The widespread use of Large Language Model (LLM)-based conversational agents (CAs), especially in high-stakes domains, raises many privacy concerns. Building ethical LLM-based CAs that respect user requires an in-depth understanding the risks concern users most. However, existing research, primarily model-centered, does not provide insight into users' perspectives. To bridge this gap, we analyzed sensitive disclosures real-world ChatGPT conversations and conducted semi-structured interviews...

Privacy is a key principle for developing ethical AI technologies, but how does including technologies in products and services change privacy risks? We constructed taxonomy of risks by analyzing 321 documented incidents. codified the unique capabilities requirements described those incidents generated new risks, exacerbated known ones, or otherwise did not meaningfully alter risk. present 12 high-level that either newly created (e.g., exposure from deepfake pornography) surveillance...

The emergence of large language models (LLMs), and their increased use in user-facing systems, has led to substantial privacy concerns. To date, research on these concerns been model-centered: exploring how LLMs lead risks like memorization, or can be used infer personal characteristics about people from content. We argue that there is a need for more focusing the human aspect issues: e.g., design paradigms affect users' disclosure behaviors, mental preferences controls, tools, artifacts...

This study examines the characteristics of mobile instant-messaging users' relationships with their social contacts and effects both relationship interruption context on four measures receptivity: Attentiveness, Responsiveness, Interruptibility, Opportuneness. Overall, overshadows as predictors all these facets receptivity; this overshadowing was most acute for Interruptibility Opportuneness, but existed factors. In addition, while Mobile Maintenance Expectation Activity Engagement were...

Social media websites thrive on user engagement by employing Attention Capture Damaging Patterns (ACDPs), e.g., infinite scroll, that prey cognitive vulnerabilities to distract users. Prior work has taxonomized these ACDPs, but we have yet measure how the presence of ACDPs impacts perceived distraction nor mechanisms suppress reduce distraction. We conducted a two-week, mixed-methods field study with 29 participants model people get distracted when browsing social websites, and might play...

AI-enabled smart-home agents that automate household routines are increasingly viable, but the design space of how and what such systems should communicate with their users remains underexplored. Through a user-enactment study, we identified various interpretations feelings toward system's confidence in its automated acts. That own mental models influenced participants wanted system to communicate, as well they would assess, diagnose, subsequently improve it. Automated acts resulted from...

Smartphone users do not deal with notifications strictly in the order they are displayed, but sometimes read them from middle, suggesting a mismatch between current systems' display and users' needs. We therefore used mixed methods to investigate 34 smartphone desired notification related it self-reported of attendance. Classifying using these two orders as dimensions, we obtained seven types notifications, which helped us only highlight distinct attributes understand implied roles well...

Researchers have long attempted to estimate instant-messaging (IM) users' attentiveness, responsiveness, and interruptibility. Yet, IM self-presentation of their receptivity, perceptions automated adjustment/revelation receptivity status (e.g., Facebook Messenger's green dot that deems a user be "active"), remain under-explored. We therefore told our 43 participants app, IMStatus, was capable automatically estimating adjusting responsive, attentive, or interruptible based on smartphone...

What has changed and what should we do about it?

The emergence of large language models (LLMs), and their increased use in user-facing systems, has led to substantial privacy concerns. To date, research on these concerns been model-centered: exploring how LLMs lead risks like memorization, or can be used infer personal characteristics about people from content. We argue that there is a need for more focusing the human aspect issues: e.g., design paradigms affect users' disclosure behaviors, mental preferences controls, tools, artifacts...

Using a mixed-methods approach over six weeks, we studied 30 smartphone users' task choices, execution and effort devoted to two commercial mobile crowdsourcing platforms in the wild. We focused on influence of activity contexts, characterized by breakpoint situations attributes. In line with their stated preferences, participants were more likely proactively perform tasks during transitions between activities than an ongoing long breaks, respectively. Their choices influenced various...

Interruptibility research is growing in computer-mediated communication (CMC). While much CMC concerns "interpersonal" communication, we have not seen a close examination of the impact who mobile interruptibility research. In this paper, propose study more closely investigating interplay between interpersonal relationship characteristics with contextual factors and their on users' receptivity to communication.

History of conversations through instant messaging (IM) contains abundant information about the communication patterns dyad, including conversation partners' mutual responsiveness to messages. We have, however, not seen many examinations using such in modeling mobile users' IM communication. In this paper, we present an in-the-wild study, which leverage participants' logs build models predicting their general responsiveness. Our based on data from 33 user achieved accuracy up 71% (AUROC)....

Premature technology, privacy, intrusiveness, power consumption, and user habits are all factors potentially contributing to the lack of social acceptance smart glasses. After investigating recent development commercial eyewear its related research, we propose a design space for ubiquitous interactions while maximising interactivity with minimal obtrusiveness. We focus on implicit explicit enabled by combination miniature sensor low-resolution display simplistic interaction modalities....

As people utilize instant messaging (IM) to communicate with of various relationships, they pay different amounts attention and have communication practices them relationships. However, we haven't seen a close investigation how users' IM patterns relate groups contacts. We collected logs 547 sender-recipient pairs from 33 smartphone users over the course 4 weeks, used k-mean clustering identify 6 clusters these patterns. illustrate characteristics distinct as well relationship between...

Despite being characterized as constantly on and connected, IM users' responsiveness varies across different contacts. While research has shown that the relationship between conversation partners plays an important role in influencing their communication patterns, characterization of such patterns simply by using information is limited [7, 8]. In this paper, we identify five distinct clusters unsupervised learning derived from 46 history. We show category sufficed to characterize three but...

Research shows that smartphone users often attend to phone notifications are in the middle of notification list. This suggests a mismatch between display order and users' attendance on notifications. Yet, we know little about how would like their be sorted presented. paper presents preliminary results mixed-methods study difference desired Our show existed nearly half cases. Specifically, many certain categories placed higher drawers than actual notification-attendance behaviors tend...

Privacy is a key principle for developing ethical AI technologies, but how does including technologies in products and services change privacy risks? We constructed taxonomy of risks by analyzing 321 documented incidents. codified the unique capabilities requirements described those incidents generated new risks, exacerbated known ones, or otherwise did not meaningfully alter risk. present 12 high-level that either newly created (e.g., exposure from deepfake pornography) surveillance...

Prior interruptibility research has focused on identifying interruptible or opportune moments for users to handle notifications. Yet, may not want attend all notifications even at these moments. Research shown that users' current practices selective attendance are through speculating about notification sources. sometimes the above information is insufficient, making speculations difficult. This paper describes first attempt examine how well a machine learning model can predict when would...

A large body of interruptibility research has attempted to minimize disruptions caused by smartphone notifications. Yet, little explored ways enable users selectively attend notifications, which can occur as early first notice the notification alert and start speculate about its source. Nevertheless, users' speculation may not be always accurate. We took step in helping make speculations notifications facilitate selective attendance. developed Notiware, an Android app that helps generating...

Instant messaging (IM) communication has been widely studied due to its prevalence in our everyday communication. Numerous factors that contribute (un)responsiveness have identified. Yet an integrated view of the influence IM responsiveness remains absent. This paper reports qualitative findings from interviews with 46 users, and identifies five main elements underlying users' response decisions: habits, need fulfillment, perceived obligation, readiness/suitability , pace/rhythm coordination...