Extended abstracts on Human factors in computing systems. CHI Conference最新文献

Advances in multimodal large language models enable automatic video narration and question answering (VQA), offering scalable alternatives to labor-intensive, human-authored audio descriptions (ADs) for blind and low vision (BLV) viewers. However, prior AI-driven AD systems rarely adapt to the diverse needs and preferences of BLV individuals across videos and are typically evaluated in controlled, single-session settings. We present ViDscribe, a web-based platform that integrates AI-generated ADs with six types of user customizations and a conversational VQA interface for YouTube videos. Through a longitudinal, in-the-wild study with eight BLV participants, we examine how users engage with customization and VQA features over time. Our results show sustained engagement with both features and that customized ADs improve effectiveness, enjoyment, and immersion compared to default ADs, highlighting the value of personalized, interactive video access for BLV users.

Patient portals are a convenient way for patients to communicate with healthcare providers and manage care. Active portal engagement is associated with enhanced knowledge, increased self-efficacy, and improved symptom management. However, gaps in portal use persist, particularly among older adults. We conducted a random-effects meta-analysis of studies published between 2005 and 2025, with publication year as a moderator. From 1,487 unique records, 18 studies were included in the meta-analysis. We found that an estimated 55% of older adults have used a patient portal at least once, and 49% were continuing users, although these rates varied widely across studies. Publication year was a significant moderator for ever having used a patient portal, suggesting increasing uptake over time, but was not significantly associated with continued use. These findings suggest that improving onboarding alone may not be sufficient; ongoing tech support is needed to sustain long-term patient portal engagement among older adults.

We describe the design and initial evaluation of a video-based explainable interface (XUI) for personal protective equipment (PPE) compliance monitoring in a dynamic medical setting of trauma resuscitation. The system integrates video input, a computer vision model for detecting PPE status, backend processing, and the video-based XUI. The initial UI underwent a major redesign and was evaluated with five nurse educators. Participants interacted with the XUI features, provided feedback on visual elements, and compared and ranked alternate design options. Using the results, we determined visuals for alert symbols, PPE status, and compliance metrics that support rapid PPE compliance monitoring among care providers. While standard computer vision outputs, such as bounding boxes, improve explainability in object detection models, we suggest moving beyond these model representations to enhance system usability in practice.

Understanding mobile Health (mHealth) user engagement patterns and its association to adherence in chronic disease populations is critical for data quality and clinical outcomes. Yet, this remains unexplored and a missed opportunity, given the unique value of mHealth data for traditionally under-documented conditions.

Methods: We analyzed 13,997 total days of data from 131 female participants with chronic pelvic pain disorders (CPPDs) and 72 healthy controls. Participants tracked daily symptoms using a research App for 90 days. Using mixed-effects regression, we investigated predictors of adherence and consistency in App engagement, with moderators of CPPD burden and temporal accessibility.

Results: Later-day engagement was the strongest predictor of user consistency, independent of pain interference or CPPD. Habitual evening tracking was associated with better adherence for those with CPPD vs healthy controls. Consistency was the strongest predictor of adherence.

Conclusion: Prioritizing habit and preference over fixed-time assessments could improve temporal accessibility and data completeness.

Recommendation systems for nutrition management have become increasingly popular. Current solutions focus on optimizing for users taste preferences, often neglecting practical factors including personal priorities, existing plans, and situational constraints. We propose similarity-based meal recommendation - a novel approach that considers users existing meal plans and recommends similar meals that also align with their nutrition goals. This approach is designed to support meal decisions in on-the-spot, particularly when individuals are in constrained settings with limited options for adjustment (e.g. a cafeteria). We conducted a feasibility study with 15 participants to evaluate the utility of similarity-based recommendations in constrained settings and refine the system design. Findings show (1) a novel approach for similarity-based meal recommendations, (2) insights into the dimensions of similarity influencing actionability across diverse, and (3) opportunities to enhance user control over the system's design. The study contributes to the advancement of human-centered AI technologies in nutrition.

Dementia is a global public health concern, with approximately 55 million individuals worldwide currently living with it. Individual cognitive stimulation therapy (iCST) has been shown to improve quality of life for persons living with dementia (PLwDs). However, providing iCST at scale remains a serious challenge, specifically as it can add considerable burden on care partners. This project focuses on a voice assistant (VA) to support care partners in delivering iCST. We developed a VA prototype and conducted a qualitative study with care partners (N=5). Our preliminary findings show that using a VA to deliver iCST is feasible and acceptable. We have also identified design requirements for the VA to effectively provide iCST, including need for personalization reflecting dementia severity and individual interests, collaboration between care partners and PLwDs, and accessible interactions to minimize frustration and distress. These findings can inform the future design of inclusive and accessible VAs.

Video games are becoming increasingly accessible, but there are still scant resources for gamers with disabilities to reliably discover and evaluate which games are accessible to them. To remedy this, I designed OURCADE: a social simulation role-playing game that utilizes player engagement to produce a wiki-based resource for game accessibility information. This design leverages existing insights around game accessibility, citizen science, and video game communities, while also exploring new questions around the potential of social video games as a medium for answering complex game accessibility questions.

MindScape aims to study the benefits of integrating time series behavioral patterns (e.g., conversational engagement, sleep, location) with Large Language Models (LLMs) to create a new form of contextual AI journaling, promoting self-reflection and well-being. We argue that integrating behavioral sensing in LLMs will likely lead to a new frontier in AI. In this Late-Breaking Work paper, we discuss the MindScape contextual journal App design that uses LLMs and behavioral sensing to generate contextual and personalized journaling prompts crafted to encourage self-reflection and emotional development. We also discuss the MindScape study of college students based on a preliminary user study and our upcoming study to assess the effectiveness of contextual AI journaling in promoting better well-being on college campuses. MindScape represents a new application class that embeds behavioral intelligence in AI.

With advances in AI, computer vision, and interface understanding, there is the potential to offload much of the work currently spent by companies' developers in making products accessible. There is also the potential to move our major accessibility approach from an 'inclusively-designed-products-plus-AT focus to a 'universal-interface-transformer focus. This would be a major reversal of approach and have significant ramifications for legislation, regulation, and the established large-scale accessibility industries that have grown up around them. Such a disruption would require concrete evidence that such a change would, in fact, be better for people with disabilities. It would also require a path from the former to the latter. This paper presents the case for such a shift, some of the benefits and ramifications, and the developments necessary to make the shift. It also outlines a hybrid approach between inclusive design and bespoke custom interfaces.

Although clinical training in implicit bias is essential for healthcare equity, major gaps remain both for effective educational strategies and for tools to help identify implicit bias. To understand the perspectives of clinicians on the design of these needed strategies and tools, we conducted 21 semi-structured interviews with primary care clinicians about their perspectives and design recommendations for tools to improve patient-centered communication and to help mitigate implicit bias. Participants generated three types of solutions to improve communication and raise awareness of implicit bias: digital nudges, guided reflection, and data-driven feedback. Given the nuance of implicit bias communication feedback, these findings illustrate innovative design directions for communication training strategies that clinicians may find acceptable. Improving communication skills through individual feedback designed by clinicians for clinicians has the potential to improve healthcare equity.