Automatic Identification of Harmful, Aggressive, Abusive, and Offensive Language on the Web: A Survey of Technical Biases Informed by Psychology Literature

The automatic detection of conflictual languages (harmful, aggressive, abusive, and offensive languages) is essential to provide a healthy conversation environment on the Web. To design and develop detection systems that are capable of achieving satisfactory performance, a thorough understanding of the nature and properties of the targeted type of conflictual language is of great importance. The scientific communities investigating human psychology and social behavior have studied these languages in details, but their insights have only partially reached the computer science community. In this survey, we aim both at systematically characterizing the conceptual properties of online conflictual languages, and at investigating the extent to which they are reflected in state-of-the-art automatic detection systems. Through an analysis of psychology literature, we provide a reconciled taxonomy that denotes the ensemble of conflictual languages typically studied in computer science. We then characterize the conceptual mismatches that can be observed in the main semantic and contextual properties of these languages and their treatment in computer science works; and systematically uncover resulting technical biases in the design of machine learning classification models and the dataset created for their training. Finally, we discuss diverse research opportunities for the computer science community and reflect on broader technical and structural issues.

Locating Identities in Time: An Examination of the Formation and Impact of Temporality on Presentations of the Self through Location-Based Social Networks

Studies of identity and location-based social networks (LBSN) have tended to focus on the performative aspects associated with marking one's location. Yet these studies often present this practice as being an a priori aspect of locative media. What is missing from this research is a more granular understanding of how this process develops over time. Accordingly, we focus on the first 6 weeks of 42 users beginning to use an LBSN we designed and named GeoMoments . Through our analysis of our users' activities, we contribute to understanding identity and LBSN in two distinct ways. First, we show how LBSN users develop and perform self-identity over time. Second, we highlight the extent these temporal processes reshape the behaviors of users. Overall, our results illustrate that although a performative use of GeoMoments does evolve, this development does not occur in a vacuum. Rather, it occurs within the dynamic context of everyday life, which is prompted, conditioned, and mediated by the way the affordances of GeoMoments digitally organize and archive past locational traces.

Building Personalized Trust: Discovering What Makes One Trust and Act on Facebook Posts

What makes one trust or distrust a post on Facebook? What makes one willing to take an action on the post, such as sharing it with friends, following its advice, or even making a donation for its cause? We hypothesize that personal factors in addition to the characteristics of a post would be associated with one's trust and willingness to take actions on the post. To test our hypothesis, we have examined the relationships between a wide variety of personal factors and how users perceived the credibility of a post on Facebook and their associated behavior—willingness to act on the post with the acceptance of potential risks. Our results from a study involving 400 participants reveal several new findings. Notably, beyond message properties, we have discovered that personal factors are significantly associated with how users perceived the credibility of Facebook posts. Not only do our findings extend prior work on studying the relationships between information properties and users' trust in Facebook posts, but they also indicate the potential of tailoring intended messages to an audience based on their personal factors for establishing desired trust. Moreover, we have found that a combination of personal factors is associated more with users' willingness to act on a post than their perceptions of credibility alone. This finding deepens the understanding of users' behavior on Facebook beyond learning their subjective perceptions. It also informs how to potentially tailor intended messages to an audience based on the audience's personal factors so that the audience can act on the messages beyond just perceiving the credibility of the messages. We also discuss the implications of our findings for building personalized trust between message authors and readers on Facebook and beyond.

Community and Capital: Experiences of Women Game Streamers in Southeast Asia

This study explores how women game live streamers in Southeast Asia make sense of their experiences as performers and gamers on streaming platforms dominated by Western products and performers. We conducted 13 in-depth interviews guided by an interpretive phenomenological approach to understand their experiences. Women streamers strive to develop audience communities and gain acceptance in the larger gaming community, in part by successfully displaying their own gaming capital. However, they face challenges regarding audience connection and maintenance, presenting their own femininity amid stereotypes and misogyny, and the influence of their respective cultures on their success as performers. We discuss directions of study that further explore gaming and streaming as a form of cultural labor in Asia and the world.

Experimental Investigation of Technical and Human Factors Related to Phishing Susceptibility

This article reports on a simulated phishing experiment targeting 6,938 faculty and staff at George Mason University. The three-week phishing campaign employed three types of phishing exploits and examined demographic, linked workstation/network monitoring audit data, and a variety of behavioral and psychological factors measured via pre- and post-campaign surveys. While earlier research studies have reported disparate effects of gender and age, the present results suggest that these effects are not significant or are of limited strength and that other underlying factors may be more important. Specifically, significant differences in phishing susceptibility were obtained for different email contexts and based on whether individuals have been successfully phished before (these people were more likely to succumb to subsequent phishing emails in our study). Further, participants who responded to phishing exploits scored higher on impulsivity than the non-clickers. Also, participants whose survey responses indicated that they had more appropriate online “security hygiene habits,” such as checking the legitimacy of links, were less likely to be successfully phished in our campaign. Participants whose post-campaign survey responses indicated that they were suspicious of a phishing email message in our campaign were far less likely to click on the phishing link than those who were not suspicious. Similar results were obtained for judgments of pertinence of the email. Participants who indicated that they thought about the negative consequences of clicking the link were less likely to do so than participants who did not think about the negative consequences. Implications for effective training and awareness are discussed.

On the Alignment Between Self-Declared Gender Identity and Topical Content from Wikipedia

Wikipedia is an important information source for much of the world. One well-established problem is that editors of Wikipedia are overwhelmingly men. This gender gap in participation has resulted in a concern that the content suffers biases as a result of the bias in participation. This problem is hard to study, because the relationships between participation, gender identity, and content have not been established. Prior studies, mostly with children, have shown some differences in topical preferences based on sex. However, this issue has not been studied with adults and has not been considered from more than a binary stance. In this study, we work to understand how gender identity relates to topical preferences. Through an empirical study, we ask participants to declare a gender identity and then present them with pairs of topical article content from Wikipedia. Through thousands of participants and tens of thousands of paired content trials, we uncover relationships between self-declared gender identity and topical preferences. Further, by focusing on topics that have a statistically significant bias, we leverage two of Wikipedia's category systems to illustrate relative categorical differences that are similar to categorical differences described in prior work. The discussion focuses on the subtly of these differences, potential future research, and the implications for interventions based on topical content. Further, the results help us reflect on relationships that might explain the persistent and worsening gender gap in participation.

Augmented Virtual Reality Meditation

In a novel experimental setting, we augmented a variation of traditional compassion meditation with our custom-built VR environment for multiple concurrent users. The presence of another user’s avatar in shared virtual space supports social interactions and provides an active target for evoked compassion. The system incorporates respiration and brainwave-based biofeedback to enable closed-loop interaction of users based on their shared physiological state. Specifically, we enhanced interoception and the deep empathetic processes involved in compassion meditation with real-time visualizations of: breathing rate, level of approach motivation assessed from EEG frontal asymmetry, and dyadic synchrony of those signals between two users. We manipulated these interventions across eight separate conditions (dyadic or solo meditation; brainwave, breathing, both or no biofeedback) in an experiment with 39 dyads (N=8), observing the effect of conditions on self-reported experience and physiological synchrony. We found that each different shared biofeedback type increased users’ self-reported empathy and social presence, compared to no-biofeedback or solo conditions. Our study illustrates how dyadic synchrony biofeedback can expand the possibilities of biofeedback in affective computing and VR solutions for health and wellness.

Deliberative Breakdowns in the Social Representation Process

This article examines breakdowns that occur when readers at partisan news websites attempt to understand a challenging news event. We conduct the work with the 2017 Alabama senate race as the empirical context marked by the nomination of Republican Roy Moore (a challenging news event for the left-leaning readers), and the story of his alleged sexual misconduct (a challenging news event for the right-leaning readers). To examine how readers attempt to understand these events, we scrape reader comments from two partisan news websites. Our analysis relies on and further elaborates the social representation theory and argumentation theory to identify obstacles that prevent successful progression of social representation processes: rhetorical, epistemological, and emotional breakdowns. The findings from our data reveal indicators of rhetorical breakdowns (greater occurrence of fallacious and non-argumentative reader comments) and epistemological breakdowns (greater use of doxastic comments) both tied to how challenging the news event is as well as indicators of emotional breakdowns (greater occurrence of attack posture) tied to the emotionally charged nature of the news event. We interpret the findings as a balancing act between protecting pre-existing representations and acknowledging the challenging news event. The indicators of potential breakdowns we find enhance our understanding of partisan political discourse viewed through the lens of social representation processes. The article discusses these contributions, including elaborations to social representation theory, and discusses implications of the work.

The Development of Connective Action during Social Movements on Social Media

In recent years, the development of information communication technologies, such as social media, has changed the way people communicate and engage in social movements. While conventional movements were fought in the streets, social media has enabled movements to take place online. In this article, we aim to investigate the role of social media during social movements that evolve online under the scope of the theory of connective action. Specifically, we examined Twitter communication during the #metoo debate. To this end, we examined two datasets (2017 and 2019) and combined methods from social media analytics to identify influential users, with a manual content analysis to classify influential users into roles. Likewise, a manual classification found distinct communication categories. Through regression analysis, we were able to gage the individual contribution of these categories and roles based on the retweet probability. This study introduces for the first time the terms of connective action starters and maintainers.

Sociability in Virtual Reality

This study investigates sociability in the context of immersive Virtual Reality (VR). A Design Science Research process was applied, and three iterative development versions of a VR application were studied. Sociability around the technology was investigated with two theoretical perspectives: social presence and social interactions. The results of qualitative interviews and observations as well as a quantitative experiment are combined to make a proposal for four design principles enhancing the beneficial sociability of VR systems. These principles address the importance of ease-of-use and personalization in the activation of a customer, customer empowerment and emancipation through familiarization, creating real-world connections and surprising content, as well as adding informative elements and streamlining the customer encounter and service process. The proposed design principles are empirically linked through social behaviors and emotions to cognitive outcomes representing beneficial sociability, e.g., improved customer motivation, willingness-to-share, value communications, co-creation, and co-innovation.