Recent terrorist attacks have increased the need to examine the public's response to such threats. This study focuses on the content of Twitter messages related to the 2016 terrorist attack on the Berlin Christmas market. We complement the collective sense-making perspective with the terror management theory (TMT) perspective to understand why people used Twitter in the aftermath of the attack. We use structural topic modeling to analyze our dataset of 51,000 tweets. Our results indicate that people used Twitter to make sense of the events and as part of typical reactions in TMT, that is, to validate their own worldviews and maintain their self-esteem. In accordance with TMT, we found that people used Twitter to search for meaning and value, show sympathy for victims and their families, or call for tolerance, but also to express nationalistic sentiment and greater hostility toward values and views other than their own. We further show that topics varied over the course of the attack and in the days that followed. Whereas in the first two days there were many emotion-related tweets and operational updates, subsequent days saw more opinionrelated tweets. Our findings contribute to the literature on collective behavior in the aftermath of terrorist attacks.

Since its foundation in 2005, YouTube, which is considered to be the largest video sharing site, has undergone substantial changes. Over the last decade, the platform developed into a leading marketing tool used for product promotion by social media influencers. Past research indicates that these influencers are regarded as opinion leaders and cooperate with brands to market products on YouTube through electronic-word-of-mouth mechanisms. However, surprisingly little is known about the magnitude of this phenomenon. In our article, we make a first attempt to quantify product promotion and use an original dataset of 139,475 videos created by German YouTube channels between 2009 and 2017. Applying methods for automated content analysis, we find that YouTube users indeed are confronted with an ever-growing share of product promotion, particularly in the beauty and fashion sector. Our findings fuel concerns regarding the social and economic impact of influencers, especially on younger target groups.

We study ethnic discrimination in the sharing economy using the example of Europe's largest carpooling marketplace. Based on a unique dataset with more than 17,000 rides, we estimate the effects of drivers' perceived name origins on the demand for rides. The results show sizable ethnic penalties. Further analyses suggest that additional information about actors in this market decreases the magnitude of ethnic discrimination. Our findings broaden the perspective of ethnic discrimination by shedding light on subtle, everyday forms of discrimination in social markets and informing ongoing discussions about ways to address discrimination in an era in which markets increasingly move online.

Past research indicates that Sociology is a low-consensus discipline, where different schools of thought have distinct expectations about suitable scientific practices. This division of Sociology into different subfields is to a large extent related to methodology and choices between qualitative or quantitative research methods. Relying on theoretical constructs of the academic prestige economy, boundary demarcation and taste for research, we examine the methodological divide in generalist Sociology journals. Using automated text analysis for 8737 abstracts of articles published between 1995 and 2017, we discover evidence of this divide, but also of an entanglement between methodological choices and different research topics. Moreover, our results suggest a marginally increasing time trend for the publication of quantitative research in generalist journals. We discuss how this consolidation of methodological practices could enforce the entrenchment of different schools of thought, which ultimately reduces the potential for innovative and effective sociological research.

Image recognition systems offer the promise to learn from images at scale without requiring expert knowledge. However, past research suggests that machine learning systems often produce biased output. In this article, we evaluate potential gender biases of commercial image recognition platforms using photographs of U.S. members of Congress and a large number of Twitter images posted by these politicians. Our crowdsourced validation shows that commercial image recognition systems can produce labels that are correct and biased at the same time as they selectively report a subset of many possible true labels. We find that images of women received three times more annotations related to physical appearance. Moreover, women in images are recognized at substantially lower rates in comparison with men. We discuss how encoded biases such as these affect the visibility of women, reinforce harmful gender stereotypes, and limit the validity of the insights that can be gathered from such data.