In response to increasing social engineering attacks, Unified Payments Interface (UPI) apps have implemented a variety of nudges to deter users from responding to fraudulent payment requests. We conducted a scenario-based semi-structured interview study with 46 Indian participants who tested the impact of various user interface nudges to help them discern a mixture of fraudulent and non-fraudulent payment scenarios. Our study revealed a multi-stage de cision-making process: participants relied on a digital literacy and security concern-based risk assessment, which is further influenced by their trust in the recipient as well as the perceived need of the financial transaction. Our results demonstrate that while most nudges help to combat fraudulent transactions, participants perceived “badges” (a nudge signaling the trustworthiness of the receiver) to be most effective in combating fraud. We conclude with concrete recommendations for current and future UPI developers.

Algorithms frequently manage online advertising markets, aligning advertisements with article topics. Our work investigates how users perceive the relevance of ads to articles when ads are placed using different keyword extraction algorithms, including Large Language Models (LLMs), and how transparency about the placement procedure influences these perceptions and behavioral intentions. We conducted an online user experiment (N = 498) where ads are matched with news articles using the keyword extraction methods TF-IDF, KeyBERT, and DeepSeek. Results indicate that lightweight methods can match advanced LLMs in delivering high user-perceived ad-article relevance, which in turn fosters click and purchase intentions. However, providing explanations for the ad-article placements by displaying extracted keywords reduces ad interest and thereby weakens behavioral intentions, while simultaneously increasing perceived relevance and moderating algorithm effects. These findings highlight the complex impact of transparency-increasing explanations and suggest that algorithmic precision metrics must be complemented by user perception and intention measures.

The increasing use of online cookies for data collection has raised privacy concerns, including risks of price discrimination and targeted advertising. Yet, many users struggle to understand different types of cookies and their implications. This highlights a need for targeted educational initiatives aimed at empowering users to better understand online cookies and online privacy. This study examines age differences in the effectiveness of four privacy education modalities—text, interactive tutorial, chatbot, and video—on improving knowledge and behavior related to online tracking. Using a mixed-design experiment, we found that all modalities enhanced privacy knowledge and behaviors among younger and older adults. However, preferences and effectiveness varied by age group: younger adults responded best to chatbot and text formats, while older adults benefited more from the interactive tutorial and video. These findings highlight the need to tailor privacy education to users’ learning preferences to improve engagement and outcomes across age groups.

This study explores whether privacy literacy, self-efficacy, and concerns mediate the age-related effects on privacy decision behavior. To study privacy decision behavior, we designed an experiment that integrates both heuristic and cognitive manipulations in the decision scenario. 625 old and younger adults participated in the experiment and used our web-based application, “RecipeDigger.” The application recorded users’ privacy decision behavior in the form of accepting or rejecting cookies, which offered a personalized service. Our findings indicate that some of the differences in privacy decision-making between older and younger adults can be traced to having different levels of privacy literacy. Older and younger adults with higher privacy literacy can better align their privacy preferences with their disclosure behavior. By bridging the gap between psychological theories and privacy research, this study provides a comprehensive understanding of the factors influencing privacy decisions among older and younger adults, offering methodological, theoretical, and policy implications.

Fairness in artificial intelligence has garnered increasing attention due to concerns about discriminatory AI-based decision-making, prompting the development of numerous mitigation approaches. However, most existing methods assume that demographic information is readily available, which may not align with real-world scenarios where such information is often incomplete. To this end, this paper tackles the pervasive yet overlooked challenge of developing fair machine learning algorithms with limited demographics. Specifically, we explore leveraging limited demographic information to accurately infer missing demographics while simultaneously evaluating and optimizing model fairness. We argue that this approach better aligns with common real-world socially sensitive scenarios involving limited demographics. Extensive experiments on three benchmark datasets highlight the effectiveness of the proposed method, surpassing state-of-the-art with significant gains in fairness while maintaining comparable utility.

Cross-cultural communication often results in misaligned norms and expectations, leading to misunderstandings or harm.As the internet increasingly facilitates cross-cultural communication online, such misalignments also increase.However, there is an opportunity to use Large Language Models (LLMs) to detect such misunderstandings and assist in addressing them.To that end, this study investigates whether cross-cultural norm violations can be detected and mitigated using popular LLMs.Using a set of carefully constructed cross-cultural communication scenarios, half of which present norm violations, we test the ability of OpenAI's GPT-4 Omni (GPT-4o) model to identify cross-cultural communication norm violations.We find that GPT-4o classification accuracy varies by the stated age, gender, and nationality of the communicators described in the scenarios, suggesting a lack of fairness and a potential cultural gap in GPT-4o's detection.

As AI technologies advance, teams are beginning to see AI transition from a tool to a full-fledged teammate. Introducing an AI teammate brings several challenges, ranging from how human teammates perceive their new AI teammates from an affective standpoint to how AI should engage in the various teaming behaviors that make up effective teamwork. The current study used a mixed factorial survey and structural equation modeling to assess how participants in hypothetical human-AI teams respond to various forms of AI information-sharing, including information related to explainability, back-up behavior, situational awareness, and augmenting team memory. The study's results found that AI design features related to situational awareness and augmenting the teams' memory had the strongest effect on participants' attitudes and perceived team cognition with their teammates. However, much of this effect was mediated by participants' affective attitudes towards the AI as a teammate, with higher ratings leading directly to higher levels of perceived team cognition constructs. These results highlight the importance of fostering positive attitudes towards AI teammates, such as trust and cohesion in human-AI teams, to support the development of effective team cognition and the ability of AI information-sharing to bring about such positive impacts.

ABSTRACT This study examines how subtitles and image visualizations influence gaze behavior, working alliance, and behavior change intentions in virtual health conversations with ECAs. Visualizations refer to images on a 3D model TV and text on a virtual whiteboard, both reinforcing key content conveyed by the ECA. Using a 2 2 factorial design, participants were randomly assigned to one of four conditions: no subtitles or visualizations (Control), subtitles only (SUB), visualizations only (VIS), or both subtitles and visualizations (VISSUB). Structural equation path modeling showed that SUB and VIS individually reduced gaze toward the ECA, whereas VISSUB moderated this reduction, resulting in less gaze loss than the sum of either condition alone. Gaze behavior was positively associated with working alliance, and perceptions of enjoyment and appropriateness influenced engagement, which in turn predicted behavior change intentions. VIS was negatively associated with behavior change intentions, suggesting that excessive visual input may introduce cognitive trade‐offs.

Several social networking sites offer automated machine translation of posts, but authors usually have no access to view or modify these translations. This may increase authors' concerns about whether the translations convey their intended meaning. To address this issue, we test a theory-driven model about human-in-the-loop translation using an online between-subjects experiment (N = 216). In our study, participants write fictitious social media status updates in a language other than English, which are then translated with one of three levels of automation: machine-provided translation with no author modifiability (MPT), where authors can view the machine translation but cannot modify it; author-provided translation (APT), where authors manually write the translation with no machine assistance; and machine-provided translation with author modifiability (MPT-AM), where authors can edit the machine translation of their post. We collect objective and subjective measures of users' experience using the assigned translation feature. The results from a structural equation modeling (SEM) analysis demonstrate that, compared to the MPT and APT conditions, participants in the MPT-AM condition reported higher measures of perceived control, ease of use, and perceived comfort, which in turn predicted higher perceived system effectiveness and ultimately increased intention to use the MPT-AM translation feature. Follow-up interviews ( n = 15) found that participants appreciate the ability to edit and/or remove translations from their posts, even if they did not always do so.

Social media platforms are an effective channel for businesses to reach potential audiences through targeted advertising. As the user base of these platforms expands and diversifies, research on targeted advertising and social media needs to go beyond well-studied Western contexts. In an online survey (n=412), we compared users' privacy-related perceptions and behaviors regarding targeted ads on social media in the United States (as a baseline representing Western contexts) and three South Asian countries: Bangladesh, India, and Pakistan. We found that participants in the US perceived significantly fewer benefits and more concerns related to security and privacy about targeted ads than those in the three South Asian countries. We also identified that individual's cultural values and religious affiliations influenced the observed cross-country variances. For instance, US participants identified less with vertical collectivism and vertical individualism than South Asian participants; these two cultural dimensions were, in turn, positively associated with perceived benefits. Our findings highlight the limitation of using one's country as a proxy for culture, as our findings show users' privacy perceptions regarding targeted advertising on social media are more fundamentally associated with their cultural values and religion. We discuss the corresponding design, education, and regulatory implications for targeted advertising on social media.

Diversifying movie recommendations is an effective way to address choice overload, a phenomenon where recommenders generate lists with highly similar recommendations that are difficult to choose from. However, existing diversification algorithms often rely on latent features, which limits their interpretability and makes it less clear why a particular set of movies is recommended. Given that movies are designed to elicit emotional responses, researchers have suggested leveraging these responses to enhance recommender system performance. This study introduces a novel “emotion diversification” approach, which diversifies movie recommendations based on emotional signals extracted from audience reviews. We evaluate this method against latent and non-diversified baselines in a controlled user study (N = 115), finding that it significantly improves perceived taste coverage and system satisfaction without compromising recommendation quality. Going beyond the traditional rating- and/or interaction data used by traditional recommender systems, our work demonstrates the user experience benefits of extracting emotional data from rich, qualitative user feedback and using it to give users a more emotionally diverse set of recommendations.

Disclosure-based cyberbullying is defined as sharing personal information without consent, often with malicious intentions. This affects young adults and leads to long-term personal damage (i.e., career, reputation, and family). Current strategies on social media platforms fall short of addressing this problem. We engaged 20 young adults (18-34) through a participatory design approach in co-designing solutions that reflect their concerns, needs, and desires. Participants were divided into three groups based on their cyberbullying experience as victim, discloser, or attacker. Our participants designed 15 solutions for prevention, intervention, support, and awareness. These solutions reveal the need for privacy control, anonymity, and autonomy for victims, a restorative justice system, social media accountability, and the involvement of influencers and content creators in awareness and education. We provide implications for progressive and adaptive social media awareness campaigns, a compassionate justice system, and soft control of personal information ownership.

In this study, we explore the effectiveness of persuasive messages endorsing the adoption of a privacy protection technology (IoT Inspector) tailored to individuals’ regulatory focus (promotion or prevention). We explore if and how regulatory fit (i.e., tuning the goal-pursuit mechanism to individuals’ internal regulatory focus) can increase persuasion and adoption. We conducted a between-subject experiment (N = 236) presenting participants with the IoT Inspector in gain ("Privacy Enhancing Technology"—PET) or loss ("Privacy Preserving Technology"—PPT) framing. Results show that the effect of regulatory fit on adoption is mediated by trust and privacy calculus processes: prevention-focused users who read the PPT message trust the tool more. Furthermore, privacy calculus favors using the tool when promotion-focused individuals read the PET message. We discuss the contribution of understanding the cognitive mechanisms behind regulatory fit in privacy decision-making to support privacy protection.

With the pervasive and evolving use of tracking and AI to make inferences about online platform users, it has become imperative for adolescents—a key demographic using such platforms—to develop a deep understanding of these practices to protect their privacy. Traditionally, K-12 cybersecurity education has largely been confined to extracurricular activities, limiting underrepresented students’ access. To resolve this shortcoming, we partnered with a rural-identifying middle school to deliver AI-related privacy education in classrooms. Using Design-Based Research methodology, we identified students’ AI-related privacy learning needs and developed six education modules. This paper focuses on the design, classroom implementation, and evaluation of module #2, covering the privacy threats of Tracking and Pervasive Personalization (TaPP). Student assessment outcomes show they developed transferable foundational knowledge of the privacy implications of tracking and personalization after participating in the TaPP module. Our findings demonstrate the benefits of integrating AI-related privacy education into existing K-12 curricula.

Recent work explores how to educate and encourage users to protect their online privacy. We tested the efficacy of short videos for educating users about targeted advertising on Facebook. We designed a video that utilized an emotional appeal to explain risks associated with targeted advertising (fear appeal), and which demonstrated how to use the associated ad privacy settings (digital literacy). We also designed a version of this video which additionally showed the viewer their personal Facebook ad profile, facilitating personal reflection on how they are currently being profiled (reflective learning). We conducted an experiment (n = 127) in which participants watched a randomly assigned video and measured the impact over the following 10 weeks. We found that these videos significantly increased user engagement with Facebook advertising preferences, especially for those who viewed the reflective learning content. However, those who only watched the fear appeal content were more likely to disengage with Facebook as a whole.

This tutorial provides practical training in designing and conducting online user experiments with recommender systems, and in statistically analyzing the results of such experiments. It covers the development of a research question and hypotheses, the selection of study participants, the manipulation of system aspects and measurement of behaviors, perceptions and user experiences, and the evaluation of subjective measurement scales and study hypotheses. Interested parties can find the slides, example datset, and other resources at https://www.usabart.nl/QRMS/.

Technical and practical advancements in Artificial Intelligence (AI) have led to AI teammates working alongside humans in an area known as human-agent teaming. While critical past research has shown the benefit to trust driven by the incorporation of interaction rules and structures (i.e. etiquette) in both AI tools and robotic teammates, research has yet to explicitly examine etiquette for digital AI teammates. Given the historic importance of trust within human-agent teams, the identification of etiquette’s impact within said teams should be paramount. Thus, this study empirically evaluates the impact of AI teammate etiquette through a mixed-methods study that compares AI teammates that either adhere to or ignore traditional etiquette standards for machine systems. The quantitative results show that traditional etiquette adherence leads to greater trust, perceived performance of the AI, and perceived performance of the team as a whole. However, qualitative results reveal that not all traditional etiquette behaviors have universal appeal due to the presence of individual differences. This research provides the first empirical and explicit exploration of etiquette within human-agent teams, and the results of this study should be used further design specific etiquette behaviors for AI teammates.

Batteryless wearables use energy harvested from the environment, eliminating the burden of charging or replacing batteries. This makes them convenient and environmentally friendly. However, these benefits come at a price. Batteryless wearables operate intermittently (based on energy availability), which adds complexity to their design and introduces usability limitations not present in their battery-powered counterparts. In this paper, we conduct a scenario-based study with 400 wearable users to explore how users perceive the inherent trade-offs of batteryless wearable devices. Our results reveal users’ concerns, expectations, and preferences when transitioning from battery-powered to batteryless wearable use. We discuss how the findings of this study can inform the design of usable batteryless wearables.

The paper reports on two empirical studies that provide the first examination into how the presentation of an AI teammate’s identity, responsibility, and capability impacts humans’ perception surrounding AI teammate adoption before interacting as teammates. Study 1’s results indicated that AI teammates are accepted when they share equal responsibility on a task with humans, but other perceptions such as job security generally decline the more responsibility AI teammates have. Study 1 also revealed that identifying an AI as a tool instead of a teammate can have small benefits to human perceptions of job security and adoption. Study 2 revealed that the negative impacts of increasing responsibility can be mitigated by presenting AI teammates’ capabilities as being endorsed by coworkers and one’s own past experience. This paper discusses how to use these results to best balance the presentation of AI teammates’ capabilities and responsibilities, as well as identifying AI as teammates.

Increased use of collaborative technologies and agile teamwork models has led to a greater need for temporary teams. Unfortunately, they lack the normal team formation processes that traditional teams use. Information sharing recommender systems can be used to share information about team members amongst the team; however, these systems rely on the team members themselves to disclose valuable information. While prior research has shown that an effective way to encourage user disclosure is through explanations to the user about what benefits they will gain from disclosure, the timing of such explanations has yet to be consideblack. In a between-subjects study with 150 participants, we assessed the content and timing of explanations on levels of disclosure in temporary teams. Our results indicate that providing benefit-related explanations during the time of disclosure can increase user disclosure, and providing benefit-related explanations during the recommendation process can increase user trust in the system. These results provide important design implications for teams and the HCI community.

Measuring instructors' adoption of learning management system (LMS) tools is a critical first step in evaluating the efficacy of online teaching and learning at scale. Existing models for LMS adoption are often qualitative, learner-centered, and difficult to leverage towards institutional support. We propose depth-of-use (DOU): an intuitive measurement model for faculty's utilization of a university-wide LMS and their needs for institutional support. We hypothesis-test the relationship between DOU and course attributes like modality, participation, logistics, and outcomes. In a large-scale analysis of metadata from 30000+ courses offered at Virginia Tech over two years, we find that a pervasive need for scale, interoperability and ubiquitous access drives LMS adoption by university instructors. We then demonstrate how DOU can help faculty members identify the opportunity-cost of transition from legacy apps to LMS tools. We also describe how DOU can help instructional designers and IT organizational leadership evaluate the impact of their support allocation, faculty development and LMS evangelism initiatives.

Money is a sensitive and complex component of everyday life, which can significantly affect people's relationships with each other. Recently, emerging digital peer-to-peer (P2P) payment applications continue to complicate how people deal with money with individuals they know in comparison to traditional payment methods such as physical money (i.e., cash). Through an online survey study (N=218), in this paper, we investigate significant differences between using digital P2P payments and traditional payment methods (i.e., physical money) in terms of perceptions of transactions by using a given payment method, perceptions of people's existing interpersonal relationships with their known contacts after such transactions, and future use intention of a given payment method. We thus offer empirical evidence to highlight multidimensional influences of using digital P2P payments on established offline interpersonal relationships compared to using traditional payment methods (i.e., physical money). We also provide crucial new insights to unpack digital P2P payments as a novel technology-mediated interpersonal dynamic in people's everyday lives, which goes beyond merely as a business payment model or a financial service. We additionally provide important directions for designing more supportive and socially satisfactory digital P2P payment platforms in the future by taking the interplay of financial exchanges and interpersonal relationships into consideration.

Newly-formed teams often encounter the challenge of members coming together to collaborate on a project without prior knowledge of each other's working and communication styles. This lack of familiarity can lead to conflicts and misunderstandings, hindering effective teamwork. Derived from research in social recommender systems, team recommender systems have shown the ability to address this challenge by providing personality- derived recommendations that help individuals interact with teammates with differing personalities. However, such an approach raises privacy concerns as to whether teammates would be willing to disclose such personal information with their team. Using a vignette survey conducted via a research platform that hosts a team recommender system, this study found that context and individual differences significantly impact disclosure preferences related to team recommender systems. Specifically, when working in interdependent teams where success required collective performance, participants were more likely to disclose personality information related to Emotionality and Extraversion unconditionally. Drawing on these findings, this study created and evaluated a machine learning model to predict disclosure preferences based on group context and individual differences, which can help tailor privacy considerations in team recommender systems prior to interaction.

The popularity of social virtual reality (VR) platforms such as VRChat has led to growing concerns about new and more severe forms of online harassment targeting one's identity characteristics (i.e., identity-based harassment ). Social VR users with marginalized identities (e.g., women, LGBTQIA+ individuals, and racial/ethnic minorities) have been reported as particularly vulnerable to such harassment. This is mainly because social VR can make one's offline identity known to others (i.e., what we term identity revelation in this work) through a unique combination of avatar design, voice use, and immersive full- or partial-body tracking. To address these safety concerns, there is an urgent need to unpack the complex dynamics surrounding how one's offline identity is (mis)perceived by others, and how these (mis)perceptions may affect identity-based harassment in social VR. This study thus utilizes a large-scale survey with 223 social VR users across six continents/regions of the world with varying social VR experiences and identities to investigate (1) the relationship between identity-based harassment in social VR and (mis)perceptions of selective identity revelation practices, (2) how embodying one's identity in social VR might actually be less risky than once thought, and (3) how who you are does still matter when it comes to identity-based harassment in social VR. It also highlights the need to better account for understudied aspects of identity-based harassment in social VR and to better educate social VR users on the interplay between harassment and (mis)perceived identity revelation in these spaces.

This paper presents a comparative legislative analysis that critically examines the precursors and implementation of online safety legislation on societies in the Global South, using the Caribbean as a focal case study. The online communities that we engage with foster greater human connections, serve as platforms to promote business services, and have become central in public discourse worldwide. Consequently, governance has emerged as a vital approach to maintain healthy digital environments and address the rise of socio-technical harms flowing from online communities. This study employs a two-tiered analytical approach to scrutinize the legislative frameworks established in the Caribbean and to surface variations in policy priorities, approaches to regulation, levels of enforcement, and identify power dynamics that impact the execution of said regulation. The findings of this work shed light on the need for additional work in social computing that delves into the societal ramifications of existing governance structures, rather than solely focusing on efforts to inform and develop new legislation. Through the lens of the Caribbean's case study, this paper provides valuable insights for policymakers and stakeholders in the region seeking to optimize the impact of online safety laws on their societies while highlighting contextually appropriate governance solutions that acknowledge the unique cultural, political, and socioeconomic contexts of the Global South.

As AI becomes more advanced in natural language processing, the research on AI's communication in teaming environments is getting more attention in CSCW/HCI. Even though AI's communication serves as an essential part of facilitating team coordination and shaping team outcomes, the impact of AI's communication modality on human-AI teamwork is still understudied. Using a mixed-design experiment and follow-up group interviews with 100 participants, we investigate the impact of AI's communication modality, one of AI's most essential communication characteristics, on team coordination and team outcomes in two different human-AI team compositions, human-human-AI teams, and human-AI-AI teams. Our findings highlight the trade-offs between AI's verbal communication and visual communication, which inspire two design recommendations on how to apply AI's verbal and visual communication to support human-AI coordination effectively. Our study generates an initial understanding of the role that AI's communication will have in ensuring team effectiveness in human-AI teams in the CSCW/HCI field.

Although older adults are increasingly adopting digital social technologies, a lack of knowledge and experience makes them vulnerable to digital privacy and security threats. It is, therefore, crucial to build digital privacy education interventions that empower older adults to take more control over their digital privacy. Most tutorials and support materials are designed for the younger generations and are not necessarily as effective for the older population. In this paper, we explore the development of education interventions suited to the learning styles of the older adult population. We particularly develop interventions that span a variety of modalities (text, videos, audio presentations, infographics, comics, interactive tutorials, and chatbots) and evaluate these interventions in a focus group study, gathering feedback from both older and younger adults regarding the education interventions and how to improve them. Our findings demonstrate that there are distinct differences in modality preferences between older and younger adults. In this paper, we discuss our findings and contribute to the development of digital privacy education interventions that are tailored to the specific needs and preferences of older adults.

Explanation of artificial intelligence (AI) decision-making has become an important research area in human–computer interaction (HCI) and computer-supported teamwork research. While plenty of research has investigated AI explanations with an intent to improve AI transparency and human trust in AI, how AI explanations function in teaming environments remains unclear. Given that a major benefit of AI giving explanations is to increase human trust understanding how AI explanations impact human trust is crucial to effective human-AI teamwork. An online experiment was conducted with 156 participants to explore this question by examining how a teammate’s explanations impact the perceived trust of the teammate and the effectiveness of the team and how these impacts vary based on whether the teammate is a human or an AI. This study shows that explanations facilitate trust in AI teammates when explaining why AI disobeyed humans’ orders but hindered trust when explaining why an AI lied to humans. In addition, participants’ personal characteristics (e.g., their gender and the individual’s ethical framework) impacted their perceptions of AI teammates both directly and indirectly in different scenarios. Our study contributes to interactive intelligent systems and HCI by shedding light on how an AI teammate’s actions and corresponding explanations are perceived by humans while identifying factors that impact trust and perceived effectiveness. This work provides an initial understanding of AI explanations in human-AI teams, which can be used for future research to build upon in exploring AI explanation implementation in collaborative environments.

Abstract When deciding where to visit next while traveling in a group, people have to make a trade-off in an interactive group recommender system between (a) disclosing their personal information to explain and support their arguments about what places to visit or to avoid (e.g., this place is too expensive for my budget) and (b) protecting their privacy by not disclosing too much. Arguably, this trade-off crucially depends on who the other group members are and how cooperative one aims to be in making the decision. This paper studies how an individual’s personality, trust in group, and general privacy concern as well as their preference scenario and the task design serve as antecedents to their trade-off between disclosure benefit and privacy risk when disclosing their personal information (e.g., their current location, financial information, etc.) in a group recommendation explanation. We aim to design a model which helps us understand the relationship between risk and benefit and their moderating factors on final information disclosure in the group. To create realistic scenarios of group decision making where users can control the amount of information disclosed, we developed . This chat-bot agent generates natural language explanations to help group members explain their arguments for suggestions to the group in the tourism domain [more specifically, the initial POI options were selected from the category of “Food” in Amsterdam (see Sect. 3.2 for the details)]. To understand the dynamics between the factors mentioned above and information disclosure, we conducted an online, between-subjects user experiment that involved 278 participants who were exposed to either a competitive task (i.e., instructed to convince the group to visit or skip a recommended place) or a cooperative task (i.e., instructed to reach a decision in the group). Results show that participants’ personality and whether their preferences align with the majority affect their general privacy concern perception. This, in turn, affects their trust in the group, which affects their perception of privacy risk and disclosure benefit when disclosing personal information in the group, which ultimately influences the amount of personal information they disclose. A surprising finding was that the effect of privacy risk on information disclosure is different for different types of tasks: privacy risk significantly impacts information disclosure when the task of finding a suitable destination is framed competitively but not when it is framed cooperatively. These findings contribute to a better understanding of the moderating factors of information disclosure in group decision making and shed new light on the role of task design on information disclosure. We conclude with design recommendations for developing explanations in group decision-making systems. Further, we propose a theory of user modeling that shows what factors need to be considered when generating such group explanations automatically.

These proceedings contain the papers selected for inclusion in the technical program for the 22st ACM Workshop on Privacy in the Electronic Society (ACM WPES 2023), held in conjunction with the 30th ACM Conference on Computer and Communication Security (ACM CCS 2023) at the Tivoli Congress Center in Copenhagen, Denmark, on November 26, 2023. In response to the workshop's call for papers, 31 valid submissions were received, including 21 full paper submissions and 10 short paper submissions. They were evaluated by a technical program committee consisting of 54 researchers whose backgrounds include a diverse set of topics related to privacy. Each paper was reviewed by at least 3 members of the program committee. Papers were evaluated based on their importance, novelty, and technical quality. After the rigorous review process, 9 submissions were accepted as full papers (acceptance rate: 29.0%) and an additional 8 submissions were accepted as short papers.

Lack of diversity and high dropout rates among underrepresented students plague the CS discipline. We developed, administered, and validated survey scales measuring social factors that impact the retention and graduation of under-represented CS undergrads at two institutions. Results revealed significant differences between students who identify as men vs. women in terms of computing identity and confidence, and between black and non-black students in terms of familiarity with future opportunities.

A growing organizational trend is to utilize ad-hoc team formation which allows for teams to intentionally form based on the member skills required to accomplish a specific task. Due to the unfamiliar nature of these teams, teammates are often limited by their understanding of one another (e.g., teammate preferences, tendencies, attitudes) which limits the team's functioning and efficiency. This study conceptualizes and investigates the use of a teammate information-sharing recommender system which selectively shares interpersonal recommendations between unfamiliar teammates (e.g., "Your voice may be overshadowed by this teammate when making decisions...") to promote teammate understanding. Through a mixed-methods approach involving 105 participants working on actual unfamiliar teams, this study explores how presentation elements such as anonymity and explanations influence system perceptions and how anonymity influences team outcomes. Results indicate that anonymizing recommendations was associated with worse team measures, particularly team satisfaction and team cohesion. Qualitative results shed light on why team members perceived privacy concerns and team benefits associated with using the system. We contribute to CSCW through a better understanding of how to support unfamiliar teams, the conceptualization and empirical investigation of a novel teammate information-sharing recommender system, and foundational design recommendations associated with such a system.

The privacy calculus framework and trust heuristics has been used to understand people’s privacy decision-making processes. However, most existing studies are mainly focused on people from developed countries. In this study, we use the privacy calculus in combination with trust heuristics to analyse how people from a developing African nation make decisions. Specifically, we conduct a web-based experiment in which 232 participants from Nigeria used a financial planning prototype app to respond to a number of disclosure questions. We examined how their perceived benefit, perceived sensitivity, and trust in the app influenced their disclosure decisions. In addition, we investigated possible moderating effects of gender and used Partial Least Squares path modelling to analyze our data. Our results show that perceived sensitivity (risks) and perceived benefits influenced the decision-making process of our participants. In addition, women were more likely to change their perception of sensitivity and benefits based on trust, while men were more likely to disclose information based on their perception of benefits. We also found that women were less likely to disclose their information to the app than men. Based on our findings, we make recommendations for educators, financial institutions, designers, and policymakers that aim to raise privacy awareness and design interventions in Nigeria and Africa at large.

Modern Web browsers provide users with tools to reduce the burden of filling out forms. Despite the widespread adoption of these tools, little is known about how they affect users’ privacy decision-making. This research compares traditional form autocompletion tools with two alternative tools designed for elaboration for this study (“add” and “remove” tools). The results show that the use of traditional form autocompletion tools significantly diminishes users’ deliberate privacy decision-making, while the proposed tools can mitigate these adverse effects, such that users (1) disclose significantly less information and (2) are more likely to assess the alignment between the type of the data requested and the goal of the entity requesting that data (i.e., context specificity ). While both proposed tools help users become more deliberate in their disclosure behavior, they prefer the “add” tool over the “remove” tool. Our results show that tools designed for elaboration can nudge users toward protecting their privacy.

We conducted a user study with 380 Android users, profiling them according to two key privacy behaviors: the number of apps installed and the Dangerous permissions granted to those apps. We identified four unique privacy profiles: 1) Privacy Balancers (49.74% of participants), 2) Permission Limiters (28.68%), 3) App Limiters (14.74%), and 4) the Privacy Unconcerned (6.84%). App and Permission Limiters were significantly more concerned about perceived surveillance than Privacy Balancers and the Privacy Unconcerned. App Limiters had the lowest number of apps installed on their devices with the lowest intention of using apps and sharing information with them, compared to Permission Limiters who had the highest number of apps installed and reported higher intention to share information with apps. The four profiles reflect the differing privacy management strategies, perceptions, and intentions of Android users that go beyond the binary decision to share or withhold information via mobile apps.

Little is known about non-Western social media users’ motivations for adopting behaviors that protect them against pervasive threats to their privacy, security, and personal well-being. Drawing on Rogers’ Protection Motivation Theory (PMT), this survey study explores Caribbean people’s (N=551) perceptions of safety threats and the factors contributing to their intention to adopt protective behaviors. Our analysis revealed that prior victimization was associated with increased perceptions of vulnerability and severity of harms, which, in turn, influenced elevated safety protection behaviors. For harassment-related harms in particular, participants’ trust in social media sites increased their intention to adopt protective behaviors. We observe significant country-to-country differences, which we contextualize through interviews with experts throughout the region. Our findings provide a new understanding of users’ mental models, behaviors, and attitudes with respect to online safety. We conclude by discussing theoretical and practical implications and outline opportunities for the design of inclusive and culturally-aware safety tools.

Machine learning technologies are increasingly being applied in many different domains in the real world. As autonomous machines and black-box algorithms begin making decisions previously entrusted to humans, great academic and public interest has been spurred to provide explanations that allow users to understand the decision-making process of the machine learning model. Besides explanations, Interactive Machine Learning (IML) seeks to leverage user feedback to iterate on an ML solution to correct errors and align decisions with those of the users. Despite the rise in explainable AI (XAI) and Interactive Machine Learning (IML) research, the links between interactivity, explanations, and trust have not been comprehensively studied in the machine learning literature. Thus, in this study, we develop and evaluate an explanation-driven interactive machine learning (XIML) system with the Tic-Tac-Toe game as a use case to understand how a XIML mechanism improves users’ satisfaction with the machine learning system. We explore different modalities to support user feedback through visual or rules-based corrections. Our online user study (n = 199) supports the hypothesis that allowing interactivity within this XIML system causes participants to be more satisfied with the system, while visual explanations play a less prominent (and somewhat unexpected) role. Finally, we leverage a user-centric evaluation framework to create a comprehensive structural model to clarify how subjective system aspects, which represent participants’ perceptions of the implemented interaction and visualization mechanisms, mediate the influence of these mechanisms on the system’s user experience.

Abstract Research finds that the users of Social Networking Sites (SNSs) often fail to comprehensively engage with the plethora of available privacy features— arguably due to their sheer number and the fact that they are often hidden from sight. As different users are likely interested in engaging with different subsets of privacy features, an SNS could improve privacy management practices by adapting its interface in a way that proactively assists, guides, or prompts users to engage with the subset of privacy features they are most likely to benefit from. Whereas recent work presents algorithmic implementations of such privacy adaptation methods, this study investigates the optimal user interface mechanism to present such adaptations. In particular, we tested three proposed“adaptation methods”(automation, suggestions, highlights) in an online between-subjects user experiment in which 406 participants used a carefully controlled SNS prototype. We systematically evaluate the effect of these adaptation methods on participants’ engagement with the privacy features, their tendency to set stricter settings (protection), and their subjective evaluation of the assigned adaptation method. We find that theautomationof privacy features afforded users the most privacy protection, while giving privacysuggestionscaused the highest level of engagement with the features and the highest subjective ratings (as long as awkward suggestions are avoided). We discuss the practical implications of these findings in the effectiveness of adaptations improving user awareness of, and engagement with, privacy features on social media.

introduction Share on Special Issue on Highlights of IUI 2021: Introduction Authors: Tracy Hammond Texas A&M University, College Station, Texas, United States Texas A&M University, College Station, Texas, United StatesView Profile , Bart Knijnenburg Clemson University, Clemson, South Carolina, United States Clemson University, Clemson, South Carolina, United StatesView Profile , John O’Donovan University of California, Santa Barbara, California, United States University of California, Santa Barbara, California, United StatesView Profile , Paul Taele Texas A&M University, College Station, Texas, United States Texas A&M University, College Station, Texas, United StatesView Profile Authors Info & Claims ACM Transactions on Interactive Intelligent SystemsVolume 12Issue 403 February 2023Article No.: 25pp 1–4https://doi.org/10.1145/3561516Published:03 February 2023Publication History 0citation47DownloadsMetricsTotal Citations0Total Downloads47Last 12 Months47Last 6 weeks9 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteGet Access

To understand the underlying process of users’ information disclosure decisions, scholars often use either the privacy calculus framework or refer to heuristic shortcuts. It is unclear whether the decision process varies by age. Therefore, using these common frameworks, we conducted a web-based experiment with 94 participants, who were younger (ages 19-22) or older (65+) adults, to understand how perceived app trust, sensitivity of the data, and benefits of disclosure influence users disclosure decisions. Younger adults were more likely to change their perception of data sensitivity based on trust, while older adults were more likely to disclose information based on perceived benefits of disclosure. These results suggest older adults made more rationally calculated decisions than younger adults, who made heuristic decisions based on app trust. Our findings negate the mainstream narrative that older adults are less privacy-conscious than younger adults; instead, older adults weigh the benefits and risks of information disclosure.

Research has demonstrated that users’ heuristic decision-making processes cause external factors like defaults and framing to influence the outcome of privacy decisions. Proponents of “privacy nudging” have proposed leveraging these effects to guide users’ decisions. Our research shows that defaults and framing not only influence the outcome of privacy decisions, but also the process of evaluating the contextual factors associated with the decision, effectively making the decision-making process more heuristic. In our analysis of an existing dataset of scenario-based smart home privacy decisions, we demonstrate that defaults and framing not only have a direct effect on participants’ decisions; they also moderate the effect of their cognitive appraisals of the presented scenarios on the decision. These results suggest that nudges like defaults and framing exacerbate the well-researched problem that people often employ heuristics rather than making deliberate privacy decisions, and that privacy-setting interfaces should avoid the effects of heuristic decision-making.

While virtual character embodiment has been studied as a mitigator of singular societal biases in fully immersive VR and empathy games, there have been no major studies on representation featuring standard game play or intersectional identities. In our study, participants played a short 2D video game with racial and gender character manipulations. They then rated a LinkedIn profile application to examine interactions of racial and gender biases. White male participants showed bias against black and female applicants, with the black female applicant experiencing both racial and gender bias. However, participants who embodied certain underrepresented characters in the game displayed reduced biases. Participants’ perceived identification with the characters moderated this effect. The study highlights a lack of homogeneity in the prevalence and potential reduction of different societal biases and incorporates intersectionality to illustrate how multiple parts of a player character’s identity can be used to combat biases.

To account for privacy perceptions and preferences in user models and develop personalized privacy systems, we need to understand how users make privacy decisions in various contexts. Existing studies of privacy perceptions and behavior focus on overall tendencies toward privacy, but few have examined the context-specific factors in privacy decision making. We conducted a survey on Mechanical Turk (N=401) based on the theory of planned behavior (TPB) to measure the way users’ perceptions of privacy factors and intent to disclose information are affected by three situational factors embodied hypothetical scenarios: information type, recipients’ role, and trust source. Results showed a positive relationship between subjective norms and perceived behavioral control, and between each of these and situational privacy attitude; all three constructs are significantly positively associated with intent to disclose. These findings also suggest that, situational factors predict participants’ privacy decisions through their influence on the TPB constructs.

Recommender systems are used with increasing frequency in a wide variety of domains ranging from e- commerce to tourism, healthcare and online learning. However, the interaction with these systems generally tends to be limited to shallow feedback, such as providing ratings or filtering. Allowing users to interact with the recommender systems in a conversational environment brings opportunities in which the preferences can effectively be elicited from the users while the users can feel more in control of the whole process. However, when the existing non-interactive recommender systems are considered, it may not be easy to build an interactive layer directly on top of them. This is because there is already a great deal of modelling and work invested in the underlying algorithm and the system itself. Enabling interaction could mean rebuilding the whole solution from scratch, as the current design may not be able to consume preferences and information learnt from the user interaction online. In this paper, we propose the Interactive Recommender Framework, which converts non-interactive recommender solutions to conversational recommenders. We demonstrate how Interactive Recommender Framework can successfully enable interactivity on top of non-interactive recommender systems by integrating it into two different recommender algorithms from literature, and validate our solution through offline simulation experiments and online user studies.

This paper addresses inconsistencies that exist in the measurement instruments HCI researchers use in cross-cultural studies. We study some commonly used measurement instruments that capture cultural dimensions at an individual level and conduct "measurement invariance tests," which test whether the questions comprising a construct have similar characteristics across different groups (e.g., countries). We find that these cultural dimensions are, to some extent, non-invariant, making statistical comparisons between countries problematic. Furthermore, we study the (non)invariance of the causal relationship between these cultural dimensions and privacy-related constructs, e.g., privacy concern and the amount of information users share on social media. Our results suggest that in several instances, these cultural dimensions have a different effect on privacy-related constructs per country. This severely reduces their usefulness for developing cross-cultural arguments in cross-country studies. We discuss the value of conducting measurement and causal non-invariance tests and urge scholars to develop more robust means of measuring culture.

Privacy has been conceptualized as a multi-dimensional construct in prior research. However, most multi-dimensional conceptualizations were developed based on populations from western countries. It remains an open question whether the underlying dimensions of privacy stays consistent in non-western countries. Through a series of factor analyses on two survey datasets, we compare the dimensions of privacy concern, information disclosure, general disclosiveness, and privacy management strategies among social network users in the US, China and South Korea. We find significant cross-country differences in the dimensions of these privacy-related concepts, indicating that the fundamental understanding of these concepts varies substantially across these countries. We discuss possible explanations of these cross-country differences and make methodological suggestions for future work.

The COVID-19 pandemic lockdown lead to the rapid adoption and use of various groupware applications ("apps'') for remote connection with colleagues, friends, and family. Different factors such as user experiences, trust, and social influences ("user-situational motivations'') were instrumental in determining how and what apps people adopted and used, especially at the onset of the COVID-19 pandemic. In this empirical study, we examine how these factors and four predominant user-situational motivations (i.e., the mandated use of an app by an employer/institution, recommended use of an app by an employer/institution, recommended use of an app by a peer(s), and self-selection of an app) influenced the rapid adoption and use of groupware applications. Specifically, we develop an "emergency adoption model" of groupware applications using 195 valid survey responses to highlight the factors that motivated these apps' use at the onset of COVID-19 pandemic lockdown. We leverage the Technology Adoption Model (TAM) and integrate it with the users' past use of the application before the COVID-19 lockdown, user-situational motivation, and their privacy-related trust in the application provider to develop a more comprehensive model. Using confirmatory factor analysis (CFA) and structural equation modeling (SEM), we find that the users who used a groupware app in the past continued to use it, and in line with TAM, users' intention to adopt and use a groupware application was largely driven by the ease-of-use and usefulness of the app. Furthermore, while not a part of the traditional TAM model, we find that privacy-related trust in the application provider plays an important role in emergency adoption. However, unlike typical adoption models, the nature of all these effects---most prominently those related to privacy-related trust---depend on the underlying situational motivation. We discuss the implications of these findings and suggest ways to improve the adoption and use of groupware applications, especially during crises like the COVID-19 pandemic.

Chatbots or conversational recommenders have gained increasing popularity as a new paradigm for Recommender Systems (RS). Prior work on RS showed that providing explanations can improve transparency and trust, which are critical for the adoption of RS. Their interactive and engaging nature makes conversational recommenders a natural platform to not only provide recommendations but also justify the recommendations through explanations. The recent surge of interest inexplainable AI enables diverse styles of justification, and also invites questions on how styles of justification impact user perception. In this article, we explore the effect of “why” justifications and “why not” justifications on users’ perceptions of explainability and trust. We developed and tested a movie-recommendation chatbot that provides users with different types of justifications for the recommended items. Our online experiment ( n = 310) demonstrates that the “why” justifications (but not the “why not” justifications) have a significant impact on users’ perception of the conversational recommender. Particularly, “why” justifications increase users’ perception of system transparency, which impacts perceived control, trusting beliefs and in turn influences users’ willingness to depend on the system’s advice. Finally, we discuss the design implications for decision-assisting chatbots.

Intelligent virtual agents increasingly use machine learning to model the needs and preferences of their users in order to give personalized support. In this paper we present a vision of "human-centric modeling" that moves this practice beyond the automated estimation of users' current preferences based on observable behaviors to a modeling process that a) considers users' hidden characteristics rather than observable behaviors, b) models possible futures rather than users' current state, and c) gives users an opportunity to interactively control this modeling process. This vision focuses on the long-term needs of individual humans and a genuine betterment of society.

The commercialization of Virtual Reality (VR) devices is making the technology increasingly accessible to users around the world. Despite the success that VR is starting to see with its growing popularity, it has yet to become widely adopted and achieve its ultimate goal- convincingly simulate real life like experiences. The inability to generate adequate levels of presence and to prevent the manifestation of cybersickness are the two prominent barriers that have hindered VR from achieving its ultimate goal. While traditional research has examined factors that influence (correlate with) the onset and severity cybersickness, there is still a gap in our knowledge about the consequences of having motion control on cybersickness in immersive virtual environments (IVE’s) achieved using tracked Head Mounted Displays (HMD’s). Furthermore, outside of a correlational capacity, it is still unclear as to what causes cybersickness to affect presence in immersive virtual environments. The success of immersive virtual reality as a technology will hence largely come down to our ability to understand the interrelationship between these variables and then address the challenges they pose. Towards this end, we investigated how the affordance of motion control affects cybersickness and presence in an HMD based VR driving simulation by conducting a between subjects study where we manipulated the affordance of control between three experimental conditions. We leverage structural equation modeling in an attempt to build a framework that explains the relationship between virtual motion control, workload, cybersickness, time spent in the simulation, perceived time and presence. Our structural model helps explain why motion control could be an important factor to consider in addressing VR’s challenges and realizing its ultimate aim to simulate reality.

Previous studies have observed an intention-behavior gap that has been labeled the “privacy paradox”: people disclose personal information (behavior) despite expressing negative sharing intentions (in surveys). However, this phenomenon has not been studied in the Internet of Things (IoT) in which users’ personal information sharing is crucial for the functionality of the technology. We explore this phenomenon by comparing participants’ intentions (via a survey) with their actual behavior (via a privacy-setting interface) and controlling the data sharing device and storage. Furthermore, we explore the decision processes underlying these privacy decisions by measuring and manipulating these processes using an aspect listing task. We find a reversed intention-behavior gap in IoT: participants disclosed less (rather than more) information in the behavior condition than in the intention condition, an effect that was associated with fewer benefits than risk aspects listed in the behavior condition. The number and type of aspects listed fully mediated the effect of decision type (intention versus behavior) on the decision, which suggests that a risk-benefit calculation guided the privacy decision-making. Moreover, this reversed intention-behavior gap vanishes if we specifically ask participants to think about positive and negative aspects of the decision, as this allows them to consider both risks and benefits, irrespective of decision type.

Abstract Virtual Private Networks (VPNs) can help people protect their privacy. Despite this, VPNs are not widely used among the public. In this survey study about the adoption and usage of VPNs, we investigate people’s motivation to use VPNs and the barriers they encounter in adopting them. Using data from 90 technologically savvy participants, we find that while nearly all (98%; 88) of the participants have knowledge about what VPNs are, less than half (42%; 37) have ever used VPNs primarily as a privacy-enhancing technology. Of these, 18% (7) abandoned using VPNs while 81% (30) continue to use them to protect their privacy online. In a qualitative analysis of survey responses, we find that people who adopt and continue to use VPNs for privacy purposes are primarily motivated by emotional considerations, including the strong desire to protect their privacy online, wide fear of surveillance and data tracking not only from Internet service providers (ISPs) but also governments and Internet corporations such as Facebook and Google. In contrast, people who are mainly motivated by practical considerations are more likely to abandon VPNs, especially once their practical need no longer exists. These people cite their access to alternative technologies and the effort required to use a VPN as reasons for abandonment. We discuss implications of these findings and provide suggestions on how to maximize adoption of privacy-enhancing technologies such as VPNs, focusing on how to align them with people’s interests and privacy risk evaluation.

Abstract Smartphone users are often unaware of mobile applications’ (“apps”) third-party data collection and sharing practices, which put them at higher risk of privacy breaches. One way to raise awareness of these practices is by providing unobtrusive but pervasive visualizations that can be presented in a glanceable manner. In this paper, we applied Wogalter et al.’s Communication-Human Information Processing model (C-HIP) to design and prototype eight different visualizations that depict smartphone apps’ data sharing activities. We varied the granularity and type (i.e., data-centric or app-centric) of information shown to users and used the screensaver/lock screen as a design probe. Through interview-based design probes with Android users (n=15), we investigated the aspects of the data exposure visualizations that influenced users’ comprehension and privacy awareness. Our results shed light on how users’ perceptions of privacy boundaries influence their preference regarding the information structure of these visualizations, and the tensions that exist in these visualizations between glanceability and granularity. We discuss how a pervasive, soft paternalistic approach to privacy-related visualization may raise awareness by enhancing the transparency of information flow, thereby, unobtrusively increasing users’ understanding of data sharing practices of mobile apps. We also discuss implications for privacy research and glanceable security.

In this article, we extend and improve upon a previously developed data-driven approach to design privacy-setting interfaces for users of household IoT devices. The essence of this approach is to gather users’ feedback on household IoT scenarios before developing the interface, which allows us to create a navigational structure that preemptively maximizes users’ efficiency in expressing their privacy preferences, and develop a series of ‘privacy profiles’ that allow users to express a complex set of privacy preferences with the single click of a button. We expand upon the existing approach by proposing a more sophisticated translation of statistical results into interface design, and by extensively discussing and analyzing the tradeoff between user-model parsimony and accuracy in developing privacy profiles and default settings.

We examined the effect of rendering style and the interplay between attention and emotion in users during interaction with a virtual patient in a medical training simulator. The virtual simulation was rendered representing a sample from the photo-realistic to the non-photorealistic continuum, namely Near-Realistic, Cartoon or Pencil-Shader. In a mixed design study, we collected 45 participants’ emotional responses and gaze behavior using surveys and an eye tracker while interacting with a virtual patient who was medically deteriorating over time. We used a cross-lagged panel analysis of attention and emotion to understand their reciprocal relationship over time. We also performed a mediation analysis to compare the extent to which the virtual agent’s appearance and his affective behavior impacted users’ emotional and attentional responses. Results showed the interplay between participants’ visual attention and emotion over time and also showed that attention was a stronger variable than emotion during the interaction with the virtual human.

Social media privacy features can act as a mechanism for regulating interpersonal relationships, but why do some people not use these features? Through an interview study of 56 social media users, we found two high-level perspectives towards social media and privacy that affected attitudes towards and usage of privacy features. Some users took a pragmatic approach to using social media and felt comfortable using various privacy features as a tool to manage their social relationships (e.g., avoiding bothersome posts, not feeling compelled to interact). However, there were also users who viewed taking such privacy actions as a relational hindrance and were concerned how using certain features to meet their own needs would harm their relationships with others. Through a subsequent survey (N=320), we reveal how these two perspectives impact user behavior across four social media platforms (Facebook, Instagram, LinkedIn, Twitter). Users who viewed social media as a pragmatic tool indeed used privacy features more. On the other hand, users who focused on how privacy can serve as a relational hindrance avoided using these features and, instead, prioritized social engagement and took a more indirect approach to protecting their privacy. Furthermore, the results show how these perspectives vary by individual rather than by privacy feature. These findings demonstrate the need to consider different perspectives towards social media and privacy when trying to understand and design for user behavior.

User testing is often used to inform the development of user interfaces (UIs). But what if an interface needs to be developed for a system that does not yet exist? In that case, existing datasets can provide valuable input for UI development. We apply a data-driven approach to the development of a privacy-setting interface for Internet-of-Things (IoT) devices. Applying machine learning techniques to an existing dataset of users' sharing preferences in IoT scenarios, we develop a set of "smart" default profiles. Our resulting interface asks users to choose among these profiles, which capture their preferences with an accuracy of 82%---a 14% improvement over a naive default setting and a 12% improvement over a single smart default setting for all users.

User system trust is critical to the uptake of recommendations, and several factors of trust have been identified and compared. In this paper we present a cross-cultural, crowdsourced study examining user perceptions of nine factors of trust and link the observed differences to trust development processes and cultural dimensions. While some factors consistently instil trust, others are preferred only in certain countries. Our findings and the discovered links are important for design of trusted recommender systems.

If one wants to study privacy from an intercultural perspective, one must first validate whether there are any cultural variations in the concept of “privacy” itself. This study systematically examines cultural differences in collective privacy management strategies, and highlights methodological precautions that must be taken in quantitative intercultural privacy research. Using survey data of 498 Facebook users from the US, Singapore, and South Korea, we test the validity and cultural invariance of the measurement model and predictive model associated with collective privacy management. The results show that the measurement model is only partially culturally invariant, indicating that social media users in different countries interpret the same instruments in different ways. Also, cross-national comparisons of the structural model show that causal pathways from collective privacy management strategies to privacy-related outcomes vary significantly across countries. The findings suggest significant cultural variations in privacy management practices, both with regard to the conceptualization of its theoretical constructs, and with respect to causal pathways.

It is important for recommender system designers and service providers to learn about ways to generate accurate recommendations while at the same time respecting the privacy of their users. In this tutorial, we analyze common privacy risks imposed by recommender systems, survey privacy-enhanced recommendation techniques, and discuss implications for users.

Current collaborative photo privacy protection solutions can be categorized into two approaches: controlling the recipient, which restricts certain viewers' access to the photo, and controlling the content, which protects all or part of the photo from being viewed. Focusing on the latter approach, we introduce privacy-enhancing obfuscations for photos and conduct an online experiment with 271 participants to evaluate their effectiveness against human recognition and how they affect the viewing experience. Results indicate the two most common obfuscations, blurring and pixelating, are ineffective. On the other hand, inpainting, which removes an object or person entirely, and avatar, which replaces content with a graphical representation are effective. From a viewer experience perspective, blurring, pixelating, inpainting, and avatar are preferable. Based on these results, we suggest inpainting and avatar may be useful as privacy-enhancing technologies for photos, because they are both effective at increasing privacy for elements of a photo and provide a good viewer experience.

Abstract The influence of cultural background on people’s privacy decisions is widely recognized. However, a cross-cultural approach to predicting privacy decisions is still lacking. Our paper presents a first integrated cross-cultural privacy prediction model that merges cultural, demographic, attitudinal and contextual prediction. The model applies supervised machine learning to users’ decisions on the collection of their personal data, collected from a large-scale quantitative study in eight different countries. We find that adding culture-related predictors (i.e. country of residence, language, Hofstede’s cultural dimensions) to demographic, attitudinal and contextual predictors in the model can improve the prediction accuracy. Hofstede’s variables - particularly individualism and indulgence - outperform country and language. We further apply generalized linear mixed-effect regression to explore possible interactions between culture and other predictors. We find indeed that the impact of contextual and attitudinal predictors varies between different cultures. The implications of such models in developing privacy-enabling technologies are discussed.

Every day, we are confronted with an abundance of decisions that require us to choose from a seemingly endless number of choice options. Recommender systems are supposed to help us deal with this formidable task, but some scholars claim that these systems instead put us inside a "Filter Bubble" that severely limits our perspectives. This paper presents a new direction for recommender systems research with the main goal of supporting users in developing, exploring, and understanding their unique personal preferences.

Many computer users today value personalization but perceive it in conflict with their desire for privacy. They therefore tend not to disclose data that would be useful for personalization. We investigate how characteristics of the personalization provider influence users' attitudes towards personalization and their resulting disclosure behavior. We propose an integrative model that links these characteristics via privacy attitudes to actual disclosure behavior. Using the E laboration L ikelihood M odel, we discuss in what way the influence of the manipulated provider characteristics is different for users engaging in different levels of elaboration (represented by the user characteristics of privacy concerns and self‐efficacy). We find particularly that (a) reputation management is effective when users predominantly use the peripheral route (i.e., a low level of elaboration), but much less so when they predominantly use the central route (i.e., a high level of elaboration); (b) client‐side personalization has a positive impact when users use either route; and (c) personalization in the cloud does not work well in either route. Managers and designers can use our results to instill more favorable privacy attitudes and increase disclosure, using different techniques that depend on each user's levels of privacy concerns and privacy self‐efficacy.

We investigate the usability of humanlike agent-based interfaces for interactive advice-giving systems. In an experiment with a travel advisory system, we manipulate the “humanlikeness” of the agent interface. We demonstrate that users of the more humanlike agents try to exploit capabilities that were not signaled by the system. This severely reduces the usability of systems that look human but lack humanlikehumanlike capabilities (overestimation effect). We explain this effect by showing that users of humanlike agents form anthropomorphic beliefs (a user's “mental model”) about the system: They act humanlike towards the system and try to exploit typical humanlike capabilities they believe the system possesses. Furthermore, we demonstrate that the mental model users form of an agent-based system is inherently integrated (as opposed to the compositional mental model they form of conventional interfaces): Cues provided by the system do not instill user responses in a one-to-one matter but are instead integrated into a single mental model.

People like variety and often prefer to choose from large item sets. However, large sets can cause a phenomenon called “choice overload”: they are more difficult to choose from, and as a result decision makers are less satisfied with their choices. It has been argued that choice overload occurs because large sets contain more similar items. To overcome this effect, the present paper proposes that increasing the diversity of item sets might make them more attractive and satisfactory, without making them much more difficult to choose from. To this purpose, by using structural equation model methodology, we study diversification based on the latent features of a matrix factorization recommender model. Study 1 diversifies a set of recommended items while controlling for the overall quality of the set, and tests it in two online user experiments with a movie recommender system. Study 1a tests the effectiveness of the latent feature diversification, and shows that diversification increases the perceived diversity and attractiveness of the item set, while at the same time reducing the perceived difficulty of choosing from the set. Study 1b subsequently shows that diversification can increase users’ satisfaction with the chosen option, especially when they are choosing from small, diverse item sets. Study 2 extends these results by testing our diversification algorithm against traditional Top-N recommendations, and finds that diverse, small item sets are just as satisfying and less effortful to choose from than Top-N recommendations. Our results suggest that, at least for the movie domain, diverse small sets may be the best thing one could offer a user of a recommender system.

Social Network Sites (SNS) are often characterized as a trade-off where users must give up privacy to gain social benefits. We investigated the alternative viewpoint that users gain the most benefits when SNSs give them the privacy they desire. Applying structural equation modeling to questionnaire data of 303 Facebook users, we examined the complex relationship between privacy and SNS benefits. We found that SNS users whose privacy desires were met reported higher levels of social connectedness (i.e., perceived relational closeness with others) than those who achieved less privacy than they desired. Social connectedness, in turn, played a pivotal role in building social capital (i.e., the benefits derived from relationships with others). These findings suggest that more openness may not always be better; SNSs should aim to achieve 'Privacy Fit' with user needs to enhance user experience and ensure sustained use.

Many users welcome personalized services, but are reluctant to provide the information about themselves that personalization requires. Performing personalization exclusively at the client side (e.g., on one's smartphone) may conceptually increase privacy, because no data is sent to a remote provider. But does client-side personalization (CSP) also increase users' perception of privacy?

Location-based systems are becoming more popular with the explosive growth in popularity of smart phones. However, the user adoption of these systems is hindered by growing user concerns about privacy. To design better location-based systems that attract more user adoption and protect users from information under/overexposure, it is highly desirable to understand users' location sharing and privacy preferences. This paper makes two main contributions. First, by studying users' location sharing privacy preferences with three groups of people (i.e., Family, Friend and Colleague) in different contexts, including check-in time, companion and emotion, we reveal that location sharing behaviors are highly dynamic, context-aware, audience-aware and personal. In particular, we find that emotion and companion are good contextual predictors of privacy preferences. Moreover, we find that there are strong similarities or correlations among contexts and groups. Our second contribution is to show, in light of the user study, that despite the dynamic and context-dependent nature of location sharing, it is still possible to predict a user's in-situ sharing preference in various contexts. More specifically, we explore whether it is possible to give users a personalized recommendation of the sharing setting they are most likely to prefer, based on context similarity, group correlation and collective check-in preference. PPRec, the proposed recommendation algorithm that incorporates the above three elements, delivers personalized recommendations that could be helpful to reduce both user's burden and privacy risk. It also provides additional insights into the relative usefulness of different personal and contextual factors in predicting users' sharing behavior.

We examine the effect of coarse-grained vs. fine-grained location sharing options on users' disclosure decisions when configuring a sharing profile in a location-sharing service. Our results from an online user experiment (N=291) indicate that users who would otherwise select one of the finer-grained options will employ a compensatory decision strategy when this option is removed. This means that they switch either in the direction of more privacy and less benefit, or less privacy and more benefit, depending on the subjective distance between the omitted option and the remaining options. This explanation of users' disclosure behavior is in line with fundamental decision theories, as well as the well-established notion of "privacy calculus". Two alternative hypotheses that we tested were not supported by our experimental data.

Prior research shows that a root cause of many privacy concerns in location-sharing social media is people's desire to preserve offline relationship boundaries. Other literature recognizes lying as an everyday phenomenon that preserves such relationship boundaries by facilitating smooth social interactions. Combining these strands of research, one might hypothesize that people with a predisposition to lie would generally have lower privacy concerns since lying is a means to preserve relationship boundaries. We tested this hypothesis using structural equation modeling on data from a survey administered nationwide (N=1532), and found that for location-sharing, people with a high propensity to lie actually have increased boundary preservation concerns as well as increased privacy concerns. We explain these findings using results from semi-structured interviews.

In a mixed-methods study on adoption of location-sharing social networks (LSSN), we discovered that variations in adoption and usage behavior could be explained by one's predisposition to communicate in a certain style. Specifically, we found that certain individuals prefer a communication style we call FYI (For Your Information). FYI communicators like to infer availability and to keep in touch with others without having to interact with them, which is the predominant style in current LSSN. Using structural equation modeling on a U.S. nationwide survey (N=1021), we show how the FYI communication style predicts the adoption of LSSN while also showing a negative effect on one's desire to call someone on the phone. Moreover, we find that as age increases, FYI preference significantly decreases. In a follow-on survey (N=180), we refine the FYI construct and show that it affects users' level of disclosure and participation in social media. Furthermore, we show that it completely mediates the effect of certain Big-5 personality traits on social media participation and LSSN usage. The results suggest that to cater to a wider segment of the population, LSSN (and arguably any social media) should support an active communication style.

Personalization relies on personal data about each individual user. Users are quite often reluctant though to disclose information about themselves and to be "tracked" by a system. We investigated whether different types of rationales (justifications) for disclosure that have been suggested in the privacy literature would increase users' willingness to divulge demographic and contextual information about themselves, and would raise their satisfaction with the system. We also looked at the effect of the order of requests, owing to findings from the literature. Our experiment with a mockup of a mobile app recommender shows that there is no single strategy that is optimal for everyone. Heuristics can be defined though that select for each user the most effective justification to raise disclosure or satisfaction, taking the user's gender, disclosure tendency, and the type of solicited personal information into account. We discuss the implications of these findings for research aimed at personalizing privacy strategies to each individual user.

We present techniques to characterize which data contributes most to the accuracy of a recommendation algorithm. Our main technique is called differential data analysis. The name is inspired by other sorts of differential analysis, such as differential power analysis and differential cryptanalysis, where insight comes through analysis of slightly differing inputs. In differential data analysis we chunk the data and compare results in the presence or absence of each chunk. We apply differential data analysis to two datasets and three different attributes. The first attribute is called user hardship. This is a novel attribute, particularly relevant to location datasets, that indicates how burdensome a data point was to achieve. The second and third attributes are more standard: timestamp and user rating. For user rating, we confirm previous work concerning the increased importance to the recommender of high and low user ratings.

The privacy of location-based services has gained attention with their increased popularity. To date, citing insufficient privacy demand and inefficient/immature privacy preserving technologies, service providers have not been willing to build private-enhanced systems in which they do not have access to users' location information. However, current practice is likely to change in coming years with increasing privacy awareness and technological advances. For instance, Narayanan et al. recently introduced a fast private equality testing protocol for proximity testing with an untrusted server. In the current work, based on basic notions of geometry and linear algebra, we describe a new three-party protocol for solving the same problem. Our proposed protocol decreases the number of encryptions needed and gives a more efficient solution for private equivalence testing.

Recommender systems increasingly use contextual and demographical data as a basis for recommendations. Users, however, often feel uncomfortable providing such information. In a privacy-minded design of recommenders, users are free to decide for themselves what data they want to disclose about themselves. But this decision is often complex and burdensome, because the consequences of disclosing personal information are uncertain or even unknown. Although a number of researchers have tried to analyze and facilitate such information disclosure decisions, their research results are fragmented, and they often do not hold up well across studies. This article describes a unified approach to privacy decision research that describes the cognitive processes involved in users’ “privacy calculus” in terms of system-related perceptions and experiences that act as mediating factors to information disclosure. The approach is applied in an online experiment with 493 participants using a mock-up of a context-aware recommender system. Analyzing the results with a structural linear model, we demonstrate that personal privacy concerns and disclosure justification messages affect the perception of and experience with a system, which in turn drive information disclosure decisions. Overall, disclosure justification messages do not increase disclosure. Although they are perceived to be valuable, they decrease users’ trust and satisfaction. Another result is that manipulating the order of the requests increases the disclosure of items requested early but decreases the disclosure of items requested later.

There is an increasing consensus in the field of recommender systems that we should move beyond the offline evaluation of algorithms towards a more user-centric approach. This tutorial teaches the essential skills involved in conducting user experiments, the scientific approach to user-centric evaluation. Such experiments are essential in uncovering how and why the user experience of recommender systems comes about.

Users of social recommender systems may want to inspect and control how their social relationships influence the recommendations they receive, especially since recommendations of social recommenders are based on friends rather than anonymous "nearest neighbors". We performed an online user experiment (N=267) with a Facebook music recommender system that gives users control over the recommendations, and explains how they came about. The results show that inspectability and control indeed increase users' perceived understanding of and control over the system, their rating of the recommendation quality, and their satisfaction with the system.

Research on recommender systems typically focuses on the accuracy of prediction algorithms. Because accuracy only partially constitutes the user experience of a recommender system, this paper proposes a framework that takes a user-centric approach to recommender system evaluation. The framework links objective system aspects to objective user behavior through a series of perceptual and evaluative constructs (called subjective system aspects and experience, respectively). Furthermore, it incorporates the influence of personal and situational characteristics on the user experience. This paper reviews how current literature maps to the framework and identifies several gaps in existing work. Consequently, the framework is validated with four field trials and two controlled experiments and analyzed using Structural Equation Modeling. The results of these studies show that subjective system aspects and experience variables are invaluable in explaining why and how the user experience of recommender systems comes about . In all studies we observe that perceptions of recommendation quality and/or variety are important mediators in predicting the effects of objective system aspects on the three components of user experience: process (e.g. perceived effort, difficulty), system (e.g. perceived system effectiveness) and outcome (e.g. choice satisfaction). Furthermore, we find that these subjective aspects have strong and sometimes interesting behavioral correlates (e.g. reduced browsing indicates higher system effectiveness). They also show several tradeoffs between system aspects and personal and situational characteristics (e.g. the amount of preference feedback users provide is a tradeoff between perceived system usefulness and privacy concerns). These results, as well as the validated framework itself, provide a platform for future research on the user-centric evaluation of recommender systems.

This paper compares five different ways of interacting with an attribute-based recommender system and shows that different types of users prefer different interaction methods. In an online experiment with an energy-saving recommender system the interaction methods are compared in terms of perceived control, understandability, trust in the system, user interface satisfaction, system effectiveness and choice satisfaction. The comparison takes into account several user characteristics, namely domain knowledge, trusting propensity and persistence. The results show that most users (and particularly domain experts) are most satisfied with a hybrid recommender that combines implicit and explicit preference elicitation, but that novices and maximizers seem to benefit more from a non-personalized recommender that just displays the most popular items.

As recommender systems are increasingly deployed in the real world, they are not merely tested offline for precision and coverage, but also "online" with test users to ensure good user experience. The user evaluation of recommenders is however complex and resource-consuming. We introduce a pragmatic procedure to evaluate recommender systems for experience products with test users, within industry constraints on time and budget. Researchers and practitioners can employ our approach to gain a comprehensive understanding of the user experience with their systems.

Even though people are attracted by large, high quality recommendation sets, psychological research on choice overload shows that choosing an item from recommendation sets containing many attractive items can be a very difficult task. A web-based user experiment using a matrix factorization algorithm applied to the MovieLens dataset was used to investigate the effect of recommendation set size (5 or 20 items) and set quality (low or high) on perceived variety, recommendation set attractiveness, choice difficulty and satisfaction with the chosen item. The results show that larger sets containing only good items do not necessarily result in higher choice satisfaction compared to smaller sets, as the increased recommendation set attractiveness is counteracted by the increased difficulty of choosing from these sets. These findings were supported by behavioral measurements revealing intensified information search and increased acquisition times for these large attractive sets. Important implications of these findings for the design of recommender system user interfaces will be discussed.

No abstract available.

In a recommender system that suggests options based on user attribute weights, the method of preference elicitation (PE) employed by a recommender system can influence users' satisfaction with the system, as well as the perceived usefulness and the understandability of the system. Specifically, we hypothesize that users with different levels of domain knowledge prefer different types of PE. While domain experts reported higher satisfaction and perceived usefulness with attribute-based PE (i.e., indicating preference levels for the domain-related attributes), novices preferred case-based PE (i.e., indicating the preference for specific examples, from which attribute-preferences can then be implicitly calculated). The paper discusses the decision-theoretical principles that are believed to lead to this distinction, as well as an experiment that provides substantial evidence for the hypothesis. Consequently, we introduce the idea of adapting the method of PE to users' domain knowledge on the fly using click stream data.