While interpretability tools are intended to help people better understand machine learning (ML), we find that they can, in fact, impair understanding. This paper presents a pre-registered, controlled experiment showing that ML practitioners (N=119) spent 5x less time on task, and were 17% less accurate about the data and model, when given access to interpretability tools. We present bounded rationality as the theoretical reason behind these findings. Bounded rationality presumes human departures from perfect rationality, and it is often effectuated by satisficing, i.e., an inclination towards "good enough" understanding. Adding interactive elements---a strategy often employed to promote deliberative thinking and engagement, and tested in our experiment---also does not help. We discuss implications for interpretability designers and researchers related to how cognitive and contextual factors can affect the effectiveness of interpretability tool use.

Online retributive harassment refers to a range of abusive online behaviors targeted at offenders with the intent of sanctioning norm violations. Online retributive harassment is common in online interactions, can be devastating in its effect, and is hard to moderate. We examined the role that negative interface cues (i.e., Dislikes, flags) might have in changing people's harassment perceptions to mitigate this activity. We conducted a 4x2 between-subjects experiment (N = 242) to test the effects of interface cues (control with Likes only vs. Dislikes outnumbering Likes vs. Likes outnumbering Dislikes vs. a flag with Likes) and harassment severity (low vs. high) on how people perceive retributive harassment. We hypothesized that Dislikes and flags, despite the presence of Likes, signal social disapproval and descriptive norms against harassment and thus reduce bystanders' belief that retributive harassment is appropriate, deserved, or justified. We found that Dislikes can be effective when they outnumber Likes in high-severity harassment but backfire when outnumbered by Likes. A flag, contrary to popular practice, does not have a significant mitigating effect on perceptions of retributive harassment. We demonstrate the potential of negative interface cues to signal anti-harassment norms to bystanders and discuss alternatives to social media platforms' one-size-fits-all content moderation approach.

Intelligent systems are limited in their ability to match the fluid social needs of people. We use affordances---people's perceptions of the utilities of a target system---as a means of creating models that provide intelligent systems with a better understanding of how people make decisions. We study affordance-based models in the context of social network site (SNS) usage, a domain where people have complex social needs often poorly supported by technology. Using data collected via a scenario-based survey (N=674), we build two affordance-based models about people's multi-SNS posting behavior. Our results highlight the feasibility of using affordances to help intelligent systems support people's decision-making behavior: both of our models are ~15% more accurate than a majority-class baseline, and they are ~33% and ~48% more accurate than a random baseline for this task. We contrast our approach with other ways of modeling posting behavior and discuss the implications of using affordances for modeling human behavior for intelligent systems.

This study documents users' changes in psychological wellbeing across the lifecycle of their participation in an online community. Through in-depth interviews with 30 long-term users of Everything2.com, and content analysis of their posts, we found that psychological wellbeing plays a large role in the evolution of how users participate in the community over time. Everything2 is a long-running user-generated content site framed as an open encyclopedia. Results suggested that negative psychological wellbeing, such as loneliness and low self-esteem, fueled initial participation; validation and criticism from other users on one's content motivation continued content contribution; but ultimately feelings of relatedness with the community, unrelated to content contribution, was what retained users. Absence of social connections in the online community, as well as improved wellbeing offline, led to exit.

Norms are central to how online communities are governed. Yet, norms are also emergent, arise from interaction, and can vary significantly between communities---making them challenging to study at scale. In this paper, we study community norms on Reddit in a large-scale, empirical manner. Via 2.8M comments removed by moderators of 100 top subreddits over 10 months, we use both computational and qualitative methods to identify three types of norms: macro norms that are universal to most parts of Reddit; meso norms that are shared across certain groups of subreddits; and micro norms that are specific to individual, relatively unique subreddits. Given the size of Reddit's user base---and the wide range of topics covered by different subreddits---we argue this represents the first large-scale census of the norms in broader internet culture. In other words, these findings shed light on what Reddit values, and how widely-held those values are. We conclude by discussing implications for the design of new and existing online communities.

Courts are increasingly adopting online information and communication technology, creating a need to consider the potential consequences of these tools for the justice system. Using survey responses from 209 litigants who had recently used an online case resolution system, we investigate factors that influenced litigants' experiences of fairness and emotional feelings toward court officials. Our results show that ease of using the online case resolution system, the outcome of the case, and a litigant's perceptions of procedural justice are positively associated both with whether the litigant views the process as fair and whether the litigant ultimately feels positive emotions toward court officials. We also analyze the online explanations litigants offer in their arguments to courts and litigant answers to an open-ended question about their court experiences, and highlight design and practical implications for online systems seeking to improve access to justice.

Online harassment is a pervasive and pernicious problem. Techniques like natural language processing and machine learning are promising approaches for identifying abusive language, but they fail to address structural power imbalances perpetuated by automated labeling and classification. Similarly, platform policies and reporting tools are designed for a seemingly homogenous user base and do not account for individual experiences and systems of social oppression. This paper describes the design and evaluation of HeartMob, a platform built by and for people who are disproportionately affected by the most severe forms of online harassment. We conducted interviews with 18 HeartMob users, both targets and supporters, about their harassment experiences and their use of the site. We examine systems of classification enacted by technical systems, platform policies, and users to demonstrate how 1) labeling serves to validate (or invalidate) harassment experiences; 2) labeling motivates bystanders to provide support; and 3) labeling content as harassment is critical for surfacing community norms around appropriate user behavior. We discuss these results through the lens of Bowker and Star's classification theories and describe implications for labeling and classifying online abuse. Finally, informed by intersectional feminist theory, we argue that fully addressing online harassment requires the ongoing integration of vulnerable users' needs into the design and moderation of online platforms.

Increasing numbers of American parents identify as lesbian, gay, bisexual, or transgender (LGBT). Shifting social movements are beginning to achieve greater recognition for LGBT parents and more rights for their families; however, LGBT parents still experience stigma and judgment in a variety of social contexts. We interviewed 28 LGBT parents to investigate how they navigate their online environments in light of these societal shifts. We find that 1) LGBT parents use social media sites to detect disapproval and identify allies within their social networks; 2) LGBT parents become what we call incidental advocates, when everyday social media posts are perceived as advocacy work even when not intended as such; and 3) for LGBT parents, privacy is a complex and collective responsibility, shared with children, partners, and families. We consider the complexities of LGBT parents' online disclosures in the context of shifting social movements and discuss the importance of supporting individual and collective privacy boundaries in these contexts.

Undergraduates interviewed about privacy concerns related to online data collection made apparently contradictory statements. The same issue could evoke concern or not in the span of an interview, sometimes even a single sentence. Drawing on dual-process theories from psychology, we argue that some of the apparent contradictions can be resolved if privacy concern is divided into two components we call intuitive concern, a "gut feeling," and considered concern, produced by a weighing of risks and benefits. Consistent with previous explanations of the so-called privacy paradox, we argue that people may express high considered concern when prompted, but in practice act on low intuitive concern without a considered assessment. We also suggest a new explanation: a considered assessment can override an intuitive assessment of high concern without eliminating it. Here, people may choose rationally to accept a privacy risk but still express intuitive concern when prompted.

Many existing studies of social media focus on only one platform, but the reality of users' lived experiences is that most users incorporate multiple platforms into their communication practices in order to access the people and networks they desire to influence. In order to better understand how people make sharing decisions across multiple sites, we asked our participants (N=29) to categorize all modes of communication they used, with the goal of surfacing their mental models about managing sharing across platforms. Our interview data suggest that people simultaneously consider "audience" and "content" when sharing and these needs sometimes compete with one another; that they have the strong desire to both maintain boundaries between platforms as well as allowing content and audience to permeate across these boundaries; and that they strive to stabilize their own communication ecosystem yet need to respond to changes necessitated by the emergence of new tools, practices, and contacts. We unpack the implications of these tensions and suggest future design possibilities.

From amateur creativity to social media status updates, nearly every Internet user is also a content creator-but who owns that content? Policy, including intellectual property rights, is a necessary but often invisible part of online content sharing and social computing environments. We analyzed the copyright licenses contained in the Terms of Service of 30 different websites where users contribute content, then conducted a survey to match perceptions of copyright terms to the reality. We found that licensing terms vary in unpredictable ways, and that user expectations and opinions differ by license and by type of website. Moreover, the most undesirable terms, such as right to modify, appear more frequently than users expect. We argue that users care about how their content can be used yet lack critical information. Site designers should take user needs and community norms into account in creating and explaining copyright policies.

While extensive research has investigated the risks of children sharing their personal information online, little work has investigated the implications of parents sharing personal information about their children online. Drawing on 102 interviews with parents in the U.S., we investigate how parents decide what to disclose about their children on social network sites (SNSs). We find that mothers take on the responsibility of sharing content about their children more than fathers do. Fathers are more restrictive about sharing to broad and professional audiences and are concerned about sharing content that could be perceived as sexually suggestive. Both mothers and fathers work to leverage affordances of SNSs to limit oversharing. Building on prior work, we explore parental disclosure management, which describes how parents decide what to share about their children online. We also describe an emerging third shift of work that highlights the additional work parents take on to manage children's identities online. We conclude with theoretical and practical implications for designing SNSs to better support family life online.

Social media sites are increasingly adopted by small nonprofit organizations (NPOs) to help them meet their public engagement goals. However, several characteristics of small organizations make it hard for them to effectively use social media sites. We present findings from interviews with 26 small NPOs' social media professionals on how they use multiple social media sites to support public engagement. Small NPOs use multiple social media sites to engage with different stakeholders toward various ends. However, these NPOs are not using social media to its full potential with regard to community-building and action mobilization. Several challenges in small NPOs, such as ineffective measurement of social media performance, deficient organizational resources, and lack of control over work, lead to strong tensions between social media engagement strategies and outcomes. Drawing on these findings, we present several practical implications for the design of successful public engagement social media tools.

Research suggests that social network sites can support social capital exchanges, which are often triggered by requests for assistance, such as seeking recommendations or asking for favors. Responsiveness to these requests for help is important to study because these interactions have the potential to affect users' overall satisfaction with the experience of using SNSs, signal social grooming functions that are an essential part of relationship maintenance, and affect social capital processes. In this paper, we study a corpus of public status updates posted to Facebook (N=7,466) in order to identify the pattern of responses to status updates that attempt to mobilize resources from the poster's Facebook network. Findings suggest that mobilization requests are treated differently than other kinds of posts; posts that attempt to mobilize help receive more comments than non-mobilization attempts. Additionally, responses occur more quickly and are shaped by the type of support requested (e.g., a recommendation vs. a favor). These findings help us better understand the role of help-seeking behaviors in the social capital conversion process as it unfolds via social media.

Past research has demonstrated a link between perceptions of social capital and use of the popular social network site, Facebook. Williams' Internet Social Capital Scales, based on Putnam's formulation, tap into sub-dimensions of social capital that have not been broadly used yet may enlighten our understanding of the different ways in which connecting with others online can facilitate access to resources embedded within our social relationships. In this study, we segment Williams' Internet Social Capital Scales into various sub-dimensions using factor analysis and explicate the distinct facets of social capital through a lab experiment in which Facebook users (N=98) request a small favor from their Facebook network. We find that some sub-dimensions play a significant role in getting favors from Facebook friends while bonding and bridging social capital do not significantly predict responses to favor requests.

Research has identified a link between Facebook use and bridging social capital, which speaks to the informational resources provided by a diverse network of connections. In order to explicate the mechanism through which Facebook may help individuals mobilize these embedded informational and support resources, this study explores the role of bridging social capital, question type, and relational closeness on the perceived utility and satisfaction of information obtained through questions posed to one's network of Facebook Friends through the status update feature. Employing a mixed-method approach, we utilize survey data collected from a sample of non-academic university staff (N=666), as well as actual Facebook question examples and responses collected during a follow-up lab session from a subset of this sample (N=71). Results indicate that question-askers' bridging social capital positively predicts the utility of responses received on SNS, while useful responses are more likely to be received from weaker ties.

Although Facebook is the largest social network site in the U.S. and attracts an increasingly diverse userbase, some individuals have chosen not to join the site. Using survey data collected from a sample of non-academic staff at a large Midwestern university (N=614), we explore the demographic and cognitive factors that predict whether a person chooses to join Facebook. We find that older adults and those with higher perceived levels of bonding social capital are less likely to use the site. Analyzing open-ended responses from non-users, we find that they express concerns about privacy, context collapse, limited time, and channel effects in deciding to not adopt Facebook. Finally, we compare non-adopters against users who differ on three dimensions of use. We find that light users often have social capital outcomes similar to, or worse than, non-users, and that heavy users report higher perceived bridging and bonding social capital than either group.

In this paper we explore effects of perceived ideology of news outlets on consumption and sharing of news in Twitter. Selective exposure theory suggests that when given access to a broad range of information, people will tend to consume and share news that confirms their existing beliefs and biases. We find that users share news in similar ways regardless of outlet or perceived ideology of outlet, and that as a user shares more news content, they tend to quickly include outlets with opposing viewpoints. This suggests that while perceived ideology does not inspire most Twitter users to treat liberal or conservative news outlets differently, it is a factor in their news consumption and sharing. Specifically, users in our sample who sent multiple tweets tended to increase the ideological diversity in news they shared within two or three tweets, and users' information diversity increased as their number of tweets sent increased.

Facebook has become an increasingly important tool for people engaging in a range of communication behaviors, including requesting help from their social network to address information needs. Through a study of 614 staff members at a large university, we show how social capital, network characteristics, and use of Facebook are related to how useful individuals find Facebook to be for informational purposes and their propensity to seek different types of information on the site. We find that bridging social capital and engagement with one's network through directed communication behaviors are important predictors of these dimensions of information seeking; furthermore, a number of demographic and usage behavior differences exist between those who choose to engage in information-seeking behaviors on Facebook and those who do not. Finally, when predicting information-seeking behaviors, we identify a significant interaction between users' perceptions of Facebook as appropriate for purposes beyond the purely social and their engagement with their network.

In this paper, we examine how user ratings of content produced for an online community are taken into account by administrators when they decide whether to delete content. Incorporating about 10 years of server data from the online peer-production community Everything2, we looked at how specific features of voting predicted deletion of posts. We found that not all types of voting are the same: negative voting of users was the strongest factor explaining deletion of a Write-up. Receiving a positive vote from a member with higher status decreases the chances of deletion, while receiving a positive vote from a user with neutral status has a very little effect on the deletion of content.

User activities in peer-production communities have mainly been examined under the assumption that individuals are rational individuals who are always cognizant of what they are doing and why. We argue that not all use is the same; while some behaviors are governed by conscious motivations, others may be a habitual response that is developed out of routine. We take a more granular approach to explaining what people are doing in online communities and how motivations and habits explain their use of specific features. In the context of the peer-production community Everything2 we employ both server log data and self-report, finding that habit is a non-conscious-driven behavior that is more associated with less cognitively-demanding tasks than content production.

Social computing and social media systems depend on contributions from users. We posit the existence of a latent demand for contribution: many users want to contribute but don't. We then test a simple interface that can induce these users to actually contribute: we display a popup window asking users to contribute. In a real-world randomized field experiment, we found that asking them to contribute right now is ineffective, but reminding the users to contribute actually leads to approximately a 23% increase in contributions with no reduction in quality. However, this effect wanes as users habituate to the popups.

One of the biggest challenges faced by social media sites like Wikipedia is how to motivate users to contribute content. Research continues to demonstrate that only a small percentage of users contribute to user-generated content sites. In this study we assess the results of a Wikimedia Foundation initiative, which had graduate and undergraduate students from 22 U.S. universities contribute content to Wikipedia articles as part of their coursework. 185 students were asked about their participation in the initiative and their intention to participate on Wikipedia in the future. Results suggest that intentions to continue contributing are influenced by the initial attitude towards the class, and the degree to which students perceived they were writing for a global audience.

A consistent theoretical and practical challenge in the design of socio-technical systems is that of motivating users to participate in and contribute to them. This study examines the case of Everything2.com users from the theoretical perspectives of Uses and Gratifications and Organizational Commitment to compare individual versus organizational motivations in user participation. We find evidence that users may continue to participate in a site for different reasons than those that led them to the site. Feelings of belonging to a site are important for both anonymous and registered users across different types of uses. Long-term users felt more dissatisfied with the site than anonymous users. Social and cognitive factors seem to be more important than issues of usability in predicting contribution to the site.

Student athletes at U.S. universities are bound by rules affecting their participation in their sport and are highly visible to their fellow students and a larger public of fans. This difference makes them more likely than other students to be sensitive to issues of impression management and use of social network sites (SNSs). In this paper, we show how student athletes at a large university engage with the social network site Facebook compared with their fellow students, including differences in the size of their networks, reported uses of the site, and perceptions about their audience. This work shows that while student athletes have a higher anticipation of being watched, they have similar uses and concerns compared to other students.

As social computing systems persist over time, the user experiences and interactions they support may change. One type of social computing system, Social Network Sites (SNSs), are becoming more popular across broad segments of Internet users. Facebook, in particular, has very broad participation amongst college attendees, and has been growing in other populations as well. This paper looks at how use of Facebook has changed over time, as indicated by three consecutive years of survey data and interviews with a subset of survey respondents. Reported uses of the site remain relatively constant over time, but the perceived audience for user profiles and attitudes about the site show differences over the study period.

Using data from a popular online social network site, this paper explores the relationship between profile structure (namely, which fields are completed) and number of friends, giving designers insight into the importance of the profile and how it works to encourage connections and articulated relationships between users. We describe a theoretical framework that draws on aspects of signaling theory, common ground theory, and transaction costs theory to generate an understanding of why certain profile fields may be more predictive of friendship articulation on the site. Using a dataset consisting of 30,773 Facebook profiles, we determine which profile elements are most likely to predict friendship links and discuss the theoretical and design implications of our findings.

Large-scale online communities need to manage the tension between critical mass and information overload. Slashdot is a news and discussion site that has used comment rating to allow massive participation while providing a mechanism for users to filter content. By default, comments with low ratings are hidden. Of users who changed the defaults, more than three times as many chose to use ratings for filtering or sorting as chose to suppress the use of comment ratings. Nearly half of registered users, however, never strayed from the default filtering settings, suggesting that the costs of exploring and selecting custom filter settings exceeds the expected benefit for many users. We recommend leveraging the efforts of the users that actively choose filter settings to reduce the cost of changing settings for all other users. One strategy is to create static schemas that capture the filtering preferences of different groups of readers. Another strategy is to dynamically set filtering thresholds for each conversation thread, based in part on the choices of previous readers. For predicting later readers' choices, the choices of previous readers are far more useful than content features such as the number of comments or the ratings of those comments.

Large numbers of college students have become avid Facebook users in a short period of time. In this paper, we explore whether these students are using Facebook to find new people in their offline communities or to learn more about people they initially meet offline. Our data suggest that users are largely employing Facebook to learn more about people they meet offline, and are less likely to use the site to initiate new connections.

Many virtual communities involve ongoing discussions, with large numbers of users and established, if implicit rules for participation. As new users enter communities like this, both they and existing members benefit when new users learn the standards for participation. Slashdot is a news and discussion site that has developed a system of distributed moderation to provide feedback about the value of posts on their site. This study examines three explanations for how new users learn to participate in a digital community: learning transfer from previous experiences, observation of other members, and feedback from other members. We find that new user behavior is affected by a combination of their viewing behavior, the moderation feedback they receive, and replies to their comments.

Can a system of distributed moderation quickly and consistently separate high and low quality comments in an online conversation? Analysis of the site Slashdot.org suggests that the answer is a qualified yes, but that important challenges remain for designers of such systems. Thousands of users act as moderators. Final scores for comments are reasonably dispersed and the community generally agrees that moderations are fair. On the other hand, much of a conversation can pass before the best and worst comments are identified. Of those moderations that were judged unfair, only about half were subsequently counterbalanced by a moderation in the other direction. And comments with low scores, not at top-level, or posted late in a conversation were more likely to be overlooked by moderators.