MobiCHAI’3 is the third edition of the International Workshop on Mobile Cognition-Altering/Augmenting Technologies (CAT) Using Human-Centered AI. Following two successful events in 2024 and 2025, the focus is shifting from exploring the design space to trustworthy, mission-oriented, and domain-specific mobile and wearable AI systems for cognitive support in real-world contexts. This time, the focus is on healthcare and other high-stakes areas, including support for surgical preparation and perioperative procedures, monitoring of stress and workload, and cognitive assistance for clinicians and care teams. With special attention to trust calibration in AI-supported decision-making. Contributions from education, work, accessibility, well-being, and everyday life are welcome. The conference addresses an interdisciplinary community spanning Mobile HCI, Ubiquitous Computing, Cognitive Science, Healthcare, AI, Design Research, and the Social Sciences. Key topics include multimodal sensing of cognitive and affective states, human-centered modeling and simulation of cognition, methods of cognitive augmentation, and the ethical, social, and trust-related implications.

Anti-Social Behaviour (ASB) is a universal issue that impacts community cohesion, involving behaviours that cause harassment, alarm, or distress to community members. The utilisation of technology in ASB management and prevention is limited, either due to resource accessibility issues or this area of research being deeply unexplored. Publicly accessible information and services are inconsistent and insufficiently circulated, potentially decreasing public knowledge and reporting of ASB. To address this, this ongoing PhD research proposes digitalising ASB management techniques to improve the communication between the general public and UK government, community cohesion, and ASB reporting and prevention. In particular, the work has built a technological toolkit that addresses: ASB engager intervention, ASB public reporting and knowledge gain, and ASB practitioner training. The technologies explored include mobile HCI interactive QR-based reporting interfaces and a digital web-based ASB engager intervention prototype. Future work regards exploration of interactive interfaces for engager and practitioner training.

New Internet users in developing contexts, like India, often access mobile technologies under conditions of limited digital literacy, financial constraints, shared device usage, and smartphone adoption under budget constraints. Existing mobile systems are largely designed for technologically proficient users with personal devices, creating gaps in privacy, security, and user experience (UX) for emerging user populations. This work adopts a mixed-methods approach combining qualitative and quantitative methods to study mobile interaction in resource-constrained contexts. Qualitative studies examine how users perceive privacy, risk, trust, and smartphone usability in everyday life. Quantitative benchmarking and system-building evaluate experiential smartphone quality across audio, video, speech, camera, battery, and network dimensions, including the development of \textit{Phone Pathology}, a diagnostic system for user-facing smartphone quality assessment. The goal is to develop an inclusive mobile UX framework integrating user capabilities, contextual constraints, and device performance to support safer and more accessible mobile experiences.

Since 2010, deploying various robots in care settings has become popular among universities and care institutions worldwide. They either focus on assistive robots or companion robots. The rise of large language models and conversational AI has boosted human-robot interaction. Primarily, it raised ethical and privacy concerns among stakeholders. My work focuses on developing various robotic applications, with and without conversational AI, and on how interaction works for older adults. My findings provide some insight that, even to some degree, the robot in idle mode can provoke older adults to feel a sense of companionship and increase their well-being by being around robots, without any proper communication between the older adults and the idle robot. In my latest work, we have deployed the humanoid Avatar, tested it at the senior fair in Germany, and compared its interactions with those of an actual robot.

This research examines the ongoing password and user authentication problem from the perspective of people accessing services on Apple smartphones. First, the current state of sign-in methods is characterised on iOS apps and the developers' rationale for their implementation. This work is going to be published at MobileHCI 2026. Building on this, a second study will study how users choose to authenticate when their choices are constrained, such as by device, policy, or method availability. The final study will propose an intervention: a cybersecurity training platform focused on user authentication security. This platform employs a conversational AI interface with a learning-by-teaching pedagogical approach to provide a unique training experience. Together, these studies aim to provide insight into the design, adoption, and education of users and authentication systems in the Apple mobile ecosystem.

People often spend more time on their smartphones than they intend. Researchers have developed various tools to reduce screen time. However, tools that rely on the blocking strategy may undermine users’ sense of autonomy, whereas tools based on the self-tracking strategy often provide insufficient exposure to effectively influence behavior. I propose integrating ambient displays with the self-tracking strategy: An always-on ambient display maintains users’ awareness of their phone usage. In Study 1, I learned that problematic smartphone usage often results from a loss of time awareness. In Study 2, I tested the effectiveness of an ambient display. In Study 3, I explored the intention-behavior gap in smartphone usage, a potential tailoring variable for just-in-time intervention. In Study 4, I will test adaptive ambient displays based on the intention-behavior gap. Through this line of research, I hope to guide users toward lasting behavior change while preserving their autonomy.

Anti-Social Behaviour (ASB) is a universal issue that impacts community cohesion, involving behaviours that cause harassment, alarm, or distress to community members. The utilisation of technology in ASB management and prevention is limited, either due to resource accessibility issues or this area of research being deeply unexplored. Publicly accessible information and services are inconsistent and insufficiently circulated, potentially decreasing public knowledge and reporting of ASB. To address this, this ongoing PhD research proposes digitalising ASB management techniques to improve the communication between the general public and UK government, community cohesion, and ASB reporting and prevention. In particular, the work has built a technological toolkit that addresses: ASB engager intervention, ASB public reporting and knowledge gain, and ASB practitioner training. The technologies explored include mobile HCI interactive QR-based reporting interfaces and a digital web-based ASB engager intervention prototype. Future work regards exploration of interactive interfaces for engager and practitioner training.

New Internet users in developing contexts, like India, often access mobile technologies under conditions of limited digital literacy, financial constraints, shared device usage, and smartphone adoption under budget constraints. Existing mobile systems are largely designed for technologically proficient users with personal devices, creating gaps in privacy, security, and user experience (UX) for emerging user populations. This work adopts a mixed-methods approach combining qualitative and quantitative methods to study mobile interaction in resource-constrained contexts. Qualitative studies examine how users perceive privacy, risk, trust, and smartphone usability in everyday life. Quantitative benchmarking and system-building evaluate experiential smartphone quality across audio, video, speech, camera, battery, and network dimensions, including the development of \textit{Phone Pathology}, a diagnostic system for user-facing smartphone quality assessment. The goal is to develop an inclusive mobile UX framework integrating user capabilities, contextual constraints, and device performance to support safer and more accessible mobile experiences.

Since 2010, deploying various robots in care settings has become popular among universities and care institutions worldwide. They either focus on assistive robots or companion robots. The rise of large language models and conversational AI has boosted human-robot interaction. Primarily, it raised ethical and privacy concerns among stakeholders. My work focuses on developing various robotic applications, with and without conversational AI, and on how interaction works for older adults. My findings provide some insight that, even to some degree, the robot in idle mode can provoke older adults to feel a sense of companionship and increase their well-being by being around robots, without any proper communication between the older adults and the idle robot. In my latest work, we have deployed the humanoid Avatar, tested it at the senior fair in Germany, and compared its interactions with those of an actual robot.

This research examines the ongoing password and user authentication problem from the perspective of people accessing services on Apple smartphones. First, the current state of sign-in methods is characterised on iOS apps and the developers' rationale for their implementation. This work is going to be published at MobileHCI 2026. Building on this, a second study will study how users choose to authenticate when their choices are constrained, such as by device, policy, or method availability. The final study will propose an intervention: a cybersecurity training platform focused on user authentication security. This platform employs a conversational AI interface with a learning-by-teaching pedagogical approach to provide a unique training experience. Together, these studies aim to provide insight into the design, adoption, and education of users and authentication systems in the Apple mobile ecosystem.

People often spend more time on their smartphones than they intend. Researchers have developed various tools to reduce screen time. However, tools that rely on the blocking strategy may undermine users’ sense of autonomy, whereas tools based on the self-tracking strategy often provide insufficient exposure to effectively influence behavior. I propose integrating ambient displays with the self-tracking strategy: An always-on ambient display maintains users’ awareness of their phone usage. In Study 1, I learned that problematic smartphone usage often results from a loss of time awareness. In Study 2, I tested the effectiveness of an ambient display. In Study 3, I explored the intention-behavior gap in smartphone usage, a potential tailoring variable for just-in-time intervention. In Study 4, I will test adaptive ambient displays based on the intention-behavior gap. Through this line of research, I hope to guide users toward lasting behavior change while preserving their autonomy.

MobiCHAI’3 is the third edition of the International Workshop on Mobile Cognition-Altering/Augmenting Technologies (CAT) Using Human-Centered AI. Following two successful events in 2024 and 2025, the focus is shifting from exploring the design space to trustworthy, mission-oriented, and domain-specific mobile and wearable AI systems for cognitive support in real-world contexts. This time, the focus is on healthcare and other high-stakes areas, including support for surgical preparation and perioperative procedures, monitoring of stress and workload, and cognitive assistance for clinicians and care teams. With special attention to trust calibration in AI-supported decision-making. Contributions from education, work, accessibility, well-being, and everyday life are welcome. The conference addresses an interdisciplinary community spanning Mobile HCI, Ubiquitous Computing, Cognitive Science, Healthcare, AI, Design Research, and the Social Sciences. Key topics include multimodal sensing of cognitive and affective states, human-centered modeling and simulation of cognition, methods of cognitive augmentation, and the ethical, social, and trust-related implications.

Mobile and wearable devices increasingly collect continuous wellbeing data, including sleep, activity, heart rate, stress, blood glucose, and blood pressure. Yet access to such data does not automatically help people interpret their condition or change behavior. Many health applications remain dashboard-first, presenting charts, thresholds, goals, and alerts while leaving users to decide what a change means and what action should follow. Conversely, generic LLM-based conversational agents (CAs) can provide fluent advice, but without personal sensor grounding, they cannot detect individualized patterns or provide contextual guidance. This three-hour tutorial teaches participants how to move from passive monitoring to actionable wellbeing dialogue. Participants examine a dashboard that combines wearable health-data visualization with conversational-agent feedback, then use Wearable Sensor-Dialogue Wellbeing Agent Studio (WSDWAS) to simulate wearables, generate sensor snapshots, configure agent personas and prompt blocks, and compare dialogue styles. Grounded in Positive Computing, the tutorial emphasizes autonomy, competence, privacy, safety, and boundaries between wellbeing support and medical advice.

MobiCHAI’3 is the third edition of the International Workshop on Mobile Cognition-Altering/Augmenting Technologies (CAT) Using Human-Centered AI. Following two successful events in 2024 and 2025, the focus is shifting from exploring the design space to trustworthy, mission-oriented, and domain-specific mobile and wearable AI systems for cognitive support in real-world contexts. This time, the focus is on healthcare and other high-stakes areas, including support for surgical preparation and perioperative procedures, monitoring of stress and workload, and cognitive assistance for clinicians and care teams. With special attention to trust calibration in AI-supported decision-making. Contributions from education, work, accessibility, well-being, and everyday life are welcome. The conference addresses an interdisciplinary community spanning Mobile HCI, Ubiquitous Computing, Cognitive Science, Healthcare, AI, Design Research, and the Social Sciences. Key topics include multimodal sensing of cognitive and affective states, human-centered modeling and simulation of cognition, methods of cognitive augmentation, and the ethical, social, and trust-related implications.

Anti-Social Behaviour (ASB) is a universal issue that impacts community cohesion, involving behaviours that cause harassment, alarm, or distress to community members. The utilisation of technology in ASB management and prevention is limited, either due to resource accessibility issues or this area of research being deeply unexplored. Publicly accessible information and services are inconsistent and insufficiently circulated, potentially decreasing public knowledge and reporting of ASB. To address this, this ongoing PhD research proposes digitalising ASB management techniques to improve the communication between the general public and UK government, community cohesion, and ASB reporting and prevention. In particular, the work has built a technological toolkit that addresses: ASB engager intervention, ASB public reporting and knowledge gain, and ASB practitioner training. The technologies explored include mobile HCI interactive QR-based reporting interfaces and a digital web-based ASB engager intervention prototype. Future work regards exploration of interactive interfaces for engager and practitioner training.

New Internet users in developing contexts, like India, often access mobile technologies under conditions of limited digital literacy, financial constraints, shared device usage, and smartphone adoption under budget constraints. Existing mobile systems are largely designed for technologically proficient users with personal devices, creating gaps in privacy, security, and user experience (UX) for emerging user populations. This work adopts a mixed-methods approach combining qualitative and quantitative methods to study mobile interaction in resource-constrained contexts. Qualitative studies examine how users perceive privacy, risk, trust, and smartphone usability in everyday life. Quantitative benchmarking and system-building evaluate experiential smartphone quality across audio, video, speech, camera, battery, and network dimensions, including the development of \textit{Phone Pathology}, a diagnostic system for user-facing smartphone quality assessment. The goal is to develop an inclusive mobile UX framework integrating user capabilities, contextual constraints, and device performance to support safer and more accessible mobile experiences.

Since 2010, deploying various robots in care settings has become popular among universities and care institutions worldwide. They either focus on assistive robots or companion robots. The rise of large language models and conversational AI has boosted human-robot interaction. Primarily, it raised ethical and privacy concerns among stakeholders. My work focuses on developing various robotic applications, with and without conversational AI, and on how interaction works for older adults. My findings provide some insight that, even to some degree, the robot in idle mode can provoke older adults to feel a sense of companionship and increase their well-being by being around robots, without any proper communication between the older adults and the idle robot. In my latest work, we have deployed the humanoid Avatar, tested it at the senior fair in Germany, and compared its interactions with those of an actual robot.

This research examines the ongoing password and user authentication problem from the perspective of people accessing services on Apple smartphones. First, the current state of sign-in methods is characterised on iOS apps and the developers' rationale for their implementation. This work is going to be published at MobileHCI 2026. Building on this, a second study will study how users choose to authenticate when their choices are constrained, such as by device, policy, or method availability. The final study will propose an intervention: a cybersecurity training platform focused on user authentication security. This platform employs a conversational AI interface with a learning-by-teaching pedagogical approach to provide a unique training experience. Together, these studies aim to provide insight into the design, adoption, and education of users and authentication systems in the Apple mobile ecosystem.

People often spend more time on their smartphones than they intend. Researchers have developed various tools to reduce screen time. However, tools that rely on the blocking strategy may undermine users’ sense of autonomy, whereas tools based on the self-tracking strategy often provide insufficient exposure to effectively influence behavior. I propose integrating ambient displays with the self-tracking strategy: An always-on ambient display maintains users’ awareness of their phone usage. In Study 1, I learned that problematic smartphone usage often results from a loss of time awareness. In Study 2, I tested the effectiveness of an ambient display. In Study 3, I explored the intention-behavior gap in smartphone usage, a potential tailoring variable for just-in-time intervention. In Study 4, I will test adaptive ambient displays based on the intention-behavior gap. Through this line of research, I hope to guide users toward lasting behavior change while preserving their autonomy.

MobiCHAI’3 is the third edition of the International Workshop on Mobile Cognition-Altering/Augmenting Technologies (CAT) Using Human-Centered AI. Following two successful events in 2024 and 2025, the focus is shifting from exploring the design space to trustworthy, mission-oriented, and domain-specific mobile and wearable AI systems for cognitive support in real-world contexts. This time, the focus is on healthcare and other high-stakes areas, including support for surgical preparation and perioperative procedures, monitoring of stress and workload, and cognitive assistance for clinicians and care teams. With special attention to trust calibration in AI-supported decision-making. Contributions from education, work, accessibility, well-being, and everyday life are welcome. The conference addresses an interdisciplinary community spanning Mobile HCI, Ubiquitous Computing, Cognitive Science, Healthcare, AI, Design Research, and the Social Sciences. Key topics include multimodal sensing of cognitive and affective states, human-centered modeling and simulation of cognition, methods of cognitive augmentation, and the ethical, social, and trust-related implications.

Mobile and wearable devices increasingly collect continuous wellbeing data, including sleep, activity, heart rate, stress, blood glucose, and blood pressure. Yet access to such data does not automatically help people interpret their condition or change behavior. Many health applications remain dashboard-first, presenting charts, thresholds, goals, and alerts while leaving users to decide what a change means and what action should follow. Conversely, generic LLM-based conversational agents (CAs) can provide fluent advice, but without personal sensor grounding, they cannot detect individualized patterns or provide contextual guidance. This three-hour tutorial teaches participants how to move from passive monitoring to actionable wellbeing dialogue. Participants examine a dashboard that combines wearable health-data visualization with conversational-agent feedback, then use Wearable Sensor-Dialogue Wellbeing Agent Studio (WSDWAS) to simulate wearables, generate sensor snapshots, configure agent personas and prompt blocks, and compare dialogue styles. Grounded in Positive Computing, the tutorial emphasizes autonomy, competence, privacy, safety, and boundaries between wellbeing support and medical advice.

Anti-Social Behaviour (ASB) is a universal issue that impacts community cohesion, involving behaviours that cause harassment, alarm, or distress to community members. The utilisation of technology in ASB management and prevention is limited, either due to resource accessibility issues or this area of research being deeply unexplored. Publicly accessible information and services are inconsistent and insufficiently circulated, potentially decreasing public knowledge and reporting of ASB. To address this, this ongoing PhD research proposes digitalising ASB management techniques to improve the communication between the general public and UK government, community cohesion, and ASB reporting and prevention. In particular, the work has built a technological toolkit that addresses: ASB engager intervention, ASB public reporting and knowledge gain, and ASB practitioner training. The technologies explored include mobile HCI interactive QR-based reporting interfaces and a digital web-based ASB engager intervention prototype. Future work regards exploration of interactive interfaces for engager and practitioner training.

New Internet users in developing contexts, like India, often access mobile technologies under conditions of limited digital literacy, financial constraints, shared device usage, and smartphone adoption under budget constraints. Existing mobile systems are largely designed for technologically proficient users with personal devices, creating gaps in privacy, security, and user experience (UX) for emerging user populations. This work adopts a mixed-methods approach combining qualitative and quantitative methods to study mobile interaction in resource-constrained contexts. Qualitative studies examine how users perceive privacy, risk, trust, and smartphone usability in everyday life. Quantitative benchmarking and system-building evaluate experiential smartphone quality across audio, video, speech, camera, battery, and network dimensions, including the development of \textit{Phone Pathology}, a diagnostic system for user-facing smartphone quality assessment. The goal is to develop an inclusive mobile UX framework integrating user capabilities, contextual constraints, and device performance to support safer and more accessible mobile experiences.

Since 2010, deploying various robots in care settings has become popular among universities and care institutions worldwide. They either focus on assistive robots or companion robots. The rise of large language models and conversational AI has boosted human-robot interaction. Primarily, it raised ethical and privacy concerns among stakeholders. My work focuses on developing various robotic applications, with and without conversational AI, and on how interaction works for older adults. My findings provide some insight that, even to some degree, the robot in idle mode can provoke older adults to feel a sense of companionship and increase their well-being by being around robots, without any proper communication between the older adults and the idle robot. In my latest work, we have deployed the humanoid Avatar, tested it at the senior fair in Germany, and compared its interactions with those of an actual robot.

This research examines the ongoing password and user authentication problem from the perspective of people accessing services on Apple smartphones. First, the current state of sign-in methods is characterised on iOS apps and the developers' rationale for their implementation. This work is going to be published at MobileHCI 2026. Building on this, a second study will study how users choose to authenticate when their choices are constrained, such as by device, policy, or method availability. The final study will propose an intervention: a cybersecurity training platform focused on user authentication security. This platform employs a conversational AI interface with a learning-by-teaching pedagogical approach to provide a unique training experience. Together, these studies aim to provide insight into the design, adoption, and education of users and authentication systems in the Apple mobile ecosystem.

People often spend more time on their smartphones than they intend. Researchers have developed various tools to reduce screen time. However, tools that rely on the blocking strategy may undermine users’ sense of autonomy, whereas tools based on the self-tracking strategy often provide insufficient exposure to effectively influence behavior. I propose integrating ambient displays with the self-tracking strategy: An always-on ambient display maintains users’ awareness of their phone usage. In Study 1, I learned that problematic smartphone usage often results from a loss of time awareness. In Study 2, I tested the effectiveness of an ambient display. In Study 3, I explored the intention-behavior gap in smartphone usage, a potential tailoring variable for just-in-time intervention. In Study 4, I will test adaptive ambient displays based on the intention-behavior gap. Through this line of research, I hope to guide users toward lasting behavior change while preserving their autonomy.

Most existing dietary feedback systems focus on calories or macronutrients, with limited support for connecting food choices to users’ subjective well-being, such as sleep, mood, and energy. We present DIIve, an exploratory dietary feedback system that uses the Dietary Inflammatory Index (DII) as a complementary lens for reflecting on possible diet–well-being relationships. A survey (N=218) showed users are interested in well-being-related outcomes but have limited familiarity with inflammation-related concepts. DIIve translates DII scores into more user-understandable feedback using intuitive tags, culturally inspired metaphors, and ecological visualizations, with meal suggestions generated via large language models (LLMs). In a 10-day field deployment (N=40), participants reported stronger connections between food choices and well-being and perceived progress toward their health goals. These results provide initial evidence on the feasibility and user experience of incorporating inflammation-aware feedback into everyday dietary tracking, suggesting that subjective well-being can serve as a complementary dimension alongside traditional nutritional metrics.

Smartphones are deeply embedded in everyday life, yet their overuse is often linked to declines in mental, physical, and social well-being. This study focuses on sleep quality, as problematic smartphone use has been shown to delay bedtimes, shorten sleep, and reduce recovery-related sleep phases. Digital self-control tools show promise in supporting healthier usage, yet few studies have examined their causal impact on sleep. We conducted a 6-week mixed-methods study with 16 participants to collect early insights on whether reducing smartphone use improves sleep quality. Using a proprietary Android application to monitor and limit screen time and a Fitbit Versa 4 to measure sleep, we found that the usage reduction features improved sleep hygiene, with earlier sleep timing, longer duration, and more time spent in REM and deep sleep. Our findings link self-motivated behaviour change to reduced PSU, better sleep quality, and lower levels of smartphone addiction.

Anorexia nervosa (AN) is a life–threatening psychiatric disorder primarily affecting teenagers. Within Family–Based Treatment (FBT), patients work closely with their families to regain agency over eating by gradually confronting “fear foods”, foods that cause intense anxiety. To support this process, we created Marbly, a mobile diary developed through patient–centred design. Marbly integrates gamification and emotional attachment mechanisms through digital and tangible companions to support gradual exposure in an engaging way. We examined Marbly through a multi–phase evaluation: exploratory in-the-wild deployments with patients (N=5) and healthcare professionals (N=6), a controlled online study with healthy participants (N=361) investigating attachment mechanisms, and extended use supported by app analytics. Findings suggest that emotional attachment and digital–tangible interaction may contribute to perceived motivation and emotional support during recovery. We contribute design insights for companion–based AN tools, explore emotional attachment in digital-tangible companions, and demonstrate a phased evaluation strategy for constrained clinical populations.

Online instructional workout videos often involve movements for different muscle groups, leading to frequent transitions between different body postures, such as prone, supine, and standing postures, which repeatedly shift the user’s viewing direction. However, users often place playback devices in a fixed location, so they may struggle to see the screen during certain movements, leading to frustration and an increased risk of injury. To mitigate this issue, we present MoveMate, a multi-device workout support system. Before the workout, MoveMate analyzes the instructional video, visualizes changes in the instructor’s face orientation and eye height, and provides device placement recommendations. During the workout, the system delivers additional visual and auditory guidance, helping users follow movements across different body postures. Our user study (N=24) showed that MoveMate improves users’ experience of following the instructor under reduced visibility. Our findings and reflections offer insights for designing future computer-assisted workout systems.

A lack of physical activity (PA) is linked to reduced well-being and health issues, yet physical inactivity continues to rise globally. Although prior research indicates that couple relationships impact health, health technologies (HTs) that support couples in adopting healthy behaviors remain underexplored. Using a Research through Design approach, this study explores HTs for cohabiting couples and their potential to promote PA. We developed a collaborative ambient HT designed to integrate into couples' shared space. To investigate couples' experiences with the prototype and its impact on PA awareness and behavior, we conducted a four-week technology probe study with seven cohabiting couples (14 individuals). Participants shared insights through diaries and interviews, and we collected usage data and step counts. We found that an ambient HT can direct couples' attention toward PA behavior, foster joint reflection, and promote PA. Our insights contribute to understanding how couple relationships can motivate healthy behaviors.

AI-driven chatbots are increasingly being used to support community health workers (CHWs) in developing regions. Yet little is known about how cultural frameworks in chatbot design shape trust in collectivist contexts where decisions are rarely made in isolation. This paper examines how CHWs in rural India responded to chatbot-interfaces that delivered identical health content but varied in one specific cultural lever: social norms. Through a mixed-methods study with 61 ASHAs who compared four normative framings: neutral, descriptive, narrative identity, and injunctive authority, we (1) analyze how framings influence preferences and trust and (2) compare effects in low- and high-ambiguity scenarios. The results show that narrative framings were most preferred but encouraged overreliance, while authority framings were least preferred yet supported calibrated trust. We conclude with design recommendations for dynamic framing strategies that adapt to context and argue for calibrated trust as a critical evaluation metric for safe, culturally-grounded AI.

Locomotion shapes the usability and comfort of immersive virtual environments. In particular, hands-free locomotion supports accessibility, yet most speech-based approaches rely on rigid command sets that limit natural and flexible interaction. We introduce a speech-based, context-aware, LLM-driven hands-free locomotion technique that enables free-form natural language navigation. We evaluate our approach against two baselines: controller-based teleportation and fixed-command voice steering. Our evaluation combines eye tracking and standardized questionnaires to assess usability (SUS), presence (IPQ), cybersickness (CSQ-VR), and cognitive load (NASA-TLX), complemented by an SHAP-based analysis. We found no evidence of differences in usability, presence, cognitive load, or cybersickness compared to established methods such as teleportation, while eye-tracking results suggest increased attention and engagement in the LLM-driven condition. In terms of performance, teleportation achieved the shortest task completion times, whereas the two hands-free voice methods (LLM-driven and fixed-command) were comparable. Overall, our results demonstrate the feasibility of LLM-driven hands-free locomotion. This technique is applicable to accessibility use cases such as hands-busy scenarios, and motivates follow-up studies with larger samples and more diverse tasks.

Turn-by-turn (TBT) and as-the-crow-flies (ATCF) navigation guidance each shape the outdoor cycling experience. Commonly, these methods present navigational cues from a cyclist perspective, disregarding world-anchored cues embedded in the environment. We investigate both perspectives combined with TBT and ATCF navigation in four implementations: (a) Turn Signs, (b) Intersection Arrows, (c) the Absolute Arrow, and (d) the Beacon. To support this investigation on an actual bicycle within a confined physical space, we additionally propose a novel Augmented Virtuality redirection technique that accommodates freedom of choice regarding cycling direction. From a controlled outdoor experiment (N=24), we found that participants using world-anchored TBT navigation successfully finished all routes and were the fastest and most efficient, and that the world-anchored perspective resulted in lower motion sickness and distraction. As a secondary contribution, our cycling redirection approach provided a high success rate while facilitating free exploration of virtual spaces larger than the physical one.

A common assumption in virtual reality (VR) is that stronger embodiment may be elicited when avatars more closely resemble users’ real-world appearance. However, avatars can be evaluated not only in terms of resemblance, but also how positively users perceive and accept the avatar they embody. In this study, we examine how avatar perceived similarity and likability relate to sense of embodiment (SoE) in a VR rhythm boxing exergame. Forty-eight participants embodied gender and skin color matched avatars and reported body ownership, agency, avatar perceived similarity, likability, and social presence of a voice assistant. Likability was consistently associated with both body ownership and agency, whereas perceived similarity was associated with body ownership only. The association patterns were stronger among female participants and under higher perceived social presence. These findings contribute empirical evidence that avatar appearance’s association with embodiment extends beyond visual resemblance, demonstrating the importance of evaluative judgments and contextual factors.

Bulky fan arrays and room-scale airflow systems have shown that wind can enhance presence in Virtual Reality (VR), but they are difficult to deploy and do not readily support head-worn, spatially coherent cues. We present AirHook, a lightweight ear-hook wind interface that delivers omnidirectional airflow around the periauricular region without interfering with a head-mounted display. AirHook renders spatial, dynamic 3D wind cues by activating a compact nozzle array with a head-contingent airflow rendering algorithm, enabling intermediate directions beyond discrete nozzle orientations. We evaluate AirHook in three studies: two perceptual studies respectively characterizing 3D direction recognition across stimulus durations and a 36° directional-resolution benchmark, and a VR study comparing No Wind, Non-directional Wind, and Multi-directional Wind, combining questionnaires and interviews. Results show that multi-directional wind provides additional experiential benefits beyond non-directional airflow and yields practical design implications for ear-based wind feedback.

Ray-cast pointing is a primary means of selecting targets in virtual reality, though small targets are difficult to select. Input assistance can help users select small targets by expanding effective target width, so they do not need to point directly at them, although this might diminish sense of control if it leads to unexpected behaviour. We investigate an input assistance method that enlarges effective target size and uses different feedback designs that hide or reveal assistance. Across two experiments (n1=20, n2=20), we evaluate its use with controllers and hand tracking in sparse and dense target layouts, with a focus on sense of agency. Results show that input assistance reduces movement time while maintaining and even strengthening the feeling of being in control. These findings show target expansion can improve VR pointing by enhancing selection without compromising agency, helping users feel in control as they achieve their interaction goals with less effort.

Finger-based locomotion promises space-efficient VR travel in public settings, yet prior work has largely centered on motion-based “finger walking,” leaving discrete tapping underexplored. This work expands the design space with TapWalk, two discrete tapping variants (position-based fore/behind zoning and posture-based tip/pad taps), and compares them to MotionWalk, a continuous sliding baseline synthesized from prior work. All techniques support forward/backward translation, yaw rotation, and strafing. In a within-subjects guided path-following study, TapWalk reduced completion time and physical/effort workload compared to MotionWalk, while MotionWalk was perceived as more “walk-like”; operation counts (a proxy for corrective effort) were comparable across methods. To probe more natural use, we further conducted an exploratory virtual-museum experience session with two hybrid techniques that layer MotionWalk onto each TapWalk variant, suggesting that users can cruise with TapWalk for step-accurate, low-fatigue navigation and switch into short MotionWalk bursts for brief accelerations and stride variation.

Future lightweight smartglasses will enable a continuous augmentation of the user's view with digital information. However, the effective integration of persistent overlays in Augmented Reality (AR) remains insufficiently understood. Where should frequently accessed persistent information, like widgets and information panels that do not refer directly to objects in the environment, be displayed in the user's view? In this work, we explore placement strategies for "AR Widgets" that, similar to traditional information widgets, should be easy to access. We propose two novel environment-aware placement strategies that dynamically reposition world-referenced widgets, and compare them to a more traditional view-fixed placement method with a user study (n=18) across sitting and walking usage scenarios. Our findings indicate that the choice of widget placement strategy impacts user experience, distraction, placement, and clear-view ratings, with the wall-aligned method scoring the highest.

Human-robot collaboration (HRC) often requires robot intentions and internal states to be conveyed to users for task efficiency and safety. Recently, augmented reality (AR) situated analytics provide such real-time robot feedback in HRC contexts. However, AR situated analytics can obstruct important real-world elements, posing safety and usability risks, especially when content is dynamically positioned relative to movements of mobile robots in a warehouse HRC scenario. In this paper, we introduce the Augmented Reality Technique Of Obstruction Deterrence while Aiding Robotic Teaming for Users (ARTOO-DARTU), an AR system tailored specifically for warehouse HRC that enables real-time robot situated analytics and control while preserving visibility of the real-world through an obstruction detection and mitigation pipeline (ODM) that is uniquely suited for AR-HRC. To evaluate ARTOO-DARTU, we developed Pocket MonstARs, a controlled gamified abstraction of HRC warehouse inventory picking in which virtual monsters serve as proxies for pick targets, while labeled and object-marked boxes preserve the real-world identification demands of the picking task. In a 34-participant user study, we found that our designed AR situated analytics yielded a 46% increase in efficiency on the overall HRC task, but only when the ODM was active. Participants with the ODM active were also 61% faster on subtasks requiring visibility of the real world. Our findings demonstrate that, when paired with our developed ODM to prevent real-world obstructions, the situated analytics in ARTOO-DARTU can significantly enhance efficiency and user experience in AR-HRC warehouse scenarios.

Many people watch videos on their phone while working out. However, the limited size and poor placement constrain how suitable the phone can be in this situation. We investigate how augmented reality (AR) screens could provide better viewing experiences during cardio workouts. Based on observations and interviews in the field, we determine requirements and then implement an AR system for watching video during exercising. We then compare phone-like, static, and adaptive AR screens in an in-situ study with 16 participants. Our results show that both AR screen options provide better viewing experiences than the phone-like screen, with adaptive AR screens being particularly well suited for rowing machine exercises.

As Extended Reality (XR) headsets increasingly capture surrounding audio-visual data, privacy concerns extend to bystanders who may be recorded or analyzed without awareness or consent. Supporting responsible XR use in public and shared spaces, therefore, requires mechanisms that make sensing visible to affected non-users. We designed three HMD-mounted bystander interfaces using an external display that varied the information shown about recordings: a recording icon, a raw data stream, and inferred bystander information. We evaluated them in an online study (N=66) and an in-situ study (N=12) with bystanders in public places. Bystanders preferred selective transparency: icons alone were insufficient, while explanatory text and raw data streams helped communicate ongoing recordings and capture scope. In contrast, inferred information could overcrowd the interface and reduce clarity. Bystanders also expected consent or intervention mechanisms that do not require direct interaction with XR users. We contribute design recommendations for privacy-enhancing XR bystander interfaces.

Cyclist fatalities in Europe have decreased by only 3.0% since 2012, compared to a 22.0% decline in overall traffic deaths. As cycling gains importance as a sustainable mode of transportation, improving cyclist safety is essential. Augmented Reality provides opportunities for effective warning systems through spatial anchoring, embedding information in the real environment rather than fixing it at a static point in the field of view. Although prior work has focused on highlighting specific Hazardous Objects, the spatial visualization of Hazardous Areas remains underexplored. This paper introduces a concept to visualize Hazardous Areas for cyclists, which we designed using a human-centered design approach and which we evaluated in a within-subject simulator study (n = 17). Results indicate that visualization increases subjective safety, lowers perceived workload, and achieves high usability and user experience. Importantly, it enables hazard perception beyond the direct line of sight, addressing key limitations of prior visual warning systems.

Diminished Reality (DR) can visually suppress physical objects and may help users manage attention in mixed-reality environments, yet live, user-controlled DR remains difficult to implement and evaluate. We present Selective Vision, a proof-of-concept MR system that combines RGB-D capture, hand-tracking input, inpainting, and 3D mesh updating to let users manually select, diminish, and restore real-world regions. In a study with 28 participants, we examined exploratory and task-oriented use. Participants quickly learned the interaction, used DR to reduce visual clutter, and generally preferred manual control over fully automatic removal. Although blending, alignment, and reconstruction artifacts remained visible, diminished regions were often perceived as acceptable and sometimes less distracting than the original objects. Our findings suggest that future DR systems should prioritize fast, low-effort selection, transparent feedback, reversibility, and user-confirmable suggestions, while improving segmentation, latency, viewpoint robustness, and temporally stable reconstruction.

Among older adults striving for healthy ageing, combining motor and cognitive exercise may help preserve functional capacity in later life. Exergames have shown effectiveness in delivering dual-task training and can complement telemedicine by offering engaging interventions with high adherence. This paper presents a mobile exergame with dual-task exercises and avatar-based coaching, designed as a low cost and pervasive solution using pose estimation from the tablet built-in RGB camera. Intuitive motion-based controls, immersive techniques, and motivating feedback mechanisms were evaluated in a usability pilot study with 27 predominantly young participants. The system achieved a mean System Usability Scale score of 81.2 (𝜎 = 11.02) and positive perceptions of the intuitiveness of the controls and user competence. These preliminary findings support the feasibility of mobile coaching-based exergames for combined motor and cognitive training and motivate further exploring of presence-enhancing features.

Parents use parental control applications (PCAs) to manage their children's smartphone use---setting screen time limits and monitoring device activity. However, as parents and children bring different needs to these tools, the tensions they raise for the parent--child relationship have only recently come into focus. Analyzing 734 reviews of \textit{Google Family Link}, a popular PCA, written by both sides in South Korea, we examined how PCAs reshape parent--child dynamics. We found that PCAs affect parent--child relations---undermining trust, constraining agency, and resolving conflict away from direct conversation. Moreover, our findings reveal a developmental concern: a PCA that offers no space for negotiation can leave children in a passive role, accustoming them to a supervised relationship in which they feel they have little say. We propose design directions that support negotiation at each stage of app use---entry, ongoing use, and end---bringing both parents and children into the same space to make decisions together.

Menstrual tracking has begun to appear inside Chinese mobile romance games, where cycle records trigger reminders and responses from fictional male partners. This late-breaking work compares four games through critical interface and disclosure analysis, drawing on official materials, privacy disclosures, dated interface records, and a closely coded thirteen-post Xiaohongshu corpus. Across the four games, the interface returns a player’s own record as evidence that a character remembers and cares, while the player performs the entry, correction, and interpretation work that makes such attention possible. This paper calls this arrangement relational self-tracking. The concept connects personal informatics to the gendered organization of care in romance games and directs design attention to editability, bodily variability, and legible data practices.

Mobile fitness apps often motivate users through individualised performance mechanisms such as streaks, badges, rankings, and leaderboards. However, these mechanisms may not fit older adults' walking routines, which are often embedded in small, familiar groups, flexible participation, and socially meaningful encouragement. Building on prior work on Group Activity Metrics (GAMs)---group-oriented activity representations designed to support social interaction around fitness---this paper presents the design and implementation of a mobile fitness app that translates GAM-informed design principles into collaborative and comparative visualisations for older adult walking groups. These visualisations support goal-relative comparison, collective progress, shared journeys, care-aware participation, and moderator-supported feedback. This Late-Breaking Work paper contributes a design translation case, a set of collaborative and comparative group visualisation patterns, and considerations for field deployment and evaluation.

People with aphasia often rely on communication partners, such as family members, friends, and caregivers, to support everyday interactions. Although communication partner training can improve communication outcomes, applying learned strategies in real-world situations remains challenging, and support outside structured learning environments is often limited. To explore how mobile technologies might address this gap, we conducted an early-stage co-design study of ANCHOR, a mobile learning and communication support application for communication partners of people with aphasia. Using paper-based prototypes, task-based walkthroughs, think-aloud activities, and semi-structured interviews, we explored participants’ needs, experiences, and design preferences. Our analysis revealed six key design requirements: (1) support beyond formal training, (2) personalized access to communication strategies and resources, (3) conversational and exploratory learning, (4) opportunities for user contribution and feedback, (5) support for therapy engagement, and (6) meaningful recognition of learning and progress. Collectively, these requirements expand the role of mobile technology from delivering educational content to providing ongoing support across everyday communication and therapy contexts. We discuss implications for designing mobile systems that integrate learning, communication support, and therapy engagement. This work contributes user-centered design requirements for mobile technologies that help bridge communication partner training and its application in everyday communication.

Depression pre-screening is moving onto consumer headsets and phones, where it reaches exactly the people least likely to reach a clinician—and where a wrong answer is not a lost benchmark point but a missed case or a false reassurance. The accuracy numbers that justify this move look strong, yet each is reported from a single context. We treat this as a measurement question: using four public EEG corpora and a deliberately conservative, subject-disjoint protocol with source-confound, permutation, and feature-validity controls, we ask how much of each number is a transferable signal and how much is an artifact of how it was measured. The one strong result (AUROC 0.94, robust across logistic regression, SVM, and random forest) entangles class with data source: the same features separate the two source datasets at 0.87, and only a within-source check (0.93) shows the signal is partly genuine. In a single-source corpus, where that entanglement cannot arise, the same features reach only 0.66 [0.54, 0.77]—far below that headline—and even there an acquisition marker tracks severity (+4.2 BDI, [+0.3, +8.0]). Two consumer-grade cohorts are indistinguishable from chance (permutation p = 0.85 and 0.17), and cross-cohort transfer fails even after harmonization. We therefore contribute evaluation hygiene—source-confound probes, permutation tests, feature-validity audits, subject-disjoint splits, and non-circular labels—as the discipline a mobile screening claim must pass before it is believed, and, because single-context signals are this fragile, a context-aware, auditable, human-supervised multimodal architecture as the design response. We implement and test its context gate, and the test yields a further result: feature-distance detection waves through a consumer-grade cohort that carries no signal, so the gate must key on provenance, such as device metadata, not feature distance alone—and a working prototype of our own makes the ungated failure concrete. Grounded in the hygiene analysis and this gate check, our central claim is constructive: evaluation hygiene together with context-gated, auditable, human-supervised screening is what lets mobile multimodal mental-health assessment make safe, auditable, and context-sensitive claims across populations, devices, and cultures.

Informal caregivers often experience substantial psychological distress, yet tailored digital interventions to support their well-being remain limited. This study evaluated CHILL, a mobile application designed for emotional regulation, to assess its perceived usefulness and identify areas for improvement. Ten caregivers (mean age = 53 years; 90\% female) engaged with app features including guided activities, relaxation exercises, sleep sounds, and an AI-based therapist. Data collected through questionnaires and semi-structured interviews were analyzed using thematic analysis. Engagement with the app was selective and influenced by users’ mood, with most participants valuing its flexibility. Relaxation features were reported as the most effective component, with most participants highlighting their calming auditory design and some willing to incorporate them into daily routines. The AI therapist was perceived as accessible and non-judgmental by half of the participants, although some users noted a lack of emotional depth in its responses. The activities section was the least effective feature: most participants found it repetitive and unclear, and they also expressed a preference for more structured, goal-oriented tasks. Overall, CHILL demonstrates potential as a digital tool for stress support among informal caregivers. However, enhancements in adaptivity, structure, and multimodal content may improve user engagement and effectiveness. Future research should examine its long-term impact on caregiver well-being.

This work presented a mobile digital coaching system that combines a locally deployed Large Language Model (LLM) with the Health Action Process Approach (HAPA) behavior change model to support physical activity through smartwatch-integrated, conversational self-monitoring. We conducted a three-week exploratory study (N=17) comparing a control condition (CON) with motivational phase and passive data visualization and an intervention condition (INT) including daily LLM-based volitional phases. Usability outcomes indicated good feasibility (SUS=72.3; BUS-11=3.44/5). Qualitative feedback valued action planning feature, goal personalization, and progress visualization, while also pointing to a clear need for more concise, adaptive, and interactive conversational content. Motivational outcomes showed heterogeneous but promising trends, with coping self-efficacy and action planning improving more in the INT than in the CON. Overall, the study supports the feasibility of system, and it identifies concrete design directions to guide a larger-scale, longer-term evaluation of the effectiveness for behavior change.

Outdoor plant guiding differs from indoor viewing experiences in that points of interest change with weather and seasons. Therefore, systems designed for indoor viewing experiences are difficult to apply directly. Through three participant observations, we identified issues such as the need to follow the preceding guide, explanations not reaching participants at the rear, the importance of direct interaction with actual plants, and the distortion of information through verbal relay. Building on these observations, we implemented a mobile prototype that places guide videos recorded by preceding guides at the same locations as the plants using location anchors. This enables following audience members to view them later at the same location. We conducted a workshop using this system in a botanical garden and report early design insights and challenges for supporting self-paced outdoor plant exploration.

Recent work has shown that a BERT-based neural decoder can convert finger-tap sequences into text at 1.6% character error rate (CER), relying only on finger identity without positional information. Wearable glove keyboards, however, can supply additional signals, including finger bend angles from stretch sensors and hand orientation from IMUs, indicating where on the layout a key was pressed. We investigate whether injecting simulated sensor tokens into BERT's input reduces CER beyond finger-identity-only decoding. Following TypeAnywhere's methodology with a smaller training set (3.6M sentences vs. the original 44M), we train four model variants on Yelp/Amazon reviews and evaluate on the Cornell Movie Dialogs corpus, varying only the presence and encoding of sensor tokens. Row tokens from stretch sensors cut CER from 2.46% to 0.83% (−66%, p < .001). Adding column tokens as separate tokens, however, increased CER to 1.48%; the penalty persists even with noiseless signals. The cause is sequence dilution: extra tokens consume the decoder's fixed input budget without proportionate benefit. A combined-token variant that packs finger, row, and column into a single token per character achieved 0.57% CER, recovering the column benefit without the sequence penalty (p < .001 vs. row-only), and remains the most noise-tolerant variant, beneficial up to σ ≈ 6°. These results demonstrate that for wearable text entry, how sensor information is encoded matters as much as which sensors are used.

Unmanned aerial vehicles (UAVs) are increasingly used in disaster response, surveillance, and environmental monitoring, driving interest in multi-drone systems. However, much research remains confined to simulation, limiting real-world applicability. We present DroneHub, a modular multi-UAV platform for rapidly developing functional research prototypes with physical drones. DroneHub comprises: (1) a drone fleet, each with a dedicated controller running a custom Android application; (2) a central server coordinating flight commands and telemetry; and (3) a web interface for configuring missions, generating flight paths through external APIs, and monitoring real-time data. Reusable software and interface components reduce development overhead and support human-drone interaction research with physical systems. We describe the design rationale, architecture, safety features, and ethical and legal considerations. Drawing on four years of development and evaluations with domain experts, we also derive transferable design considerations for modularity, observability, experimental control, recovery, and adaptability in HCI research infrastructures for networked physical systems.

This paper investigates the role of physical context in virtual nature experiences by exploring VR use in winter outdoor environments. We explore the user experience (UX) with two virtual nature scenes—a winter landscape and a summer landscape—experienced in two physical settings: a snowy winter outdoor environment and an indoor laboratory. Our aim is to understand how winter outdoor conditions influence the holistic user experience, including immersion, presence, and multisensory experiences. The results suggest that participants experienced the virtual environments not only visually but also in a multisensory, bodily, and contextual manner. Our work contributes to research on in-the-wild VR by examining its use in winter outdoor settings, a study context that has received little attention in prior research.

We present FretPhone, a guitar-inspired side-of-device input method that augments mobile e-reading through tactile interaction. The design leverages string-based gestures, touch, strum, and slide performed along the device-edge to support navigation and control without relying solely on the touchscreen. We developed a working prototype featuring three tensioned guitar strings and a sensing mechanism for gesture detection. A user study (N=10) evaluated gesture recognition accuracy, usability, and user-defined gestures for common e-reading tasks (scrolling, skipping chapters, activating text-to-speech, changing font-size). Results show that FretPhone could reliably classify inputs, with touch gestures outperforming strumming and sliding gestures in accuracy. Participants reported finding the interaction as engaging and easy-to-use, while experienced guitarists highlighted improvements for guitar-inspired input. These findings suggest that instrument-inspired, edge-based interactions offer a promising direction for enriching mobile reading. We outline opportunities for improving sensing fidelity, expanding gesture vocabularies, and refining hardware's form factors.

Digital food-sharing platforms can help reduce waste and strengthen local communities, but little is known about how effectively they support the redistribution of surplus fresh apples. We conducted semi-structured interviews with three apple growers and two staff members from a food redistribution organization and derived nine collaboration requirements. We then used these requirements to review five existing surplus food redistribution tools. Our review shows that current tools support collaboration through features such as location-based discovery, collaboration profiles, and mechanisms for describing surplus items. However, support for making expected value visible before commitment, assessing apple-specific feasibility, and maintaining reliable coordination remains limited. This work addresses collaboration requirements and highlights design opportunities for mobile applications that support surplus redistribution.

Synthetic voice technologies are often discussed as a detection problem: can people distinguish AI-generated voices from real ones? We report early qualitative findings that suggest a different emphasis. Based on semi-structured interviews with 11 participants familiar with synthetic voice technologies, we found that participants described suspicious voice encounters less as failures of authenticity detection and more as situations in which urgency, emotional pressure, and the synchronous nature of voice calls reduced opportunities for reflection. They described these encounters as emotionally demanding situations marked by urgency, fear, and reduced time for reflection. Even when participants were aware that synthetic voice deception exists, they felt that real-time calls could override caution by creating social and emotional pressure. We therefore argue that synthetic voice deception should be considered not only a media-authenticity problem but also an interaction design problem for voice communication systems, with implications for MobileHCI. We discuss exploratory design implications, including friction in high-risk calls and intentional desynchronization of suspicious voice interactions.

Individuals with breast cancer often experience cancer-related cognitive impairment, affecting attention and daily functioning, yet these difficulties remain challenging to assess outside clinical settings. This paper presents a low-cost, home-based mobile virtual reality (VR) platform integrating brief symptom self-reports with an immersive task to assess selective attention. We report findings from a preliminary study with six individuals with breast cancer who used the system independently at home over one week. Results indicate high usability and a moderate sense of presence, supporting the approach's feasibility. The VR task was sensitive to cognitive load, with increased difficulty leading to slower and less accurate performance, providing preliminary support for its validity as a cognitive-attentional measure. Exploratory analyses suggest that daily symptom fluctuations, particularly stress and energy, were associated with cognitive accuracy. These findings highlight the potential of mobile VR as an accessible, ecologically valid tool for real-world cognitive monitoring in cancer care.

Selecting the right emoji on a mobile keyboard remains a persistent usability challenge. We present MojiMatch, an Android keyboard that integrates two automated emoji prediction modalities: text-based sentiment analysis and facial-expression recognition, alongside a standard manual selection baseline. In a within-subjects experiment, text-based prediction matched the standard keyboard on typing speed and input time, suggesting it can enrich emoji selection at near-zero efficiency cost. By contrast, facial-expression prediction was significantly slower, likely because it required additional computation and imposed greater attentional demands. Despite this efficiency gap, the majority of participants preferred the face-based modality for its perceived expressive richness, revealing a preference-performance dissociation. These findings provide the first controlled, within-subjects empirical characterisation of the efficiency-expressiveness trade-off that future hybrid emoji prediction systems must navigate.

Pet loss is among the most common yet least socially acknowledged forms of bereavement. For many owners, grief resurfaces through everyday routines once shared with the animal. Existing continuing-bonds technologies rarely engage with these moments: digital memorials are largely static, while griefbots designed for human loss remain reactive, responding only when users initiate contact. We present MemoryAn, a proactive, generative conversational agent designed to sustain an owner's continuing bond with a deceased pet. Grounded in formative interviews with six bereaved owners, and using a small set of photographs and owner narrations, MemoryAn builds a Pet Identity Model combining a visual identity anchor and a repertoire of habitual behaviors. A context-aware controller decides when to initiate interactions, while an identity-preserving generation pipeline renders personalized images and short videos of the pet. We contribute an exploratory prototype, a design-space framing of proactive, generative re-presencing for companion-animal loss, and an ongoing evaluation agenda examining its benefits and risks.

Advances in multimodal large language models have enabled mobile GUI agents that can interpret screen content and autonomously perform tasks across applications. While prior research has largely focused on task completion and agent autonomy, less attention has been paid to how users can remain involved in the delegation process and shape agent behavior. We present PRIMA (PReference- Informed Mobile Agent), a mobile GUI agent that supports user agency through six mechanisms: specifying preferences, defning execution steps, establishing approval conditions, clarifying requests, reviewing and revising generated plans, and allowing user interference. We further report fndings from a preliminary feld deployment. Results suggest that participants actively established delegation boundaries, developed trust through accumulated experiences, learned the agent’s capabilities and limitations through use, and adapted the agent through preferences. Participants also valued clarifcation when it reduced ambiguity but preferred it to be used selectively. These fndings provide insights for more user-aligned mobile GUI agents.

We present SENSi, a smart career mentoring agent powered by a large language model (LLM), designed to emulate real-world mentoring behaviors through structured dialogue. Drawing on insights from live mentoring sessions and informed by cognitive task analysis (CTA), we developed an expert model of mentor strategies using the Hierarchical Task Analysis (HTA) framework. This model guided a prompt engineering process that enabled the agent to provide personalized, reflective, and context-aware career guidance. We evaluated SENSi in a formative usability study with twelve participants from engineering and technology fields. Participants rated the system highly on relevance, clarity, personalization, and motivational support, and emphasized its effectiveness in providing actionable and trustworthy guidance. These findings highlight the potential of LLM-driven agents to expand access to mentorship and inform the development of more adaptive career support systems.

Voice messaging remains a unimodal experience centered on recording and listening to audio. Current speech analysis models could supplement messages with automated transcriptions and visual emotion cues, such as emojis based on the speaker's tone or the verbal content. However, how these proactive features affect UX is not yet well understood. To address this gap, we introduce Speejis, a proactive agent that leverages a Speech Emotion Recognition (SER) model to automatically augment voice messages with emojis and colored waveforms, alongside automatic speech recognition for transcriptions. We present the Speejis prototype and report findings from an exploratory user study (N = 12). Our results detail how Speejis impacts UX, highlighting a trade-off between increased stimulation and reduced dependability due to the fully automated nature of emotion cueing.

Wearables expose signals about a person's physiological, cognitive, and affective state, and large language models (LLMs) can turn them into personalized guidance. Existing systems typically run in the cloud and emit free-form text. We present Syni, an agent built on a strict separation: a deterministic upstream engine produces a versioned Human State Interface (HSI); Syni is purely its consumer. This lets Syni run offline-first on-device, condition through an audited seam, guarantee schema-safe output via grammar-constrained decoding with a generate–validate–repair–fallback pipeline, and keep raw biosignals off the model—simplifying deployment, validation, and privacy. In preliminary measurements, an unversioned renderer change between training and inference silently collapses state-reflection from 85% to 0% while schema-validity stays 100%—a failure no schema check catches. Grammar-constrained decoding lifts a 1B model's first-pass validity from 0% to 100%, and coach inference runs on an iPhone 17 Pro in ∼2 s—showing contract-mediated, on-device agents are practical.

Myoelectric control uses surface EMG signals from muscle contractions to operate prosthetic limbs. However, learning intuitive myoelectric control of multi-joint movements and diverse grasp patterns remains a major challenge for broader user adoption of multi-articulating prosthetic hands. The training required to attain these control skills is intensive and often inaccessible in areas with limited rehabilitation resources. To address this gap, we developed a mobile virtual reality (VR) solution that offers multi-articulating prosthetic training and assessment in ecologically valid tasks. We evaluate platform usability through real-time performance metrics capturing user interactions, and semi-structured interviews capturing end-user perspectives, conducted during VR training sessions with a clinician and a participant with upper-limb loss (ULL). Training and assessment functionalities were verified, and users reported positive usability outcomes. These findings suggest our VR platform's potential as an accessible and effective prosthetic rehabilitation tool in remote areas.

Bulky fan arrays and room-scale airflow systems have shown that wind can enhance presence in Virtual Reality (VR), but they are difficult to deploy and do not readily support head-worn, spatially coherent cues. We present AirHook, a lightweight ear-hook wind interface that delivers omnidirectional airflow around the periauricular region without interfering with a head-mounted display. AirHook renders spatial, dynamic 3D wind cues by activating a compact nozzle array with a head-contingent airflow rendering algorithm, enabling intermediate directions beyond discrete nozzle orientations. We evaluate AirHook in three studies: two perceptual studies respectively characterizing 3D direction recognition across stimulus durations and a 36° directional-resolution benchmark, and a VR study comparing No Wind, Non-directional Wind, and Multi-directional Wind, combining questionnaires and interviews. Results show that multi-directional wind provides additional experiential benefits beyond non-directional airflow and yields practical design implications for ear-based wind feedback.

Finger-worn IMUs have been shown to be very accurate at detecting touch events on physical surfaces, enabling input to physically-bound XR interfaces. In this work, we describe a new XR headset method for detecting finger touch events on ad hoc surfaces, without needing the user to wear anything on their fingers or hands. Instead, our method uses a pair of high-speed headset cameras to track and triangulate the position of the user's fingertips in 3D space. We take the first and second derivatives of this signal to estimate a finger's velocity and acceleration. At almost 800 Hz, our system's output is similar to a worn IMU, but with bare hands. Using this signal, we can then detect rapid and characteristic decelerations of the finger, indicative of touch.

This paper presents a practice-based study of eight generative, biosignal-driven interactive installations deployed across six diverse locations in Europe and Asia between 2024 and 2026. From these deployments, we develop a modular framework for designing context-aware generative installations for adaptive public space. We examine how system transparency and audience interpretation shift across varying spatial and institutional contexts, introducing two core constructs: the system transparency spectrum, which distinguishes between installations that reveal their underlying computation and those that conceal it, and the contextual interaction threshold, which defines the level of system exposure an audience is willing to engage with. These constructs are underpinned by a modular, node-based generative media pipeline featuring interchangeable sensing, mapping, and output layers, which enables rapid redeployment across diverse venues without requiring architectural redesign. We support this framework with a live demonstration of the pipeline, grounded in the landscape and cultural history of its host city. Our findings indicate that the successful deployment of mobile and situated HCI systems in public spaces depends not only on technical robustness but on contextually adaptive design strategies that negotiate the relationship between system legibility and audience expectation.

Mathematical answer input remains a practical obstacle in online assessment systems, especially on mobile devices. In Moodle/STACK, one of the online assessment system for mathematics and STEM, students can submit algebraic expressions that are evaluated by computer algebra system, but they must enter answers using STACK/Maxima-compatible textual syntax. This creates a gap between natural handwritten mathematical notation and textual answer input. We present STACK Input Helper, a Moodle local plugin that supports camera-based mathematical answer input for STACK questions. The plugin enables students to upload handwritten expressions from a smartphone, recognizes them using mathematical OCR, converts the recognized LaTeX into STACK-compatible syntax, and inserts the selected result into the original answer field after user confirmation. The prototype supports QR-code-based mobile upload, multi-line answer selection, final-line recommendation, and partial-expression selection. This demo allows attendees to experience the full workflow of capturing handwritten mathematics, reviewing recognized candidates, and inserting a converted STACK answer.

Prolonged sedentary behavior is difficult to interrupt because conventional reminders often rely on users to self-initiate an activity. We present Haysun, a phone-native micro-exergame that transforms sedentary-break prompts into short motion-controlled battles. Holding the smartphone in hand, users perform movements such as simulated jump rope, squats, or punches, which are detected through the phone’s inertial sensors and mapped directly to attacks against a Desk Monster. Counter-attacks require rapid directional phone movements, while visual, audio, and haptic feedback support interaction when sustained screen attention is impractical. The installable mobile web prototype requires no wearable, camera, or external equipment. Haysun demonstrates how commodity smartphones can reduce the interactional and motivational friction between receiving a sedentary reminder and beginning an active microbreak.

University departments utilise surveys to assess student life and improve the overall campus experience. However, declining response rates create significant challenges in securing accurate data. Building on the concept of gamified physical surveys, this research introduces a demo of an onsite campus exploration game integrated with a digital survey system. The demo explores a new framework for institutions to alleviate survey fatigue, moving away from low-engagement forms to capture the student voice through active, onsite events.

Full paper abstract: Anorexia nervosa (AN) is a life–threatening psychiatric disorder primarily affecting teenagers. Within Family–Based Treatment (FBT), patients work closely with their families to regain agency over eating by gradually confronting “fear foods”, foods that cause intense anxiety. To support this process, we created Marbly, a mobile diary developed through patient–centred design. Marbly integrates gamification and emotional attachment mechanisms through digital and tangible companions to support gradual exposure in an engaging way. We examined Marbly through a multi–phase evaluation: exploratory in-the-wild deployments with patients (N=5) and healthcare professionals (N=6), a controlled online study with healthy participants (N=361) investigating attachment mechanisms, and extended use supported by app analytics. Findings suggest that emotional attachment and digital–tangible interaction may contribute to perceived motivation and emotional support during recovery. We contribute design insights for companion–based AN tools, explore emotional attachment in digital-tangible companions, and demonstrate a phased evaluation strategy for constrained clinical populations.

Museums often use objects, space, light, colour, and display order to teach visitors. These choices can exclude visitors who are blind and visually impaired (BVI) when access depends on sight or long verbal description. This article presents four prototypes for new MobileHCI interaction techniques in museum access. The main prototype is a microcontroller touch system that connects tactile exploration with spatial sound. Its new enabling technology is the joining of sensors, tactile imaging, embedded control, and spatial audio in one touchable museum object. Visitors touch zones on a model and hear short audio, tones, or sound textures. Three related prototypes follow the same concept: a sonified tablet painting that vibrates, a spatial soundscape using The Starry Night, and a face-tracking system which plays with the change of sound depending on the wearer's face motion. The article is contributing with a clear design argument on sonified touch as a novel interaction technique in MobileHCI.

Aerobic exergames can make physical activity more engaging, but many still rely on dedicated motion hardware or adopt game mechanisms that conflict with the bodily constraints of exercise. In particular, conventional failure and replay loops can make workout duration unpredictable, while repetitive exercise sessions may lose novelty over time. We present Pulse Raid, a phone-native rhythm-boxing exergame that explores how accessible mobile sensing, meaningful challenge, and structured exercise can be combined in a playable demo. Using a smartphone camera, players punch, dodge, duck, and block in response to rhythm-based enemies. The prototype introduces two design directions for mobile exergames: fail-forward recovery, which turns failure into a bounded physical penalty rather than a full restart, and between-song roguelite upgrades, which add session-level variation while preserving the prescribed movement sequence. Future work will evaluate how these mechanisms affect enjoyment, exertion, perceived fairness, and continued participation.

The growing market for used smartphones presents challenges for evaluating device quality beyond cosmetic condition and battery health. Degradations that affect everyday experience—including audio, speech, video, and camera quality—often remain hidden during conventional inspection. We present \textit{Phone Pathology}, an interactive smartphone diagnostic system that evaluates experiential quality through automated multimedia assessment. The system combines mobile-based testing, perceptual quality analysis, and an interpretable report card summarising performance across video, audio, speech, and camera dimensions. To date, Phone Pathology has been deployed across 76 smartphones through voluntary participation, demonstrating the feasibility of experiential quality assessment in real-world settings. Phone Pathology illustrates how perceptual quality metrics can support more transparent smartphone evaluation and informed decision-making in second-hand device markets.

Sign language users frequently encounter communication barriers when interacting with hearing individuals due to the limited availability of accessible technologies that support natural sign-based communication. We present SignTalk, an interactive AI-powered system that enables bidirectional communication between Deaf and hearing users through real-time sign language recognition, conversational AI, and avatar-based sign generation. The system combines computer vision techniques for sign recognition, transformer-based language understanding, and a 3D avatar capable of producing sign language animations from text inputs. The demonstration allows attendees to experience multiple interaction modalities, including live sign-to-text translation, text-to-sign avatar generation, and synchronized visualizations of video input, pose landmarks, and avatar animations. SignTalk showcases how recent advances in artificial intelligence and human-computer interaction can support inclusive communication and improve accessibility for sign language communities. The demo highlights practical applications in healthcare, education, and everyday communication while illustrating the potential of multimodal conversational systems for underserved sign languages such as Ghanaian Sign Language.

We present Touch to Listen, a mobile interaction system where a single tap of an NFC‑enabled smartphone on a tagged everyday object triggers immediate audio playback — no menus, no navigation, and no screen interaction. In this demo, attendees directly perform both interactions side‑by‑side in the booth: the one‑tap physical trigger and a conventional multi‑step app interface, making differences in speed, attention, and effort immediately tangible. The system uses only passive NFC tags, a standard smartphone, and locally stored audio, and has been explored in a home deployment. Attendees are invited to move tags themselves and discuss how shifting control from screens to physical placement changes the rhythm of everyday information access.

We developed an iOS application that enables STACK-based mathematics e-learning in a mobile environment, along with a standalone iOS custom keyboard application ``iFlickMath'' derived from it. The conventional STACK system was designed for PC browser use; on smartphones, the complexity of mathematical input and poor screen visibility significantly impair learning efficiency. We developed an iOS app that connects to STACK using the Moodle Mobile Web Service API without requiring additional server-side plugins. Furthermore, we redesigned and released the math-input-specialized keyboard ``iFlickMath'' as a standalone iOS custom keyboard app. iFlickMath employs the flick input method widely used in Japan, enabling input of integral symbols and function names with fewer taps. By switching between STACK mode and TEXT mode, it supports mathematical input not only for STACK exercises but also in any app such as word processors.

Voice over IP (VoIP) has existed since the mid-1990s, and guidelines from the International Telecommunication Union (ITU) recommend a maximum one-way delay of 400 ms for acceptable call quality. However, this guideline was established without explicitly considering users' awareness and does not consider how mobile interfaces can shape users’ understanding of network behavior. We design a smartphone interface that provides real-time visual feedback of speaker’s audio status, making delay and packet loss visible during calls. Using this interface, we study two research questions: (RQ1) whether users prefer higher reliability over lower delay under poor mobile network conditions, and (RQ2) whether appropriate notification increases tolerance for delays beyond the ITU threshold. User study results show that all participants (61/61) preferred higher reliability over lower delay in degraded conditions. Across personal and professional mobile contexts, users tolerated one-way delays exceeding 1.2 seconds when informed in advance. These findings highlight how interface-level awareness can reshape delay tolerance for VoIP communication.

research categorizes social smartphone use as either convergent or divergent, yet observational studies miss the internal states defining these behaviors’ appropriateness. To bridge this gap, we conducted a video-elicited interview study with 29 participants (12 groups), utilizing immediate playback to unpack the discrepancy between visible behavior and internal perception. Our analysis surfaces three patterns of intra-group perception: (1) Asymmetry of Intent, where pro-social divergent behaviors are misconstrued by observers as withdrawal; (2) Burdened Convergence, a paradox where users perform socially correct engagement at the cost of invisible cognitive strain; and (3) Intersubjective Alignment, where screen visibility serves as a safety net for conversational repair. We conclude with design implications for mitigating information asymmetry and alleviating the hidden cognitive burden of social presence.

Location sharing has increasingly become an integrated part of mobile phone use. In this paper we present a study that investigated the trajectories of how parental strategies of providing youth with smartphones in childhood led to long term impact on youth’s own location-sharing behavior in adulthood. Through retrospective interviews with 17 emerging adults (18-22), we found that most of their parents took a gradual approach, sometimes starting with basic phones to help their children adapt. As youth transitioned to full smartphones, tensions arose—parents used them for easier communication and safety, while youth primarily saw them as social tools. Our findings revealed that youth tracked by parents during childhood often continued sharing their location in college, despite discomfort, and also shared with others for safety reasons. These results highlight the need for a balance between youth privacy and parental peace of mind in location-based technologies.

The social environment determines the activities we can perform. On trains for example, reading is difficult if your seat neighbor is making a phone call. Likewise, making a phone call can feel uncomfortable if everyone around is working. Through such co-located activities, we collectively produce social situations and ``soundscapes'' that prohibit and facilitate certain activities. This paper models this collective process with an Agent-based Model of readers, callers, and listeners on trains. We found that variations in the number of passengers, their considerateness, and activity plans can explain fundamentally different soundscapes, which in turn shape possible activities and (user and witness) experiences. Moreover, we compare the effects of three interventions (noise-canceling headphones, silent speech devices, and quiet zones) on experiences and possible activities. We found positive but differential effects of each on readers, and additional positive effects of silent speech devices on callers.

As Mixed Reality (MR) technologies are increasingly adopted, understanding users’ and bystanders' privacy perceptions is crucial. Prior studies have examined MR privacy perceptions in online, lab, and field settings, but rarely compared these settings within the same study design. We address this gap with a between-subject study, comparing online, lab, and in-situ settings with 23 participants each (total N=69). After interacting with an MR headset, participants completed a questionnaire about privacy implications for users and bystanders. Quantitative privacy ratings were largely consistent across settings, suggesting that broad MR privacy perceptions can be elicited robustly across contexts. However, settings shaped how participants expressed these perceptions: online participants provided more qualitative statements, whereas lab participants produced a higher proportion of positive statements. Our findings suggest that research settings may matter less for eliciting broad MR privacy concerns than for shaping the richness and tone of qualitative accounts.

Many individuals living abroad rely on technological mediation such as video and phone calls to bridge the distance. Recently, telepresence robots (TPRs) have gained traction as an alternative means of interactive communication for interpersonal relationships in domestic settings. This paper explores the potential of TPRs as a communication medium between people living abroad and their loved ones. We employ autobiographical accounts, semi-structured interviews, and questionnaires to analyze how embodiment and sense of presence are experienced during TPR use in domestic spaces. Our findings show that embodiment is often fragile and transient, and while the spatial presence is elevated through TPR mediation, social presence is reconfigured as social dynamics are negotiated anew during communication through the TPR.

The growth of mobile dating platforms has coincided with a rise in romance scams, in which offenders construct convincing personas to defraud users. While research on romance scams is expanding, victims’ lived experiences of recognizing and responding to deception in mobile-mediated interactions remain insufficiently understood. To address this gap, we conducted in-depth interviews with 24 victims of online dating romance scams in Iran, where legal, social, and cultural constraints limit formal support. Our analysis identifies suspicion cues and the investigative strategies victims use to verify identities across platforms. We show that victims are not passive recipients of deception but engage in active, iterative detection practices under significant emotional, social, and relational pressure. Based on these findings, we contribute empirically grounded insights into deception cues and user-driven detection work, and we discuss implications for the design of mobile technologies that better support users in identifying, resisting, and recovering from romance scams. Content Warning: This paper discusses sexual violence.

Augmented reality (AR) has shown promising results in training for medical procedures. However, it can lead to attentional tunneling, where only the examination area is focused on. Observing the patient's face is an important part of medical examinations to detect signs of pain. To address this, four attention guidance cues were developed to complement medical AR training scenarios. Two visual cues, a subtle, mid-peripheral flashing and an overt circle segment, and two auditory cues, a spatial non-diegetic ticking sound and spatial diegetic breathing sounds, were implemented. In a mixed-design user study (n = 80), performance and gaze behavior were compared before, directly after, and one hour after training with the cues in a dual-task involving Training significantly improved task performance for cue and control groups. Although no proposed cues individually differed significantly from the control group, cues overall, and auditory cues in particular, improved viewing behavior.

In-the-wild automotive user studies are time-consuming, costly, and challenging to scale. Yet, early and fast feedback is crucial for designing new intelligent in-vehicle applications in a user-centric way. This paper explores whether generative agents can simulate participants and produce synthetic data for such studies. We replicated two real-world cases based on interruptibility assessment and emotion recognition, with agents embodying synthetic personas exposed to multimodal driving contexts. We evaluated the alignment with human data through majority-vote consensus and a machine-learning generalization test, focusing on personas, contextual templating, and drive–goal conditioning. Our results show that agents approximate human behaviour in tasks with explicit decision rules, e.g., interruptibility (F1-scores $\in [.72, .84]$), but perform poorly when simulating internal states, e.g., emotion recognition (F1-scores $\in [.17, .43]$). Generative agents may complement participants due to their effectiveness in narrowly defined studies that rely on filtered and task-relevant data, but final validation must include human responses.

Adults with Attention-Deficit/Hyperactivity Disorder (ADHD), a neurodivergent population of approximately 15 million in the United States, face daily challenges with focus, organization, and sustaining support. Yet existing digital tools rarely reflect their lived realities. We conducted a study on the potential of Generative Artificial Intelligence (Gen AI) to support adults with ADHD. Semi-structured interviews with 17 adults showed that participants used mental health and general-purpose mobile applications (apps) to stay organized. Participants described frustration with subscription costs, reminders that quickly lost impact, and overly generic personalization. 15 participants (88%) reported using Gen AI, such as ChatGPT, to organize tasks and manage routines. We also introduced a mid-fidelity prototype combining a Gen AI chatbot with a focus timer and a body-double workspace. Participants valued the timer for creating structure and the body-double for reducing isolation, while expressing mixed views on the chatbot’s reliability. These findings highlight opportunities for more inclusive and trustworthy mobile support systems.

Algorithmically curated social media feeds frequently trigger anxiety during everyday mobile use, yet little is known about how AI can provide momentary emotional support in these contexts. We present TriggerDetector, a large language model (LLM)–powered technology probe that provides AI-generated explanations of anxiety and coping suggestions during social media use. Through an in-the-wild deployment with RedNote users (n = 241), we examine how users engage with AI-mediated support in emotionally charged moments. We find that AI-generated explanations function as resonance-based sensemaking resources, helping users feel understood while often accepting multiple plausible interpretations. At the same time, AI outputs introduce a sense of authority that shapes both interpretation and action. While this authority can enable action, it may also lead users with low self-efficacy to interpret difficulty in following advice as personal failure. We derive design implications for mitigating implicit correctness and supporting user reflection and negotiation in everyday use.

In avatar-mediated hybrid VR communication, audio-only attendees contribute verbally but are represented by static avatars, forcing others to infer attention and addressee solely from speech. To address this, we present AudioAvatar, which synthesizes an audio-only attendee’s avatar gaze based on interlocutors’ voices in conversation to convey attention without altering what the audio attendee says. We propose two synthesized gaze strategies, Speaker-Following and Context-Corrected. The former provides immediate, audio-driven attention toward the active speaker, while the latter supplements this with post-utterance corrections using an LLM to infer contextual meaning (e.g., implicit addressee) from live dialogue. A within-subject study (N=24) shows that synthesized gaze increases VR→Audio direct gaze and perceived mutual gaze from the VR attendee’s perspective, and improves social presence relative to the baseline (i.e., static avatar). Context-Corrected provides additional benefits during speaking moments, further increasing VR→Audio direct gaze and improving attentional allocation and perceived message understanding. We conclude with design implications for deploying default gazesynthesis, emphasizing context understanding for audio-only participation, while discussing asymmetry in awareness of synthesized gaze.

We present BezelBraille, a novel eyes-free, braille-inspired text entry technique for blind and low vision users. It maps the left and right columns of the braille cell to gestures performed with the left and right thumbs, using the smartphone's bottom corners as tactile landmarks. Gestures are flexible in both direction and size, can be initiated from any direction, and may be performed sequentially or simultaneously, as long as they pass through the intended zones, reducing the need for precise target selection. We evaluated BezelBraille with 12 visually impaired participants of varying braille proficiency in eight short sessions. Despite the limited training time, entry speed improved from 3 wpm in the first session to 4.2 wpm in the last, reaching 6.9 wpm without the confirmation step used for novice training. Braille proficiency improved significantly, especially among novices (from 12% to 79%). Participants rated BezelBraille as easy to learn, accurate, and suitable for regular use, underscoring its potential as both an accessible text entry method and a promising tool for braille learning.

We present a field deployment of a mobile system that enables blind and low-vision (BLV) adults to receive AI-generated descriptions of children's artwork, crafts, or picture books. We conducted a three-week deployment using a modified version of ArtInsight, a prior open-source tool, with seven families consisting of BLV relatives and their sighted children. Participants captured 135 images, 91 of which were children’s artworks; other captures included picture books, architecture, scenic landscapes, family photographs, and house furnishings. We found that ArtInsight sparked meaningful conversations by revealing new levels of detail, with BLV family members noting children felt reduced pressure to describe their art and initiated more discussion about artwork during the weeks. We reflect on the role of AI-powered creative visual interpretation systems in daily life, highlighting impacts on all mixed-ability group members---not only enabling access for BLV adults, but also supporting children’s emotions and evolving understanding of AI.

Spatial navigation for people who are blind or vision-impaired (BVI) often relies on embodied interaction, where perception and movement are closely coupled. In contrast, wearable assistive technologies increasingly deliver verbal instructions that shift navigation toward semantic information processing. To examine how these approaches shape navigation, we conducted a mixed-methods study with eight BVI participants who navigated using auditory strategies that varied in cue type and information structure. Our findings show that step-wise strategies support more accurate and consistent movement, whereas full-path strategies enable faster and more confident recall but reduce spatial precision. While speech strengthens semantic understanding and recall, it does not improve navigation accuracy. Crucially, we identify mismatches between subjective and objective measures, where users prefer and feel confident in semantic guidance, yet achieve higher accuracy with embodied, step-wise cues. We conclude that effective mobility aids should balance semantic understanding with embodied interaction to support performance and user experience.

Smartphone-based context-awareness holds significant promise for wheelchair users—from detecting everyday accessibility barriers to enabling ability-based adaptations. Such capabilities often build on passive context inference through mobile sensing, yet their accuracy hinges on how and where phones are carried and the resulting signal quality. While prior work documents phone-carrying behaviors in the general population, patterns specific to wheelchair users remain underexplored. Through a mixed-methods approach combining a survey of 91 and interviews with 15 wheelchair users, we systematically investigate their phone-carrying locations and influencing factors. Our findings reveal distinct patterns extending beyond pocket storage to diverse wheelchair-mounted accessories and around-body placements, shaped by the interplay of physical ability, wheelchair design, and everyday contexts, including social, activity, and device factors. Grounded in these findings, we articulate how carrying location can serve as a proxy for user context to enable novel context-aware experiences, and discuss design implications for developing inclusive and effective mobile context-aware applications.

"High-quality cardiopulmonary resuscitation (CPR) requires stable control of compression rhythm and depth, yet most training systems presuppose instructor mediation, repeated practice, and explanatory guidance—assumptions that do not hold in the Tibet Autonomous Region, where instruction is fragmented and learners' linguistic and educational backgrounds are heterogeneous. We present TibetCPR, a low-cost, self-guided CPR training system that pairs depth-driven electrotactile feedback with rhythm-driven visual cues within a Tibetan-language narrative. In a randomised study with 40 lay community members aged 19–56, the experimental group showed progressive minute-by-minute stabilisation of rhythm and depth across a 10-minute intervention, substantially exceeding an unguided-practice control, with gains transferring to an unscaffolded one-minute post-test. Qualitative accounts described the feedback as legible through participants' bodily action, and usability was high (SUS = 84.3). We synthesise three transferable design principles for self-guided embodied training: feedback as a calibration reference, not an immediate corrector; modality temporal granularity matched to behaviour's temporal structure; and autonomous interpretability as a deployment prerequisite, not an after-effect of usability."

In firefighting, aerial robots (a.k.a drones) are primarily used as remote sensors, leaving their potential to actively support firefighters unrealized. We propose FireFly, a mini drone designed to serve as an interactive teammate, providing on-the-go support during indoor firefighting tasks. For example, during reconnaissance, FireFly can locate victims and guide firefighters to their position. We present a proof-of-concept built with low-cost off-the-shelf hardware, combining gestural interaction and visual feedback on a low-resolution display, which was iteratively designed. We evaluate Firefly in a user study with professional and volunteer firefighters, who performed simulated operational procedure tasks. All participants completed tasks successfully, highlighting the potential of FireFly to enhance operational efficiency. We provide design implications and discuss the challenges that must be addressed for real-world deployment and adoption. This work contributes to framing drones as interactive assistants in emergency response and disaster relief operations.

Mobile sensing can advance understanding of social functioning in individuals with mental health disorders. However, prior mobile sensing studies often focus on non-clinical populations and emphasize detection rather than therapeutic response. We address these gaps through a blended intervention study combining passive smartphone sensing with cognitive behavior therapy (CBT) in individuals with serious mental illness. We integrate objective audio (e.g., conversation duration) and location (e.g., distance travelled) features with ecological momentary assessments capturing maladaptive social cognitions, including asocial beliefs, defeatist attitudes, worries, and anxious defenses. Using these measures, we categorize individuals as responders or non-responders based on post-intervention cognitive improvement. Permutation tests show that responders exhibit greater social engagement and mobility. Predictive modeling further identifies digital biomarkers; specifically, mobility features correlate with anxiety-related cognitions, while multimodal features prove more effective in predicting asocial beliefs. This work demonstrates the potential of passive sensing to objectively assess therapy response and inform personalized interventions.

Prolonged digital device use has made poor posture and musculoskeletal discomfort pervasive among knowledge workers. Existing ergonomic wearables rely solely on posture thresholds, frequently interrupting users during high-focus moments and leading to alert fatigue and abandonment. Yet posture and cognitive load are closely coupled, and most systems remain cognitively unaware. We present \system, a head-worn ergonomic assistant that detects poor posture using IMU-based head tracking and estimates task-induced cognitive load using a consumer-grade EEG headband for continuous everyday use. In a controlled lab study, \system achieves 81\% posture classification and 90.2\% task induced cognitive load estimation accuracy under leave-one-subject-out evaluation. In a preliminary real-time deployment, cognition-aware alerting reduces alert frequency by 81\%, improves perceived usability by 43\%, task performance by 25\%, and improves posture correction rate by 38\%, delivering fewer but better-timed interventions rather than merely suppressing alerts.

Despite their increased integration, wearable technologies and mobile health apps continue to face usability and human-centered design challenges, particularly in clinical contexts, where hardware constraints, interaction, and technology acceptance intersect. Existing usability frameworks often assess devices and apps in isolation, overlooking their combined use in practice. We present HEXSSYS, a flexible system that guides usability evaluation of integrated wearable-app systems across expertise levels. HEYSSYS was developed in a multi-phase process informed by a synthesis of prior usability research, expert feedback from HCI/UX practitioners, and empirical insights into clinical researchers' existing evaluation workflows. These insights guided the design of an interactive prototype that scaffolds the planning and execution of usability assessments. We user-tested the prototype to evaluate its utility and fit with real-world clinical research practices. We discuss implications for usability evaluation across domains, including product design, clinical research, and technology adoption.

Wearable devices such as smartwatches and fitness trackers are now central to health and activity monitoring, yet their accuracy remains highly sensitive to how they are worn. Suboptimal fitting (whether too loose, too tight, or misoriented) introduces artifacts that compromise data reliability and limit their usefulness. We present MalWear, a system that detects improper device fitting using on board sensors. By combining accelerometer and optical heart rate (PPG) data, MalWear achieves 95% accuracy in identifying suboptimal wearing conditions. Beyond wear state detection, we also evaluate how misfitting impacts fine grained activity recognition, suggesting its significant effect on classification performance and, by extension, underscoring the importance of considering wear state in wearable sensing pipelines. Unlike prior approaches that rely on hardware modifications or heavy post-processing to mitigate placement and wearing issues, MalWear provides a foundation for detecting wear conditions at the source, which could support future feedback mechanisms.

Finger-worn IMUs have been shown to be very accurate at detecting touch events on physical surfaces, enabling input to physically-bound XR interfaces. In this work, we describe a new XR headset method for detecting finger touch events on ad hoc surfaces, without needing the user to wear anything on their fingers or hands. Instead, our method uses a pair of high-speed headset cameras to track and triangulate the position of the user’s fingertips in 3D space. We take the first and second derivatives of this signal to estimate a finger’s velocity and acceleration. At almost 800 Hz, our system’s output is similar to a worn IMU, but with bare hands. Using this signal, we can then detect rapid and characteristic decelerations of the finger, indicative of touch.

Social robots employ a range of emotional expression aesthetics, from lifelike faces and voices, to stylised or abstract cues. However, it remains unclear how varying these aesthetics impacts user perceptions of robot affective expressions in terms of preference, clarity, appropriateness, and empathy. Furthermore, directly comparing these aesthetics is constrained by the physical and logistical limitations of traditional hardware prototyping. To address this, we conducted a comparative user evaluation of robot expression aesthetics using an open-source augmented reality (AR) platform to deploy ten face and sound styles, drawn from existing robots, research, and media, onto the zoomorphic robot Qoobo. A mixed-methods evaluation revealed that participants prioritised styles they perceived as evocative and appropriate for the robot, over clarity or recognisability, among other insights. We present design recommendations for affective faces and sounds for zoomorphic robots, derived from our findings, and discuss the implications of AR for rapid robot prototyping and evaluation.

Zoos aim to provide engaging learning experiences for visitors while supporting conservation efforts. A common challenge in zoos is providing educational information in an engaging manner to diverse target audiences. We examine how emulated interactions with animal avatars, powered by large language models, can be designed to support visitor engagement at zoos. We present Zoodex, a mobile application that allows zoo visitors to discover animals and ‘chat’ with digital avatars representing zoo animals. We evaluated Zoodex through an in-situ study with twenty participants aged 12–14. Our findings indicate that the application facilitates engagement with zoo exhibits and social interactions among teenagers. We reflect on the use of emulated interaction to support experiences that bridge the analogue and digital worlds and discuss considerations for representing animals through conversational AI.

Traditional Chinese scrolls possess extreme information density that challenges human perception. We investigated two character-based VR exploration paradigms---Open vs. Linear Mode---using Qiu Ying's Along the River During the Qingming Festival. A user study (N=34) revealed that both modes significantly improved historical knowledge, with no significant differences in immersion or motivation. Crucially, qualitative findings indicated a strong user preference for embodied experiential activities (e.g., pouring wine) over passive information consumption (e.g., watching videos or NPC dialogues). This suggests that action-oriented tasks are primary drivers of engagement in virtual heritage. Our work contributes design insights for digitizing large-scale scrolls, proposing that balancing navigational autonomy with action-centric learning transforms static digital artifacts into interactive spaces. This study provides a reproducible reference for creating immersive historical narratives and managing cognitive load in expansive virtual environments.

Magnetic resonance imaging (MRI) procedures can be stressful and unfamiliar experiences, motivating the development of preparatory tools that improve comfort and expectation management. Virtual reality (VR) has shown promise for MRI anxiety reduction, but deployment is limited by the cost and availability of head-mounted displays (HMDs). Smartphone-based approaches offer a more accessible alternative with reduced immersion and interactivity. We report two exploratory studies examining how immersion and delivery modality influence MRI-related anxiety and user experience. One study evaluated a HMD-based application with 2D video, 360° video, and interactive VR, while the other evaluated a smartphone-based application with equivalent 2D and 360° content. Across both studies, we combined self-report and physiological measures. Anxiety decreased following exposure scenarios, while perceived presence increased with immersion level. Physiological responses varied and were not consistently aligned with self-report measures, highlighting trade-offs between immersion and accessibility in MRI preparation design.

Virtual museum tours are increasingly used to broaden access to cultural heritage, yet it remains unclear how closely they reproduce the experience of visiting the same exhibition in person. This initial exploration study directly compares how visitors navigate and experience a physical museum exhibition and its image-based 360-degree virtual tour (VT). Thirty-two participants completed either an on-site walkthrough or the VT, followed by standardized questionnaires and custom measures focusing on spatial clarity and navigation behavior. Behavioral logs captured viewing time, walking time, room transitions, and exhibit coverage across both conditions. Overall results showed that experiential ratings were mostly similar across conditions, suggesting that VTs can elicit comparable levels of engagement and satisfaction. Behavioral measures likewise showed broad similarity in total visit time, viewing time, and the number of exhibits visited. However, clear differences emerged in navigation patterns, revealing challenges for spatial clarity and intuitive navigation in image-based 360-degree virtual tours.

Extended Reality could improve in-transit entertainment, but must balance immersion with situational awareness. Prior ARTV Ornamentation work suggests content-themed augmentation can approach VR-like immersion while preserving awareness, but evidence comes from simulated AR-in-VR studies. We evaluated ARTV Ornamentation with 19 participants across six conditions spanning pass-through AR, VR-simulated AR, VR Cinema, and VR Videosphere, conducted in a public-facing museum subway carriage and adjacent lab. We collected subjective experience measures, a stop-announcement awareness task, rankings, usage-intent ratings, and interviews. ARTV Ornamentation improved hedonic and engagement-related aspects of viewing, but current pass-through implementation introduced pragmatic and usability costs, including texture instability and rendering artefacts. AR-based viewing was preferred over fully immersive VR for everyday journeys because it preserved flexible attention to passengers, belongings, and journey cues. VR simulation reproduced broad design trends but missed bystander behaviour, ambient noise, pass-through artefacts, and the perceptual accountability of real-world augmentation.

Absolute touchscreen-based distal pointing on large high-resolution displays offers speed but often suffers from reduced selection accuracy. We investigate whether this limitation can be mitigated using a target expansion technique. Our approach is based on a Voronoi partitioning with a transient display of Voronoi cells. We conducted a controlled experiment evaluating this approach for absolute touchscreen pointing. In this study, Voronoi cell visibility and target highlighting were systematically combined to define four pointing techniques used by participants. The results indicate that brief exposure to the Voronoi cells at the onset of targeting, combined with simple target highlighting (EVTarget pointing technique), resulted in high selection accuracy (Mdn = 0.969 [0.951–0.988]) and competitive selection times (Mdn = 2.868 [2.614–3.179] s). These findings demonstrate that ephemeral visual feedback of Voronoi cells preserves the speed of absolute mapping while maintaining high selection accuracy.

Recalling story characters and how they relate is essential when reading fiction. Yet, on mobile devices, frequent interruptions, small screens, and linear navigation make it difficult to recover such context without disrupting the reading flow. Although contemporary reading applications can dynamically augment text, they provide little support for accessing character information in situ during reading. We present four reading interface prototypes for accessing spoiler-aware character information: an out-of-page list based on Kindle’s X-Ray, an in-page textual pop-up, a page-overlay relationship diagram, and an in-line augmentation. In a within-subjects study with 109 participants, the pop-up emerged as the preferred interface, offering the best trade-off between ease of access, visual clutter, and reader control. We discuss design implications for intelligent reading interfaces, arguing that effective mobile support should surface context on demand while remaining visually lightweight and easy to disengage from, preserving the narrative flow.

We propose hand-to-nose gestures as a discreet input method for smartglasses in everyday settings. Existing alternatives face different barriers. Voice control can be socially disruptive; touch interaction with eyeglass frames are noticeable in public; and gesture techniques evaluated in laboratories often leave the trade-off between recall and false positives unresolved, which is critical because non-gesture periods outnumber intentional inputs in everyday use. We enhance hand-to-nose recognition with temporal stabilization, null class detection, and scenario-specific training, then evaluate the system across five everyday scenarios and assess social noticeability from an observer's perspective. The proposed system achieves 87.9\% recall with 0.95 false positives per hour, and hand-to-nose gestures are no more noticeable than conventional methods. Our results suggest that gesture timing and repetition matter more for noticeability than gesture type alone. Together, these findings suggest that hand-to-nose gestures can serve as an additional smartglasses input method, supporting reliable recognition with low noticeability in everyday settings.

Individual perception of colors is an important, often overlooked, issue when recreating or finding colors. Few studies have tackled this problem, and fewer studies have examined recreating a physical color onto a digital device. The conversion of color from physical to digital, or vise versa, is complicated. Digital displays create colors in a manner physically different from, say, prints on paper or colors on textiles and do not necessarily represent these colors correctly. To better understand this difference and overcome potential mismatches, we conducted a study with 50 participants where they were tasked to recreate 22 physical colors, twice, on an iPhone to investigate how close colors can be recreated on the digital device from the ground truth of the physical colors. Our prior assumption was that users would recreate colors normally distributed around the ground truth, but our study showed that participants generally recreated the digital colors lighter than the physical counterparts.

Freehand pointing is a fundamental gesture commonly used for cursorless interactions. Prior work in HCI often elicits the same pointing behaviour—facing the target with an outstretched dominant arm and index finger. However, freehand pointing outside of HCI shows more variability across hand pose, usage, and coordination with gaze. To understand what variability exists and how it affects pointing performance, we collected data (N?=?23) using a hybrid motion capture system. To elicit a wide variety of pointing behaviours, we included different levels of user effort and attention, as well as the widest range of target placements studied. We systematically characterised and described three distinct pointing behaviours, each with three different traits, ranging from accurate stereotypical pointing observed in prior works to more casual hip fire-style pointing. Our analysis demonstrates how different pointing behaviours affect pointing performance and highlights their importance when designing interactive systems for more naturalistic freehand pointing.

Electrotactile haptics is a novel form of skin stimulation. It is versatile and allows manipulation of many design parameters, broadening the range of haptic experiences. However, how parameter changes influence qualitative experiences is not well understood. We systematically investigated (n = 45) the influence of single parameter changes (e.g., pulse width, frequency), capturing the breadth of experiences elicited. We compared differences between calibration and ordering task performance, first between left and right hands, then between single and both hands simultaneously, all with an open hand and while holding an object, fully exploring the design space. Results showed the wide range of qualitative experience, from concrete to abstract. Calibration should be performed on both hands simultaneously to ensure perception, as participants preferred lower pulse width values for simulation on both hands as compared to each hand alone. The results present the first detailed exploration into the qualitative aspects of electrotactile stimulation across the hands.

Voice over IP (VoIP) has existed since the mid-1990s, and guidelines from the International Telecommunication Union (ITU) recommend a maximum one-way delay of 400 ms for acceptable call quality. However, this guideline was established without explicitly considering users' awareness and does not consider how mobile interfaces can shape users’ understanding of network behavior. We design a smartphone interface that provides real-time visual feedback of speaker’s audio status, making delay and packet loss visible during calls. Using this interface, we study two research questions: (RQ1) whether users prefer higher reliability over lower delay under poor mobile network conditions, and (RQ2) whether appropriate notification increases tolerance for delays beyond the ITU threshold. User study results show that all participants (61/61) preferred higher reliability over lower delay in degraded conditions. Across personal and professional mobile contexts, users tolerated one-way delays exceeding 1.2 seconds when informed in advance. These findings highlight how interface-level awareness can reshape delay tolerance for VoIP communication.

Mobile authentication is both an interaction and security problem. Users encounter sign-in flows on a small touch-screen device, with constrained input and platform-mediated affordances. This paper examines how iOS apps authenticate users, and how developers reason about it, through a mixed-methods design combining a feature analysis of 660 apps across 22 App Store categories with a developer questionnaire. It was found that email–password remains dominant, NIST-aligned features are only partially implemented, password configurations do not differ between high- and low-risk apps, and lower-friction primitives the platform already exposes (Sign in with Apple, biometrics, passkeys, autofilled SMS codes) are under-used. Developers report following platform defaults, perceived norms, and friction concerns rather than formal risk assessment. This is characterised as a templated mobile sign-up UX, a security plateau at the interaction layer, and argue for platform-level defaults, IDaaS templates, and tooling that make standards-aligned mobile sign-in the path of least resistance.

Prior work on mobile notifications primarily investigates specific features introduced by researchers, observing how users interact with them and evaluating their effectiveness in supporting notification delivery and management. Less is known about what features and capabilities users themselves desire from future intelligent notification systems (FINS). We conducted interviews and design workshops to understand these desires. Our findings reveal that participants envisioned FINS as proactive agents extending beyond passive notification delivery. They wanted FINS to manage attention by surfacing actionable takeaways and progressively disclosing notification details, assist follow-up actions, and associate notifications with related ones and relevant external context. Participants also imagined FINS evolving into adaptive companions that infer personal context to anticipate needs, reinforce self-defined goals to support personal development, and connect users with relevant trends beyond their followed sources. Finally, participants desired distributing notification awareness through co-present contacts and objects embedded in routines to time and contextualize delivery.

Smartphones are widely used even in emergent contexts, where devices are often shared or ageing and where mistakes can be costly and difficult to recover from. Yet mobile user experience frequently focuses on interface usability and individual learning, overlooking how everyday interaction is shaped by social and economic constraints. We present a qualitative study of smartphone users in emergent contexts operating under conditions of financial vulnerability and shared device use. Participant accounts show that user experience is organized around risk-oriented practices rather than feature exploration or progressive mastery. Participants prioritize avoiding severe harms, particularly financial loss, while negotiating privacy and usability trade-offs. Delegated configuration, conservative interaction, selective platform trust, feature avoidance, and non-technical interpretations of breakdowns emerge as strategies for maintaining stability. We discuss implications for MobileHCI design and evaluation, arguing for approaches that focus on recoverability, legibility of consequences, and support for shared and assisted use.

Voice assistants are promising user interfaces for driving as they reduce the need for manual interaction. Current systems typically remain reactive, though, requiring explicit input. Proactive assistants could address this, yet we have little knowledge of how they affect driving and social contexts where drivers and passengers engage with the system. Therefore, we derived conversational scenarios and assessed proactive behavior through an online survey (N=93), followed by a driving simulator study (N=36). Results demonstrate that situational context plays an important role in user experience and user acceptance. If not considered, participants may consider proactive behavior as intrusive or "creepy,'' especially when it interrupts private or non-driving conversations. Certain topics reveal differences between drivers and passengers, further informing expectations in a multi-user setting. Our findings explore both the promise of proactive in-car voice assistants and the need to evaluate contextual design guidelines to balance helpfulness, intrusiveness, and multi-user interaction.

As smartphones become integral to daily life, researchers have sought to identify when the use becomes problematic. Previous studies have operationalized problematic smartphone usage (PSU) from either an intention or a behavior perspective. Both risk delivering interventions not welcomed by users. We propose a novel approach to operationalizing PSU as the intention–behavior gap (IBG). We collected self-reported data on intentions to stop phone usage, alongside usage behavior data, from 37 participants over two weeks. We calculated IBG, examined effects of demographic and contextual variables, and developed machine learning models to predict IBG in real time. We found that IBG was explained by gender, time, app, and input interactions, among other factors. Intention was predicted most accurately with only personal data, whereas behavior and IBG were predicted most accurately with both personal and global data. Our findings can inform the design of future intervention tools optimized for timing and adaptive intensity.

Mixed reality (MR) 3D storybooks, in which virtual 3D storybooks are embedded in the real-world environment, have demonstrated considerable potential for supporting students’ learning. However, creating such 3D stories remains time-consuming and technically demanding, limiting their accessibility for non-expert users. Recent advances in AI have made text-to-3D content creation increasingly feasible, yet it remains unclear how AI can support accessible 3D story authoring and whether AI-generated 3D stories can serve as usable materials in children’s story-based learning. To address these questions, we first conducted a two-day workshop to examine where novice users break down when creating 3D stories from text without AI, and how their AI-assisted workflows differ from those of domain experts. Informed by the design insights derived from the workshop, we developed 3DStoryWeaver, a proof-of-concept mobile-to-headset pipeline for AI-assisted 3D story authoring. Our Study 1 with 16 novice users shows that 3DStoryWeaver enables novice users to create satisfactory 3D stories within an acceptable operation time and waiting time. Study 2 with 22 children further showed that the generated 3D stories could function as supplementary story-based learning materials that enhance interest, support retention, and reduce mental demand and effort, while maintaining comparable comprehension to traditional paper-based reading. We conclude by discussing human roles in AI-assisted story authoring and using 3D stories generated by AI-assisted pipeline in story-based learning.

Hand-based input is widely used for interaction in Extended Reality (XR). While the most established input techniques, direct-touch and pinch-to-click, have been extensively explored and combined with gaze- and hand-ray casting, previous research has proposed more continuous pinching gestures to set a greater range of values by adjusting the distance between the thumb and index finger. While promising, prior studies have not systematically assessed the factors contributing to these input techniques and their impact on usability and efficiency relative to established techniques. We address this gap by systematically combining and comparing selection (direct, by gaze, and by ray-casting) and manipulation methods (touch, pinch-to-click, and continuous pinching) to understand effective hand-based input techniques in XR. To this end, we transfer the concept of continuous thumb-index pinching to XR and present the results of a controlled experiment comparing it with established hand-based XR input.

Amid a global “crisis” in abortion care, marked by expanding legal restrictions, persistent stigma, and dwindling international aid, technologies become central yet precarious infrastructures through which people navigate, negotiate, and endure barriers to reproductive care. Yet, HCI infrastructure scholarship often discusses abortion as a taboo, controversial, or purely political example. This framing tends to eclipse abortion as a lived, embodied, and relational experience. To bridge this gap, we engaged in a polyethnographic walkthrough of abortion- related mobile applications. We collectively explored what applications encode about abortion when examined as part of larger socio-technical and socio-political infrastructures. We find that abortion is ‘designed out’ of reproductive health applications through omission, over-simplification, and misrepresentation. Taking an infrastructure lens, we show how abortion applications disembed abortion from the user’s relational context, as well as larger transnational and historical systems. To conclude, we offer guidance on grounding technical infrastructures to support local community infrastructure.

We present Airveil, a haptic interface that leverages air cushion effects to dynamically modulate the tangential friction between the fingertip and a surface. By recontextualizing industrial aerostatic lubrication, Airveil regulates airflow through a micro-patterned orifice array to create a pressurized air film. This mechanism enables subtractive friction modulation—physically reducing the coefficient of friction to render a dynamic range of tactile sensations from rough to smooth, while preserving direct finger contact. Airveil’s thin and conformal form factor makes it highly adaptable to ungrounded handheld controllers and curved physical 3D props. Technical evaluations characterize the system's wide friction range and geometric adaptability. User studies confirm that Airveil significantly enhances tactile immersion in tangible and mixed-reality scenarios. Results demonstrate that users can reliably distinguish complex material properties and perceive a heightened sense of realism, establishing aerostatic lubrication as a lightweight, high-fidelity approach for enriching materiality in mobile and immersive environments.

Behavioral variation across students from different academic years may have implications for the design of student-focused mHealth interventions: should intervention frameworks rely on a unified model intended to generalize across the student population, or should they be tailored to students at different stages of their academic journey? To address this question, we investigate whether passively sensed behavioral patterns differ across students from different academic years using smartphone sensing data collected over a two-week period from 58 undergraduate students across three academic years. Our results show that behavioral features such as physical activity and overall smartphone interaction were broadly similar across academic years, whereas limited differences were observed in contextual mobility, communication behavior, and application-level phone use. These findings suggest that academic year is associated with variation in selected behavioral patterns and indicate that student mental health interventions may benefit from a common intervention framework.

Reading fiction is a generative process in which readers transform text into personal mental imagery. Generative AI and mixed-reality systems can increase immersion by surrounding text with visuals, yet detailed depictions can over-define settings, contradict imagery, and pull attention from the page. We target shared acceptability, examining how ambient backdrops that vary in visual commitment and inspectability support an imagination-preserving, reading-first experience. In a formative workshop (n=8), readers externalized scenes with generative tools; they prioritized ambience and lighting, and found stylistic cues polarizing unless specified. Informed by these findings, we developed a confidence-aware GenAI pipeline that extracts passage-level environment attributes with evidence-based confidence scores, composes stagecraft-style backdrops that abstract low-confidence aspects, and renders 360° VR scenes. We explore two ambient designs: Gaze-Faded and Persistent-Detail. In a within-subjects VR study (n=18) comparing Passthrough (No backdrop), Gaze-Faded, and Persistent-Detail, participants preferred ambient backdrops for comprehension, presence, and engagement; Interviews and gaze data reveal an assist-versus-contradict dynamic.

Word prediction models are widely integrated into virtual keyboards and are expected to improve text entry performance. However, several studies show that such benefits are not consistently observed, and the reasons behind this discrepancy remain unclear. This paper examines how users visually attend to and interact with word predictions presented as a list of three suggestions during text entry, and how these behaviors relate to performance. We conducted a controlled text-copying study comparing a keyboard augmented with word prediction to a baseline keyboard without prediction, combining keystroke logging with eye-tracking data. Results indicate that users primarily fixated and selected the first two predictions, while the third prediction was rarely attended to or used. We also observed a relationship between visual attention to the word prediction list and prediction usage. These findings question the utility of maintaining three word predictions and provide design insights for more effective text entry systems.

Digital fitness coaches are gaining popularity as they become more sophisticated owing to the use of Large-Language Models (LLMs). However, it remains unclear which design qualities to consider when creating LLM-powered virtual coaches that effectively benefit users. To address this, we created KineSenseAR, an augmented reality (AR) coach for fitness exercises to support individuals during training. The system demonstrates exercises and provides personalized feedback. We evaluated the system in a within-subject user study (N=27) where participants experienced different versions of the coach. The versions differed in verbal behavior (pre-scripted and LLM-based dialogue) and motion awareness (including or excluding motion sensor data in feedback generation). Our results indicate that participants preferred the motion-aware coach, which used sensor data to personalize the feedback, because it provided the most details.This work provides insights into design rationales behind LLM-powered, motion-aware fitness coaches to support amateurs with their fitness routine

Life storytelling helps older adults reflect on experiences, strengthen identity, and connect across generations, yet their meaningful engagement is often constrained by limited opportunities and rigid digital tools. Existing digital storytelling systems often rely on rigid templates and lack emotional sensitivity, limiting their effectiveness. Therefore, we present LifeLoom, a mobile emotionally adaptive digital life storytelling system for older adults, powered by large language models. LifeLoom combines a structured topic framework with adaptive prompting strategies that dynamically adjust disclosure depth based on users’ emotional engagement. Through multi-turn conversations, the system guides reflection and generates first-person narrative summaries to support memory reconstruction and self-expression. We deployed LifeLoom in a twelve-day in-the-wild field study with older adults. Our findings show the potential of emotionally responsive LLM-powered conversations on mobile devices to help older adults’ digital storytelling and offer design insights for future mobile AI-assisted narrative tools for aging populations.

Mental health concerns are increasing worldwide, highlighting the need for interventions that support everyday emotional well-being. Prior work has demonstrated the potential of wearable and mobile technologies to deliver data-driven interventions. However, developing effective data-driven systems requires access to emotion data that captures individuals’ emotional variability and change in everyday contexts. Existing approaches to data collection largely rely on frequent, prescheduled prompts and predefined scales or questionnaires. These methods often fail to account for participants’ availability, agency, or the complexity of their emotional experiences, resulting in shallow, context-poor data. In this paper, we present a feasibility study of a participant-centric, multimodal emotion-annotation application designed around users’ emotional intensity and availability. Our findings show how multimodal emotion logging can shape participants’ experiences and data logging behaviors, and demonstrate its potential to support the collection of richer, more nuanced emotion data.

As Large Language Models (LLMs) become increasingly integrated into daily routines, understanding how users interact with these systems is crucial for effective human-AI collaboration. This work investigates keystroke dynamics as a behavioral measure of user mental effort and perceived output usefulness in human-LLM interaction. We conducted a user study (N = 36) to examine how task difficulty (easy vs. hard) and device type (desktop vs. mobile) influence typing behavior and workload (NASA-TLX) during interactions. Our results indicate that hard tasks led to significantly more keystrokes, slower typing, increased pauses, and higher self-reported workload. Device type had weaker effects, with mobile use slightly reducing input length and typing speed. While keystrokes captured differences in cognitive effort, they did not predict perceived LLM output usefulness. These findings highlight the potential of keystroke dynamics as real-time indicators of cognitive effort during LLM prompting, while also showing their limitations in capturing perceived collaboration success.

The physical world quietly encodes human activity and object context through microvibrations produced by everyday interactions, from tapping and rubbing to swinging and striking. While such vibrations have been explored as a sustainable power source for ubiquitous sensing, interaction-driven harvesting remains underutilized due to limited awareness of vibration-based sensing opportunities, fabrication complexity, and uncertainty about whether harvested energy can support practical sensing. We introduce VIB3D, a design and fabrication approach for harvesting everyday microvibrations as distinct profiles for context-aware sensing. VIB3D combines a scanner and parametric design framework that enables non-experts to customize low-cost FDM-printed harvesters from off-the-shelf materials, based on energy needs, spatial constraints, and target interactions. By tuning batteryless sensing to specific objects and contexts, VIB3D enables everyday objects to become sustainable sensing sites for mobile and ubiquitous computing.

While research on menstrual health applications (MHAs) has focused predominantly on active users, the considerations of those who choose not to adopt or discontinue use remain underexplored. We conducted semi-structured interviews with 33 participants who had initial interest in MHAs but ultimately did not adopt or discontinued use. Our analysis identifies three non-user groups: non-downloaders, non-activators, and discontinuers, and organizes their considerations into five categories: functionality, data privacy and security, utility, trust, and design. We find that engagement stage shapes how shared considerations are reasoned about and used to justify disengagement, with rationales evolving from anticipatory to experience-based as engagement deepens. We derive design implications for supporting user decision-making around MHAs.

Hand microgestures are becoming increasingly popular in the context of wearable devices. Despite this growing interest, there is limited understanding of how to design effective help interfaces for microgestures, which is critical for their widespread adoption. In a between-subject experiment with 36 participants, we compared 3 types of help interfaces. The experiment tested a subset of 24 microgestures (12 taps and 12 swipes) associated with abstract commands. Users discovered the help interface by themselves and then passed learning, test and retention blocks. Based on the results, we discuss the advantages and drawbacks of each help interface tested and provide design implications. We also evaluate the mean time needed to perform a microgesture (with or without a help interface) given a specific command and propose a first approximation of the confusion between two microgestures. These findings contribute to the democratization of microgesture-based interaction.