Recent work has shown that front-end code generated by Large Language Models (LLMs) can embed deceptive designs. To assess the magnitude of this problem, identify the factors that influence deceptive design production, and test strategies for reducing deceptive designs, we carried out two studies which generated and analyzed 1,296 LLM-generated web components, along with a design rationale for each. The first study tested four LLMs for 15 common ecommerce components. Overall 55.8% of components contained at least one deceptive design, and 30.6% contained two or more. Occurence varied significantly across models, with DeepSeek-V3 producing the fewest. Interface interference emerged as the dominant strategy, using color psychology to influence actions and hiding essential information. The first study found that prompts emphasizing business interests (e.g., increasing sales) significantly increased deceptive designs, so a second study tested a variety of prompting strategies to reduce their frequency, finding a values-centered approach the most effective. Our findings highlight risks in using LLMs for coding and offer recommendations for LLM developers and providers.

Online platforms are riddled with harassment, which significantly impacts the well-being of users. Unfortunately, the content moderation solutions provided by platforms often disappoint end-users as they fail to equip individuals with sufficient controls for their personal situations. In this work, the author, who personally experienced a sustained harassment campaign on Twitter, decided to regain control by constructing an automated, personalized, and collaborative anti-harassment system to protect herself, which has proven itself to be effective. The experience of developing—and re-developing in the face of repeated platform API changes and restrictions—this personalized content moderation system highlights many design issues that make managing severe online harassment such a challenge and invites critical study of the power dynamics between large online platforms and individual users. Through this analysis, this report aims to inform better designs to help platforms more effectively protect victims.

YouTube is the world's most widely used video platform, with over 70% of content viewed through algorithmic recommendations. While prior audits have examined polarization in YouTube's long-form video recommendations, the platform's fast-growing Shorts feature remains understudied. In this paper, we present the first large-scale audit comparing political content exposure and engagement dynamics across short-form and long-form videos on YouTube. We design a matched audit based on the insight that many news media organizations publish both short and long versions of the same content and collect 50,000 pairs of long-form and short-form video recommendations from both political and nonpolitcal seed videos. We analyze recommendations along several dimensions: the frequency of political recommendations, the diversity of retrieved videos, the engagement those videos receive, and finally, the partisan alignment between recommended videos and seed videos. Our results highlight fundamental differences between each algorithm, which we hope we can inform future research in analyzing the impact of YouTube recommendations.

Peer Reviewed

Voice-based social media platforms that enable attendees to have real-time, ephemeral interactions with each other—such as X-Spaces, Discord, and Clubhouse—have seen considerable growth in recent years. While prior research on these spaces has predominantly focused on moderating harms, our work seeks to understand emergent practices employed by hosts to proactively shape their discussion space— focusing on the facilitation aspect of moderation duties. Drawing on facilitation strategies, we study these practices through three comprehensive studies using mixed-methods: survey of social-audio users, co-design interviews, and analyzing training sessions for hosts. Our findings reveal insights into the issues faced by hosts and attendees, current facilitation practices, opinions on technological solutions, and factors that could be responsible for some of the identified issues such as the available training for hosts. We found that hosts themselves are often significant sources of issues due to practices such as focusing more on self-promotion than facilitating discussions. In addition, host training sessions seem to encourage behaviors that contribute to the negative perception of hosts. We draw on outcomes from co-design interviews to guide the design of future tools to support hosts in facilitating social-audio spaces. Our findings provide insights that could help create a more positive experience for both hosts and attendees.

Microaggressions are subtle, everyday comments or actions that communicate discrimination towards historically marginalized groups. While these can be small experiences, they can cause measurable harm. Researchers have investigated how to reduce the negative experiences that result. But there have only been limited investigations into how to support those efforts with technology. To address this, we conduct a series of fifteen participatory design workshops aimed at designing to reduce the occurrence of and negative impacts from microaggressions. The workshops included 47 participants drawn from communities frequently targeted by microaggressions, with groups formed around gender, race/ethnicity, and disability & accessibility. Our study findings identified four primary themes in designing effective interventions for microaggressions: proactive measures, reward and accountability frameworks, community support mechanisms, and long-term educational resources. Through axial coding, we observed that different groups respond uniquely to these intervention approaches, with distinct preferences for timing, intervention style, and accountability. We identify challenges participants face when intervening around microaggressions and suggest directions for future solutions.

Studying computer science is a journey: people start at different times, travel at different paces, and pause along the way. In this experience report, we describe a peer-led, year-long program designed to welcome students to Computer Science and Engineering as a discipline, department, and academic program. We detail the logistical, curricular, and personnel structures of this program, highlighting design choices we made to (a) open multiple ways to join the program all year, (b) de-emphasize "getting ahead", (c) prioritize reflection, and (d) connect students to existing resources. Throughout, we emphasize the critical role of peer mentors in leading and shaping this space. We share our own lessons learned, as well as reflections from students and mentors on the value of this learning community outside of formal classroom structures.

Selective exposure has long been a concern of HCI researchers as it can lead to ideological polarization and distrust in society. Efforts have tried to reduce selective exposure online by serving diversified news content, but their effectiveness has been limited by users' lack of motivation to engage with the diverse content offered. To address this, we design the NewsGuesser system, which leverages the insight that curiosity can prompt motivation and engagement, by asking readers to guess the source of their news. In interviews with 40 participants, balanced for partisan affiliation, we use NewsGuesser as a probe tool to explore how guessing affects their perceptions of selective exposure. Participants struggled with the guessing game, which revealed a misalignment between users' expectations of different news sources and reality. Faced with the visualizations of the (often inaccurate) guessing results, participants were able to reflect on their own biases and selective exposure. In a number of cases, the guessing process changed participants' impressions of news organizations and some expressed an interest in engaging with more diverse news sources. While many also found the guessing game frustrating, the system and interview results suggest a number of new directions for designing social media and news media platforms.

Recent years have seen a rise in social virtual reality (VR) platforms that allow people to interact in real-time through voice and gestures. The ephemeral nature of communication on these platforms can enable new forms of harmful behavior and new challenges for moderators. We performed virtual field research on three VR environments (AltspaceVR, Horizon Worlds, Rec Room). Based on observing 100 scheduled events, our analysis uncovered 13 distinct types of potentially harmful behaviors enabled by real-time voice, embodied interactions, and platform affordances. We witnessed potential harm at 45% of our observed events; only 24% of these incidents were addressed by moderators. To understand moderation practices, we conducted interviews with 11 moderators to investigate how they assess real-time interactions and how they operate within the current state of moderation tools. Our work sheds light on how moderation tools and practices must evolve to meet the new challenges of social VR.

Online discussion boards, designed to facilitate learning from peers and instructors in an accessible space, are a vital part of course design, especially in large scale computer science classes. Previous work has shown that women in computer science tend to use anonymity more often than men on these boards, a trend not found in humanities, social science or business courses. In this work, we build on these findings using an intersectional lens, analyzing both gender and race/ethnicity. We find this combined analysis reveals differences in anonymity that are not apparent when examining gender alone. For example, we find a significantly greater difference in anonymity use between Hispanic men and women than would be expected from analyzing race/ethnicity and gender independently. We additionally analyze type of content (e.g., questions, answers), course, platform, and data source to characterize the many factors at play in measuring students’ choice to participate anonymously. In doing so, we show that different approaches used in prior work for eliciting information on gender — whether using registrar data, a survey, or imputing gender based on name — changes how over of students are classified, particularly affecting nonbinary students and Asian students. Understanding when students participate anonymously can help educators and platform designers to make students’ experience of online discussion boards more welcoming.

At our large U.S. research-intensive university, Chicano/Latino and Black/African-American students have been disproportionately leaving the Computer Science and Engineering (CSE) majors at a higher rate than students without these identities. To uncover possible reasons for this, we invited students in these majors who identify as Chicano/Latino and Black/African-American to participate in focus groups. Twelve students, all identifying as Latinx/Hispanic, partici- pated in the focus groups. We identify several themes related to challenging aspects of the student experience, spanning physical campus environment, department curriculum and policies, and connections between students. We triangulate these findings with results from a survey measuring sense of belonging, confidence, and obstacles for thousands of students across eight introductory CSE courses. We discuss how these themes relate to actions that departments can take to address these challenges.

Differential privacy is one of the most popular technologies in the growing area of privacy-conscious data analytics. But differential privacy, along with other privacy-enhancing technologies, may enable privacy theater. In implementations of differential privacy, certain algorithm parameters control the tradeoff between privacy protection for individuals and utility for the data collector; thus, data collectors who do not provide transparency into these parameters may obscure the limited protection offered by their implementation. Through large-scale online surveys, we investigate whether explanations of differential privacy that hide important information about algorithm parameters persuade users to share more browser history data. Surprisingly, we find that the explanations have little effect on individuals' willingness to share data. In fact, most people make up their minds about whether to share before they even learn about the privacy protection.

Content moderation systems for social media have had numerous issues of bias, in terms of race, gender, and ability among many others. One proposal for addressing such issues in automated decision making is by designing for contestability, whereby users can shape and influence how decisions are made. In this study, we conduct a series of participatory design workshops with participants from communities that have experienced problems with social media content moderation in the past. Together with participants, we explore the idea of designing for contestability in content moderation and find that users' designs suggest three fruitful, practical avenues: adding representation, improving communication, and designing with compassion. We conclude with design recommendations drawn from participants' proposals, and reflect on the challenges that remain.

Control settings are abundant and have significant effects on user experiences. One example of an impactful but understudied area is feed settings. In this study, we investigated awareness, navigation, and use of feed settings. We began by creating a taxonomy of feed settings on social media and search sites. Via an online survey, we measured awareness of Facebook feed settings. An in-person interview study then investigated how people navigated to and chose to set feed settings on their own feeds. We discovered that many participants did not believe ad personalization feed settings existed. Furthermore, we discovered a misalignment in the expectation and the function of settings, especially of ad personalization settings for many participants. Despite all participants struggling to find at least one setting, participants overall wanted to use settings: 94% altered at least one setting they encountered. From these results, we discuss implications and suggest design guidelines for settings.

Interest has grown in designing algorithmic decision making systems for contestability. In this work, we study how users experience contesting unfavorable social media content moderation decisions. A large-scale online experiment tests whether different forms of appeals can improve users' experiences of automated decision making. We study the impact on users' perceptions of the Fairness, Accountability, and Trustworthiness of algorithmic decisions, as well as their feelings of Control (FACT). Surprisingly, we find that none of the appeal designs improve FACT perceptions compared to a no appeal baseline. We qualitatively analyze how users write appeals, and find that they contest the decision itself, but also more fundamental issues like the goal of moderating content, the idea of automation, and the inconsistency of the system as a whole. We conclude with suggestions for -- as well as a discussion of the challenges of -- designing for contestability.

Algorithms exert great power in curating online information, yet are often opaque in their operation, and even existence. Since opaque algorithms sometimes make biased or deceptive decisions, many have called for increased transparency. However, little is known about how users perceive and interact with potentially biased and deceptive opaque algorithms. What factors are associated with these perceptions, and how does adding transparency into algorithmic systems change user attitudes? To address these questions, we conducted two studies: 1) an analysis of 242 users' online discussions about the Yelp review filtering algorithm and 2) an interview study with 15 Yelp users disclosing the algorithm's existence via a tool. We found that users question or defend this algorithm and its opacity depending on their engagement with and personal gain from the algorithm. We also found adding transparency into the algorithm changed users' attitudes towards the algorithm: users reported their intention to either write for the algorithm in future reviews or leave the platform.

Advances in computer vision and machine learning are changing the way people dress and buy clothes. Given the vast space of fashion problems, where can data-driven technologies provide the most value? To understand consumer pain points and opportunities for technological interventions, this paper presents the results from two independent need-finding studies that explore the gold-standard of personalized shopping: interacting with a personal stylist. Through interviews with five personal stylists, we study the range of problems they address and their in-person processes for working with clients. In a separate study, we investigate how styling experiences map to online settings by building and releasing a chatbot that connects users to one-on-one sessions with a stylist, acquiring more than 70 organic users in three weeks. These conversations reveal that in-person and online styling sessions share similar goals, but online sessions often involve smaller problems that can be resolved more quickly. Based on these explorations, we propose future highly personalized, online interactions that address consumer trust and uncertainty, and discuss opportunities for automation.

Algorithmic prioritization is a growing focus for social media users. Control settings are one way for users to adjust the prioritization of their news feeds, but they prioritize feed content in a way that can be difficult to judge objectively. In this work, we study how users engage with difficult-to-validate controls. Via two paired studies using an experimental system -- one interview and one online study -- we found that control settings functioned as placebos. Viewers felt more satisfied with their feed when controls were present, whether they worked or not. We also examine how people engage in sensemaking around control settings, finding that users often take responsibility for violated expectations -- for both real and randomly functioning controls. Finally, we studied how users controlled their social media feeds in the wild. The use of existing social media controls had little impact on user's satisfaction with the feed; instead, users often turned to improvised solutions, like scrolling quickly, to see what they want.

Medical crowdfunding helps patients receive financial support from their distributed social networks online. However, little is known about who the patient's supporters are, what support they provide, and why. To address this, we interviewed fifteen people involved in medical crowdfunding, including both beneficiaries and supporters. We found that support networks were larger than beneficiaries expected, with strangers offering support. Supporters offered not only monetary but also volunteering contributions including campaign creation, promotion, and external support. However, the emphasis medical crowdfunding interfaces place on monetary contributions led to social issues. Beneficiaries' close friends felt pressured to donate money they could not afford to give. And beneficiaries promoting the campaign worried they would be judged for requesting money. To mitigate these concerns, we suggest making the variety of volunteering contributions more visible and discuss the design challenges of including such signals in existing systems.

Many online platforms use curation algorithms that are opaque to the user. Recent work suggests that discovering a filtering algorithm's existence in a curated feed influences user experience, but it remains unclear how users reason about the operation of these algorithms. In this qualitative laboratory study, researchers interviewed a diverse, non-probability sample of 40 Facebook users before, during, and after being presented alternative displays of Facebook's News Feed curation algorithm's output. Interviews revealed 10 "folk theories' of automated curation, some quite unexpected. Users who were given a probe into the algorithm's operation via an interface that incorporated "seams,' visible hints disclosing aspects of automation operations, could quickly develop theories. Users made plans that depended on their theories. We conclude that foregrounding these automated processes may increase interface design complexity, but it may also add usability benefits.

The outfits people wear contain latent fashion concepts capturing styles, seasons, events, and environments. Fashion theorists have proposed that these concepts are shaped by design elements such as color, material, and silhouette. A dress may be "bohemian" because of its pattern, material, trim, or some combination of them: it is not always clear how low-level elements translate to high-level styles. In this paper, we use polylingual topic modeling to learn latent fashion concepts jointly in two languages capturing these elements and styles. Using this latent topic formation we can translate between these two languages through topic space, exposing the elements of fashion style. We train the polylingual topic model (PLTM) on a set of more than half a million outfits collected from Polyvore, a popular fashion-based social net- work. We present novel, data-driven fashion applications that allow users to express their needs in natural language just as they would to a real stylist and produce tailored item recommendations for these style needs.

Our daily digital life is full of algorithmically selected content such as social media feeds, recommendations and personalized search results. These algorithms have great power to shape users' experiences, yet users are often unaware of their presence. Whether it is useful to give users insight into these algorithms' existence or functionality and how such insight might affect their experience are open questions. To address them, we conducted a user study with 40 Facebook users to examine their perceptions of the Facebook News Feed curation algorithm. Surprisingly, more than half of the participants (62.5%) were not aware of the News Feed curation algorithm's existence at all. Initial reactions for these previously unaware participants were surprise and anger. We developed a system, FeedVis, to reveal the difference between the algorithmically curated and an unadulterated News Feed to users, and used it to study how users perceive this difference. Participants were most upset when close friends and family were not shown in their feeds. We also found participants often attributed missing stories to their friends' decisions to exclude them rather than to Facebook News Feed algorithm. By the end of the study, however, participants were mostly satisfied with the content on their feeds. Following up with participants two to six months after the study, we found that for most, satisfaction levels remained similar before and after becoming aware of the algorithm's presence, however, algorithmic awareness led to more active engagement with Facebook and bolstered overall feelings of control on the site.