User experience quality(UEQ) plays a critical role in multilingual intelligent voice assistant(VA), where real-time feedback directly affect interaction experience. However, current UEQ evaluation methods primarily depend on surveys, lack automation and timeliness, posing limitations for dynamic and user-centered adaptation. We propose a non-intrusive real-time system, UEQManager, for automated UEQ recognition and management. First, UEQManager incorporate interpretable deep learning models to predict UEQ in seven subdimensions. Second, leveraging LLM and expert group, we design adaptive interaction interface based on recognized UEQ states. Third, we implement integration system capable of real-time, non-intrusive UEQ recognition and management. We validate effectiveness of UEQManager through a user testing experiment. Results reveal that UEQManager significantly outperformed baseline, yielding an average UEQ improvement of 27.29% over baseline and demonstrating statistical significance across all subdimensions. This work contributes a proof of concept HCI system that translates webcam gaze cues into adaptive design decisions for multilingual VAs and illustrates how designers can couple interpretable sensing with proactive interaction design.
Voice assistants (VAs) are typically evaluated through task performance metrics and self-report questionnaires, but people’s voices themselves carry rich paralinguistic cues that reveal affect, effort, and interaction breakdowns. We present a within-subjects study (N=49) that systematically compared three VA personas across three usage scenarios to investigate whether speech-derived audio features can serve as a proxy for user experience (UX). Participants’ speech was analyzed for temporal, spectral, and linguistic markers, alongside standardized UX measures, brief mood and stress ratings, and a post-study questionnaire. We found correlations between specific speech features and self-reported satisfaction and experience. Furthermore, a machine learning model trained on speech features achieved promising accuracy in classifying UX levels, indicating that this might be a reasonable alternative to self-report instruments. Our findings establish speech as a viable, real-time signal for implicitly measuring UX and point toward adaptive VUIs that respond dynamically to emotional and usability-related vocal cues.
Eye-tracking and questionnaires are typically treated as separate methods for measuring usability and user experience (UX). Recent studies show that machine learning models trained solely on eye movements can predict pragmatic and hedonic quality ratings. Building on this, this study examines which gaze patterns predict usability and UX and whether models can generalize across stimuli. Five models were trained on eye movements from 121 users browsing six websites. A feature-importance analysis revealed that saccadic patterns, such as regressions and successive forward movements, are more associated with UX, whereas longer consecutive saccades are indicative of usability. When trained separately for each website, the best-performing models achieve Matthews Correlation Coefficient scores of 0.751 and 0.780, with only small negative effect sizes on holdout data. Trained across websites, holdout scores dropped to 0.196 for usability and 0.338 for UX, suggesting that cross-stimuli generalizability is limited and, at best, achievable for hedonic interaction aspects.
Digital health interventions in the Global South often rely on transmission models, assuming that delivering correct medical information yields better care. We challenge this view through an analysis of a multi-platform social media intervention for Community-Based stroke Rehabilitation(CBR) in rural Thailand. Following a collaborative development process with clinicians and a deployment across roughly 2,000 villages, we interviewed 28 caregivers, patients, and health volunteers. We found that communities appropriated the technology in unexpected ways, such as using videos as social objects to manage family hierarchies, integrating rehabilitation into Buddhist merit-making, and prioritising offline peer networks over online discussion. Our findings suggest that effective Human-Computer-Interaction (HCI) for digital health in Low- and Middle-Income Countries (LMICs) should look beyond engagement metrics to support the appropriation of digital tools, enabling communities to integrate clinical protocols into their existing cultural and relational fabrics.
The Wizard-of-Oz (WoZ) method has long been a core prototyping technique in Human-Computer Interaction (HCI), in which users interact with systems that seem autonomous but are actually controlled by hidden human operators. Advances in interactive technologies have expanded the landscape of future system behaviors, broadening both where and how WoZ is used. However, as more envisioned behaviors become technically feasible, the distinction between engineering a system and simulating an interaction becomes blurred, making it essential to clarify when and why to employ wizarding. This paper presents the first systematic review of WoZ in HCI, drawing on 194 papers from SIGCHI venues to identify ten application domains, five wizard control types, eight motivations, and five categories of concerns. Building on these findings, we propose a reciprocal evolution framework that interprets how technology and wizarding shape each other, and derive guidelines for the rigorous application of WoZ. We further illustrate the framework through emerging prototyping practices with Large Language Models (LLMs).
Climate anxiety often makes the future feel distant and overwhelming. We present Branching Foresight, a self-guided scenario tool that helps individuals explore climate futures through AI-generated scenarios framed as positive, neutral, and challenging trajectories. We position the system as a coping and reflective support tool, where each scenario is enriched with imagery, a conversational coach, and reflection cards. In a single-session study (N = 30), we measured climate transilience, climate-related emotions, creativity support, and user experience, complemented by telemetry and interviews. Results suggest within-subject pre-post improvements in transilience and reductions in negative emotions, with participants reporting strong creativity support and positive experience. As this evidence is single-session and non-comparative, it should be interpreted as initial indications of short-term change rather than superiority over alternative approaches. Qualitative feedback highlighted how the scenarios, imagery, and dialogue made climate futures more concrete and actionable. Our contributions are: (1) an interaction prototype for AI-mediated scenario exploration, (2) initial within-condition pre-post evidence of short-term improvements in perceived coping capacity and transilience during a self-guided session, and (3) design implications for creativity-supportive human-AI interaction in scenario planning.
The Wizard of Oz (WoZ) method is a common and popular approach for simulating interactive systems in Human-Computer Interaction. Running such studies is demanding for researchers because the human wizard must manage human–agent interactions in real time while keeping participants safe and the interaction natural. Many WoZ systems struggle to reproduce complex agent behaviours without minimal delays or heavy workload for the moderator. We introduce AI of Oz, a framework that uses large language models to support researchers by monitoring ongoing interactions, detecting sensitive moments, and suggesting contextually appropriate responses. In a study with 20 HCI-related researchers, the system improved participants’ ability to manage interactions and maintain control compared to a version without AI support. We outline implications for WoZ research and note current limitations and future directions.