABSTRACT. The use of country-wide administrative register data enables the discovery of population-scale insights into contemporary problems such as segregation, inequality and poverty. This talk considers responsibly anonymized population-scale social network data on all 17 million inhabitants of the Netherlands. I will discuss how the type of formal links in this social network require one to critically rethink network analysis concepts such as the unit of analysis, measurement errors effects and the boundary specification problem. Moreover, I will show how the analysis of a population-scale multilayer network of family, work, school, household and neighborhood relations enables us to revisit the well-known small-world phenomenon from a unique angle. Finally, I outline the possibilities of population-scale network data for various areas of social science research.

ABSTRACT. Critical Race Theory examines the role of race in the law, questioning how law constructs race and reinforces discrimination. Just as Critical Race Theory introduced an approach to deconstruct race in legal contexts, a new theory is now required to deconstruct the impact of digital data and AI in legal contexts. Critical Data Theory examines the role of big data and algorithmic decisionmaking at its intersection with the law. The tools of big data, AI, and data science allow for legal, scientific, socioeconomic, and political frameworks of power that parallel the manner in which tools of race negotiation and definition have shaped these constructions. It is argued here that the constitutionality of big data governance generally and cybersurveillance specifically can be better conceptualized through Critical Data Theory. While the broader umbrella of Critical Theory features prominently in the work of surveillance scholars in the social sciences, legal scholars have not significantly deployed these theoretical tools to address recent surveillance and privacy law challenges. The article on which this keynote is based proposes that Critical Data Theory can help assess the computational and data science impact of technological developments on core constitutional rights and principles.

ABSTRACT. Using a 75-year newspaper archive from Sweden, we map the evolution of a shared understanding of immigration to demarcate eras of meaning based not on their significance to today's beholder but based on the mental associations people of the day made.

ABSTRACT. This study examines how the scientific community has evolved its semantic structure about the revolutionary gene-editing technology CRISPR-Cas9 by analyzing over 23,000 CRISPR-related scientific papers from 2012 through 2020 with a text network approach.

ABSTRACT. We address the problem of modeling bursty diffusion of text-based events over a social-network of users. We propose Dual-Network-Hawkes-Process(DNHP) that disentangles, and analyzes overlapping social-conversations. Using experiments over large collection of tweets by US-politicians, we show that DNHP generalizes better and also provides interesting insights about user and topic transitions.

ABSTRACT. We examine the discourses of the so-called New Right in Germany using algorithm-based topic modelling. Our data basis are 61,211 articles of the Junge Freiheit published between 1997 and 2020. Results show, amongst others, that with the rise of the right-wing party AfD, historical themes plays a decreasing role.

ABSTRACT. Combining insights from media psychology with computational methods, we conducted a large-scale automated content analysis of user-generated video game reviews to determine the extent to which these reviews can inform readers about game features that are critical for players' gaming experience and can, hence, potentially influence purchase intentions.

ABSTRACT. Based on a longitudinal analysis of far-right Telegram groups and channels (2016-2021), this paper shows that – although Telegram currently functions as essential communication medium for far-right milieus – major platforms (e.g. YouTube) continue to have key significance for far-right strategic communication in terms of recruitment, mobilization, and financing.

ABSTRACT. We describe our use of computational methods to address obstacles in comparative analyses of discourses, drawing on the statistics of information theory to quantitatively describe the relationships between the latent semantic structures of lexically distinctive discourses underlying their superficial distinctions and resemblances.

ABSTRACT. On the analysis of emotion-oriented covert propaganda in images and texts.

ABSTRACT. Using deep learning and NLP techniques, we create an algorithm to automatically extract sources and quotes in news articles. Specifically, we train a two-layer bidirectional LSTM neural network with ELMo word embeddings. Our model achieved 0.94 accuracy on a pre-annotated test set of 2200 Wall Street Journal news articles.

ABSTRACT. Based on computerized text analysis of writings by Chinese top leaders, this paper attempts to trace the changing meanings of a primary political concept of the Chinese Communist Party--“reform”. Findings show how leaders have diverging conceptualization of reform in relation to various policy aspects.

ABSTRACT. There is a growing body of research detecting and measuring such strategies of information control, as distortion and “fake news”. However, large pro-government bias may lead to less media consumption, as citizens will choose not to engage with unreliable news sources and form a demand for alternative media. Autocrats tend to use more subtle information manipulation tactics that are harder to detect. Based on the analysis of news reports of the Russian primary state-controlled channel spanning from 1999 to 2020, the author tested the theory that during economic decline state-controlled media tend to increase the coverage of (1) international news; (2) nationalistic content; (3) stories that affirm regime’s governing abilities. Using a semi-supervised Bayesian classification model and hierarchical topic model with additive regularization (ARTM), the author matched news with a country that it is about and the topic it is covering. Results of regression analysis showed that there is a positive connection between the share of news about the E.U. and the RUB/USD exchange rate, oil price. Granger’s causality also suggests that overall there is some causality running from oil price and RUB/USD exchange rate towards a daily share of news about the E.U., while vice versa is not true. There is ambiguous evidence regarding the share of news about the U.S. and international news in general. Hypotheses about the use of nationalistic content and stories about the government’s activities were not supported.

ABSTRACT. By detecting sentiment and emotional contents from the text in Trump’s Twitter data, this research makes a fine-grained interpretation of Trump's sentiments and emotions in the presidency using Time Varying-Vector Auto-Regression (TV-VAR). Combining text analysis and time-series analysis methods provides a new perspective in the existing text analysis framework.

ABSTRACT. We have applied a novel sentiment analysis tool to private messages exchanged on mobile communication channels, including WhatsApp and SMS, over the course of several months. With the thus-obtained historical sentiment data, we have explored the application of time series methods to reveal possible relationship-altering happenings and their impact.

ABSTRACT. The evolutionary aim of moral content is to provide information on who can be trusted –or not. Yet, online platforms' attention economy has changed the motivations and incentives to engage, create, and share online content. Thus, it remains unclear how moralization impacts people's digital content engagement. Here, using data on more than two million online posts from four online social platforms –Reddit, 8Chan, Twitter, and Facebook– we identify two mechanisms that drive online moral engagement: i) content moralization, and ii) moral penalization. We find that online moral engagement scales sub-linearly with the number of moral words, suggesting decreasing returns to scale in moralization. It also decays exponentially with the ratio of moral words, suggesting that the excess of moralization is penalized. Both effects unveil a previously unknown and universal pattern of moral saturation, which mathematically leads to an optimal level of moralization that maximizes the online moral content engagement. These results provide insights on how people engage in moral content in the digital era and generalize previous research across a variety of online platforms and topics.

ABSTRACT. This study uses a real-world criminal network data to simulate the effect of various types of network disruption on its structural properties and subsequently uses stochastic actor-oriented models to simulate the recovery of the network.

ABSTRACT. Gossip has been proposed to foster cooperation. Empirical studies have mainly examined pro-social gossip. However, gossip is spread for various motives. We propose an ABM in which agents gossip for pro-social, pro-self and emotion-venting reasons. We show that gossip decreases cooperation compared to first-hand-reputation. Pro-social gossip is only marginally beneficial.

ABSTRACT. I present an alternative model of market miscategorization: miscategorization reflects rational producer entry into rugged terrain instead of an audience's bounded rationality. Through retrospective rationalization, categories reflect, but do not cause, producer success. This model accounts for the basic findings of category theory as well as work on category emergence.

ABSTRACT. We study a modified version of the Axelrod model where the agents are allowed to hesitate. We add a disagreement process, intrinsic noise driven, coupled to the dynamics of the system. We show there is no ordered state for any non-zero amount of noise in the thermodynamic limit.

ABSTRACT. We demonstrate the use of cellular automata in modeling and predicting internal migration driven by infrastructure; using annual data from the United Kingdom National Health Service to test the effects of the Northern Powerhouse Rail to increase employment and housing within the area.

ABSTRACT. Impact of COVID-19 policies on a city is full of unknowns. To support policymakers we propose a full-scale replica of a city in an ABM. We found that hard lockdown policy has a controversial spatiotemporal effect. It shifts infections hotspots to the zones that are typically considered to be safe.

ABSTRACT. A complex system, trophic cascade model is used to analyze the environmental consequences of wolf reintroduction to Yellowstone National Park in the western United States.

ABSTRACT. This work investigates the relationship between classical centrality measures, community-aware centrality measures, and network topology. With an extensive analysis using artificial and real-world networks, results show that the mixing parameter and transitivity are the main features affecting the correlation variation between classical and community-aware centrality measures.

ABSTRACT. Faculty hiring networks exhibit sharp inequalities. Understanding the mechanisms driving these patterns would inform new efforts to diversify the academy and shed new light on the role of hiring in shaping which scientific discoveries are made. Here, we investigate mechanisms that can (cannot) describe observed structural inequality.

ABSTRACT. Analyzing human-based acknowledgement networks of academia, we found that reciprocal authors predominantly tend to cite other reciprocal authors rather than non-reciprocal ones and topological features of gender diversity in pairs of reciprocal authors.

ABSTRACT. Here we use data from ICEWS and GDELT to evaluate protest co-occurrence using network theory. The networks of protests for both data sets present a small-world effect indicating that protests can quickly diffuse from one region to any part of the globe.

ABSTRACT. We test some graph and text embedding methods and evaluate how well they capture the disciplinary structure of research published in the American Physical Society (APS). The vector representations that embeddings provide may be helpful in conducting paper-level analyses that provide a new understanding on how scientific knowledge is produced.

ABSTRACT. People who suffer Opioid Use Disorder often call on online communities to seek support during detoxification. In this work, we aim at measuring the behavioral shift of users participating in opioid recovery discussions on Reddit, as well as the social feedback they receive from the community.

ABSTRACT. A complementary currency is a medium of exchange independent of the national currencies, which is based on the agreement of the users and is usually implemented to serve social goals. The Sarafu network is a complementary currency system active in Kenya and managed by the non-profit organization Grassroots Economics. In 2019, they digitize the payment system infrastructure by providing each new member with 400 tokens. The percentage of acceptance as means of payment at each transaction is up to the seller and buyer. From the weighted indegree and outdegree distributions, it is possible to observe that group accounts have a prominent role in the network. In this paper, we will describe the timestamped datasets of transactions and users - from the 26th of January 2020 until the 21st of April 2021. The richness of these datasets gives net-work scientists the opportunity to analyze the structure and the dynamics of the economic network. The economic network includes 50,487 individuals who transacted in total 179,684,432.548 Sarafu (330,939 transactions; $ 1,668,379.12) in the whole observed period. The Sarafu transaction network also includes 149 official group accounts; these are 'Chamas'. Roughly 90% of group accounts are registered with Sarafu as savings cooperatives, specifically, although other forms of the collective financial activity may also be taking place. We focus on the transaction patterns of these group accounts and find them to be consistent with collective savings.

ABSTRACT. This article analyzes how fintech ecosystems have emerged in Austria, Germany, and Switzerland. In particular, we study which role the network position and geographic location of a fintech startup has on bank-fintech alliances. We use a novel dataset of 604 fintech startups and 75 banks over the period from 2015 to 2019 to determine measures of network and geographic centrality. We find that fintechs with a higher network centrality and fintechs located geographically closer to other fintechs are significantly more likely to engage in bank-fintech alliances.

ABSTRACT. Significant gender disparities in economics remain in funding, inclusion in research collaborations, selection of prizewinners. Women represent 30.23% of all the authors in the WOS papers. Women are distributed unevenly across papers and are dramatically concentrated in publications with none or low impact factor.Out of the 770 journals considered, women are the majority in only 30(or 3.90%). Second, women are underrepresented in leadership positions in scientific work. In the full WOS sample, 24.02% of 412,729 papers are solo authored, of which just over one-fourth(26.85%) are authored by women. Interestingly, on average women publish their solo-authored work in higher-impact journals than men, however, they receive fewer citations.Scientific prizes are one of the most important forms of recognition a scientist can receive.We identified 1,260 prize winners of which 1,106 were men and 154 women. While, the presence of women among award winners has been slowly increasing, going from non-existenting in the 1950s and reaching its all time high in the last decade with 15.44% . It is still far away from the 50% needed for parity. For the 25 prizes that offered monetary compensation, we also found gender differences.Between 1999 and 2019, women represented only 13.57% of the winners and on average they received a cash award of $38,131.59,while men received $71,108.02. In other words, for every dollar a man received, female prize winners received $0.53 cents. The situation gets even worse, when one considers the five highest-value awards. Here, out of 72 award winners, only two of them(or 2.78%) were women.

ABSTRACT. We present a novel framework that leverages NLP techniques to quantify the alignment between course syllabi and job postings. Based on neural representations of these documents, three complementary perspectives are included: economic (identifying overlapping skills), educational (mapping instructional design features), and computational (computing overall semantic similarity).

ABSTRACT. We characterize demand for interpretable AI among non-experts, showing it is strongest in settings involving high stakes and scarcity. However, these same factors cause people to sacrifice interpretability for the sake of accuracy. These preferences could drive a proliferation of AI systems making high-impact decisions that are difficult to understand

ABSTRACT. Male dominance in the news is well documented, whereas females are seen as “eye candy” or “inferior”, and are underrepresented and under-examined within the same news categories as their male counterparts. In this paper, we present an initial study on gender bias in abstracts of news articles.