As artificial intelligence (AI) products diversify, it remains unclear whether trust mechanisms generalize across systems or vary by users’ construal of AI. In a video-based vignette study (N = 242), participants evaluated 11 AI systems ranging from disembodied assistants to highly anthropomorphic agents. Exploratory factor analysis and hierarchical cluster analysis identified four perception-based categories: Functional Machines, Social Companions, Anthropomorphic Agents, and Disembodied Assistants. Multilevel models showed that perceived competence and warmth predicted trust across categories. At the same time, anthropomorphism showed mixed effects: familiarity and animacy increased trust, whereas mind attribution and uncanniness decreased trust, with stronger negative effects for embodied agents. We further observed a dissociation between trust and willingness to interact, with willingness to interact more sensitive to visceral discomfort and category ambiguity. These findings indicate that trust in AI is category-dependent and that effective design should align social cues with users’ mental models.

This study examines the relationship between legal frameworks and web accessibility compliance across e-commerce platforms in the United States, United Kingdom, Canada, Australia, and South Africa. Using a mixed-methods design, 56 websites were evaluated against WCAG 2.0 and 2.1 Level AA standards, focusing on visual, auditory, operational, and content-related barriers. In parallel, national legal systems were analyzed using six variables: subject matter, definition, clarity, exemptions, penalties, and reporting. The findings underscore the role of accessible technologies in reducing barriers for people with disabilities and frame digital exclusion as a product of broader structural and systemic inequalities. Although WCAG is widely referenced, substantial implementation gaps persist, particularly in commercial contexts where legal requirements are ambiguous or weakly enforced. While structural and navigational accessibility showed relative strength, major disparities emerged in regulatory clarity, with the United States and United Kingdom leading and South Africa trailing, especially in large commercial e-commerce environments.

This study explores team emergent states (team cohesion, team identification, team psychological safety) in Human-AI teams (HATs) compared to human-only teams (HTs). In a laboratory experiment, participants (N = 67 teams; 134 individuals) completed two tasks. The first round involved two human teammates. The second round introduced a third teammate – either another human (HTs) or an AI (HATs). We hypothesized that HATs would exhibit lower levels of team emergent states. Results suggest that HATs exhibit lower team cohesion and identification than HTs, an effect that seems to occur indirectly through reduced levels of self-rated team performance and team trust. There was no difference in perceived team psychological safety. Participants in HATs identified less with the AI than with the human teammate. These findings suggest that traditional team dynamics might not be directly applied to HATs. This research advances our understanding of the implications of AI teammates, providing practical insights for implementing AI while maintaining effective teamwork.

Cognitive workload (CW) recognition is crucial in real-time and dynamic decision-making environments. Electroencephalogram (EEG) provides a cost-effective, flexible, and high-resolution approach for CW recognition.Various research studies has been developed that faces less accuracy in prediction and also leads error in estimation. Hence, a robust deep learning (DL) model for CW estimation is proposed. Initially, EEG signals from EEG workload dataset are pre-processed then third-order spectral cumulant features are extracted using residual dilated bidirectional long short-term memory network for obtaining higher-level feature representation. Subsequently, temporal and spectral features are fused using an attention mechanism, and an ensemble spatial attention with an enhanced stack autoencoder is employed for CW recognition. Proposed method attained an accuracy of 95.1% in no-task experiment and 95.5% in simultaneous capacity-based multitasking activity. Experimental findings demonstrate the effectiveness of proposed method in EW prediction. Proposed model has the potential to enhance decision-making and performances in constraint environment.

Human-AI collaboration has demonstrated unique potential in various decision-making scenarios. However, the dynamic and personalized nature of online decision-making (e.g., shopping, trip planning) poses challenges to achieving alignment between AI support and user requirements. To address these challenges, we took online shopping as a representative scenario and conducted experience prototyping to explore human-AI collaboration in decision-making tasks. It revealed a dynamic transition in user requirements from ambiguity to clarity, while limited transparency and decision control hindered communication efficiency. Building on this, we formulated a design strategy comprising 10 AI actions that dynamically contribute to the human-AI interaction loop, aiming to achieve alignment between AI support and user requirements. We implemented this strategy in ContrXAI, an LLM-powered mixed-initiative system, and evaluated it as a technology probe. Findings demonstrated that ContrXAI effectively instantiated the design strategy, enhancing human-AI alignment and enriching the collaborative decision-making experience. We also identified typical user interaction patterns with ContrXAI and highlighted their expectations for AI collaborators.

This study examines how two facial attributes of robots—the facial width-to-height ratio (fWHR) and eye shape—influence users’ judgments of trustworthiness, and how these features are processed at the neural level. To achieve this goal, we conducted a within-subject Electroencephalogram (EEG) experiment using a 2 (fWHR: low/high) × 3 (eye shape: round/rectangular/obround) full factorial design. EEG signals were analyzed using event-related potentials (ERPs). Results reveal that eye shape significantly influences trustworthiness, with round and obround eyes rated higher than rectangular ones. High fWHR robots with round or obround eyes elicited more negative N1 and N170 amplitudes, while round eyes reduced P3 amplitudes compared to other shapes. These findings not only clarify the neural mechanisms underlying the evaluation of robotic facial cues but also provide practical insights for designing socially trustworthy robots.

This research presents a novel design and implementation of an Intelligent Virtual Agent (IVA) in mixed reality that has advanced speech capabilities from large language models and integrates computer vision to perceive the user’s environment and actions in the real-world context. Scene understanding allows the IVA to navigate in the user’s physical space, demonstrate an understanding of the user’s actions, and dynamically interact with real-world entities. We propose a comprehensive framework for this multimodal integration, which enables the IVA to tailor its assistance and provide adaptive guidance to the users based on the actions they take in the real-world. To demonstrate and evaluate the proposed framework, we implemented two novel scenarios. Results demonstrated that participants consistently reported higher engagement, interactivity, and effectiveness with the IVA despite taking more time to complete the task. Moreover, all participants valued the IVA’s ability to adapt to their actions, offering a more personalized experience.

Graphical assets play an important role in design and development of games. There is potential in the use of AI-driven generative tools to aid in creation of such assets, improving pipelines. However, there is little research to address how generative methods can fit into the wider pipeline, and no guidelines or heuristics for creating such tools. Hence, we conducted a user study with 16 game designers and developers to examine their behaviour and interaction with such tools. Findings highlight that early design stage is preferred by all participants. Designers and developers prioritise rapid variations over initial quality of assets. Results also strongly raised the need for better integration of such tools in existing design/development environments and pipelines, specifically regarding common data formats and output manipulability. Informed by these results, we provide a set of heuristics for creating tools that meet the expectations and needs of game designers and developers.

This bibliometric study maps research at the intersection of artificial intelligence (AI), wearable devices, and aging, emphasizing human–computer interaction implications. We analyzed 1,057 Web of Science articles (2005–2024) using science mapping and visualization to identify key contributors, collaboration networks, and research hotspots. Publication output has grown steadily, led by the United States and China, with a nine-year citation half-life indicating both foundational and emerging influence. The field forms an interdisciplinary ecosystem integrating wearable sensing, machine learning, and older adult care. Research hotspots focus on intelligent data processing for health management and chronic disease monitoring, alongside increasing attention to usability, accessibility, and user-centered design. However, theory-driven HCI work remains limited, particularly in addressing older adults’ cognitive, perceptual, and practical needs. These findings highlight links to innovation and health-related Sustainable Development Goals and suggest directions for more inclusive, theory-informed interaction design.

The rising demand for services and scarcity of hospital workers call for innovations to decrease workload and support occupational well-being. Recent technological advancements have enabled the implementation of autonomous mobile robots as service robots for use in service delivery. We studied perceptions of service robots through retrospective data and a technology acceptance-based survey of 64 hospital workers (Study 1) and how service robots co-constructed the occupational well-being with 16 experienced hospital workers through a survey and qualitative interviews (Study 2) at hospitals in Denmark and South Korea. Results indicated equivalence in use of service robots across the two hospitals, high work engagement, and experiences of predictability and flow, which enabled hospital workers to prioritize work tasks and focus throughout the day. We found that service robots are complex human-robot interaction technologies that carry demands and resources, including casting anthropomorphism and playful work design as important job resources in digitalized workplaces.