CHIRA 2026 Abstracts

Area 1 - Interactive Devices
Full Papers Short Papers
Area 2 - Adaptive and Intelligent Systems
Full Papers Short Papers
Area 3 - Human Factors for Interactive Systems, Research, and Applications
Full Papers Short Papers
Area 4 - Interaction Design
Full Papers Short Papers

Area 1 - Interactive Devices

Full Papers
Paper Nr: 47
Title:

Playing for Sustainability: Exploring Robot-Assisted Serious Games to Support Sustainable Living in Families with Children at Home
Authors:

Nasim Beheshtian, Aino Ahtinen and Kaisa Väänänen
Abstract: This study explored the role of robot-assisted serious games (RASG) in supporting sustainable living within families’ homes. Two physical board games, a collaborative and a competitive game, were de￾signed and facilitated by a social robot to explore if the games had the potential to foster eco-friendly living. Eight families (n=26 family mem￾bers) participated in the study. A mixed-methods approach was utilized to collect qualitative and quantitative data on players’ intrinsic motiva￾tion, their overall experience of playing the games, and the potential of robot-assisted serious games in supporting sustainable living. The find￾ings showed both games were well-received, with the collaborative game demonstrating greater potential to support sustainable living practices, particularly among children. The robot’s persuasive features, including gestures and feedback, were also perceived as impactful during the collab￾orative game. Overall, age-related differences were evident, and RASG proved to be an engaging platform for fostering eco-awareness, with chil￾dren exhibiting the greatest enthusiasm and potential for adopting sus￾tainable practices.
Paper Nr: 77
Title:

Social Robots at Recess: Exploring Children’s Social Interaction and Acceptance across Play Formats
Authors:

E. Liberman-Pincu and T. Oron-Gilad
Abstract: Social robots are emerging as promising tools in education, not only for supporting learning but also for promoting children's emotional and social well-being. School recess is an important yet underexplored context for children's social development. To inform the study design, teacher feedback was collected through a preliminary online survey of 65 educational professionals. Responses suggested that robot-mediated recess activities may support children's social interaction, while highlighting the need to adapt the robot to the school context. The child study was conducted with 24 fourth-grade students, divided into six small groups, using a between-subjects design. Each group interacted with the temi robot through one of three game formats: cooperative group play, individual competitive play, or competitive group play. Children then completed a child-adapted Almere questionnaire and participated in a class discussion. Preliminary findings suggest that the robot's appearance, responsiveness, and communication style should be refined to better support trust, inclusion, and emotional connection during recess.
Short Papers
Paper Nr: 30
Title:

Can Iterated Learning Help Us Optimise Hand-Based Gestural Control in AR? Not Really: A Pilot Study
Authors:

Jacek Matulewski, Piotr Ablewski, Monika Boruta and Jacek Jabłoński
Abstract: To our knowledge, there have been no attempts to use iterated learning to optimise hand gestures designed to control software. Inspired by Namboodiripad's research on gesture optimisation in human communication, we verified the effectiveness of this approach in human-computer interaction. We conducted a pilot study to evaluate a method of gesture optimisation based on a set of gestures designed to control AR software. The experiment involved 28 participants, assigned to four learning chains comprising seven generations each. Each participant learned and reproduced 33 gestures recorded by a prior participant. Each of the participants was the learner and the source of gestures for the next-generation participant. The video and motion-capture recordings were used to verify the conformity of the produced gestures with the original pattern, as well as their volume and trajectory length. Assessments of the usability of the gestures were analysed. The results indicate that the iterative learning method does not lead to the optimisation of gestures.
Paper Nr: 67
Title:

Virtual Reality Teleprompter-Based Public Speaking Simulation
Authors:

Thanh-Hai Nguyen, Duc-An Tran, Tam V. Nguyen, Viet-Tham Huynh, Duy-Nam Ly and Khanh-Duy Le
Abstract: Public speaking with teleprompters can improve delivery fluency and reduce cognitive load, yet without rehearsal it often results in rote reading and reduced audience engagement. We present a virtual reality (VR) teleprompter system that supports public speaking training through multimodal tracking and real-time feedback. The system combines speech recognition, semantic alignment, and gaze analysis to monitor user performance in simulated presentations. Spoken content is transcribed and semantically matched to the script, allowing the teleprompter to advance smoothly even when users paraphrase. To promote skill development, the system provides (1) a responsive virtual audience that reacts to the speaker’s head orientation, and (2) quantitative visual feedback including attention heatmaps and word-rate graphs, and (3) a synchronized playback video review that fuses third-person footage, headset audio, and subtitles for post-session reflection. A GPT-based module further evaluates semantic coherence and offers real-time suggestions for improving clarity and structure. The performance evaluation shows that our system effectively supports real-time VR presentation training by achieving low online latency dominated by ASR inference and robust script-to-speech alignment tolerant to paraphrasing. In addition, a pilot study suggests that participants perceived the VR teleprompter–based approach as useful for presentation practice, highlighting its potential for more effective, accessible public speaking rehearsal.
Paper Nr: 74
Title:

Do Input Devices Matter for Cognitive Assessment? Investigating Interaction Modality in a Digital Tower of London Task
Authors:

Maxime Macé, Emmanuelle Richard, Paul Richard and Teddy Louisy
Abstract: This paper argues that interaction modality should not be treated as a neutral implementation detail in digital cognitive assessment. Different input devices may introduce varying levels of extraneous cognitive load, potentially influencing the measurement of executive functions. Despite extensive research on digital cognitive assessment and human–computer interaction, relatively little attention has been paid to interaction modality as a fundamental design dimension that may shape the validity and interpretation of cognitive assessment outcomes. We discuss this challenge through a digital implementation of the Tower of London (ToL) task, comparing indirect mouse-based interaction with direct touch interaction on a touch table. Performance will be assessed using behavioral measures of planning and problem solving, together with subjective workload ratings.
Paper Nr: 92
Title:

Reducing Interaction Ambiguity in Social VR through Visual UI Interaction-State Cues
Authors:

Ammar Yasser and Amal Yassien
Abstract: Social virtual reality (VR) enables users to collaborate through embodied avatars, but interactions with virtual objects and user interface (UI) elements can resemble social gestures, making users' intentions difficult to interpret. This interaction ambiguity can hinder communication and coordination during collaborative tasks. This paper proposes a lightweight visual interaction-state cue that changes an avatar's color while the user is interacting with virtual objects, allowing collaborators to distinguish object interaction from social communication. The cue was implemented in a multiplayer virtual café and evaluated through a between-subjects user study with 40 participants performing a collaborative object manipulation task. Participants using the cue were better in assessing their partner's reaction and perceived the interaction as modern and leading edge. The findings suggest that simple visual interaction-state cues can reduce interaction ambiguity and improve collaborative communication in social VR.
Paper Nr: 98
Title:

Stepping Inside the Future: A Virtual Reality Testbed for More Realistic Evaluation of Conversational Home Assistants
Authors:

Pedro Carneiro, Inês Águia, Bernardo Marques, Ana Patrícia Rocha, Nuno Almeida and António Teixeira
Abstract: Conversational assistants have emerged as a natural interaction paradigm for smart-home environments, yet their iterative development and refinement typically involve user-centred evaluation within realistic domestic contexts. Although physical Living Labs provide high ecological validity, they are costly to build, difficult to reconfigure, and unsuitable for rapidly evaluating conversational systems during early development stages. This paper proposes an immersive Virtual Reality (VR) testbed that enables the iterative development and pre-deployment evaluation of conversational assistants within a representative virtual smart-home environment. The proposed approach is founded on the hypothesis that a sufficiently realistic virtual representation preserves the contextual cues required for meaningful conversational interaction before deployment in physical installations. To validate this hypothesis, a proof-of-concept integrating a voice-based domotics assistant was implemented and evaluated through an exploratory user study involving five participants. Results indicate that users were able to naturally interact with the virtual environment despite limited prior VR experience and perceived it as immersive and representative of a smart-home setting. Furthermore, participants were able to evaluate the conversational assistant without significant interference from the VR environment. These findings provide initial evidence that immersive VR environments can serve as a suitable pre-deployment platform for obtaining meaningful user feedback during the iterative development and evaluation of conversational home assistants.
Paper Nr: 24
Title:

Humanoid Telerobotics as a Bridge for Expectation Calibration and Requirements Elicitation: An Exploratory Industrial Pilot Study
Authors:

Frederic Schubert and Alexandros Giannakidis
Abstract: Early industrial assessment of humanoid robots requires stakeholders to relate general expectations to the requirements of specific tasks and the capabilities of a particular robot and interface configuration. This paper presents an exploratory qualitative case study of the Reachy 2 robot in a workshop setting. Five stake-holders first provided an initial qualitative assessment of their expectations and subsequently participated in a two-week practical pilot comprising a platform and teleoperation introduction, guided hands-on familiarisation, collaborative scenario definition, expert-operated scenario trials and individual interviews. Case-level comparison showed that initial expectations became more task-specific, particularly regarding grasp reliability, force transmission, positioning stability, obstacle coverage, teleoperation, and process fit. Handheld milling could not be completed, while sanding exposed an inadequate fit between gripper and tool geometry. Third-hand assistance enabled coarse holding and presentation, but grasping and positioning remained limiting factors. Across scenarios, the analysis yielded traceable requirements concerning end effectors, platform mechanics, perception, interface feedback, control assistance, task redesign, and organisational implementation. The study further developed a diagnostic perspective on operator compensability: the extent to which limitations can be compensated through teleoperation depends on available feedback, operator practice, task tolerances, physical system capabilities, control assistance, and opportunities for task redesign. The findings indicate that teleoperation can serve as a situated mode of early requirements elicitation within exploratory and configuration-specific assessments of humanoid robots.
Paper Nr: 50
Title:

User Study of 3D Shipbuilding Models Viewed in a VR Environment for Material, Design, and Installation Deviation Tracking
Authors:

Milla Pöyry, Jan Lehtinen, Krister Stenius, Ali Khan and Kimmo Tarkkanen
Abstract: Large shipbuilding projects involve complex geometries, dense system integration, and strict regulatory requirements, making the identification and management of material, design, and installation deviations challenging. Although Virtual Reality (VR) technologies and commercial CAD review solutions have been available for years, many existing systems rely on proprietary cloud platforms, support only limited file formats, or require substantial software investments. These limitations may hinder adoption, particularly among small and medium-sized companies operating with sensitive design data. This study investigates the use of a locally deployed VR environment for deviation tracking in shipbuilding. An application prototype was developed using the Unity engine to enable immersive visualization and interaction with CAD-derived shipbuilding models. The prototype was evaluated with industry professionals using real shipbuilding models and realistic inspection scenarios. The results suggest that immersive full-scale inspection can support the early identification of design- and installation-related deviations. Participants identified measurement tools, materials’ metadata integration, transparency controls, and annotation capabilities as important requirements for practical adoption. The findings indicate that locally operated VR solutions have potential as a cost-effective complement to existing deviation management practices.
Paper Nr: 70
Title:

Making Absence Visible: Visualizing Digitally Reconstructed Heritage of Nazi Persecution with Augmented Reality in Contemporary Landscapes
Authors:

Anna Pyrmer, Chris Hall and Stefan Jänicke
Abstract: This work explores how Augmented Reality (AR) can make destroyed or erased heritage visible at memorial sites related to Nazi persecution. Using historically informed 3D reconstructions, we developed a mobile AR application that allows visitors to experience lost structures at their original locations. We present a Unity-based framework and two case studies: the Great Choral Synagogue in Riga, Latvia, and the Ferramonti Internment Camp in Tarsia, Italy. Through a human-centered, participatory design process with local memorial institutions, we developed historically authentic and context-sensitive experiences. User feedback indicates that AR improved visitors' understanding of the historical environment and strengthened their connection to the sites' histories. The effect was particularly evident at the Great Choral Synagogue, where the reconstruction conveyed the scale and presence of a deliberately destroyed landmark. We further discuss challenges surrounding authenticity, ethics, privacy, and respectful engagement. Our findings suggest that AR can reconnect contemporary memorial landscapes with their lost historical fabric and bring invisible histories into public view.

Area 2 - Adaptive and Intelligent Systems

Full Papers
Paper Nr: 105
Title:

GWAIHIR: Generating Weekly Action Plans by Intelligent Agents Collecting Habits, Interests and Routines
Authors:

Maani Beigy, Roy van den Munckhof, Laura Genga and Pieter Van Gorp
Abstract: Mobile health applications recommend activities, such as walking 5000 steps, eating vegetables, or doing mindful breathing, but seldom specify when to perform them. In the dynamic context of daily life, deciding ``when'' is itself a burden, and many people abandon the goals that brought them to the app. Scheduling these activities to fit a person's evolving context, routines, and commitments is hard, and apps rarely personalize it. We mitigate this burden by facilitating scheduling with GWAIHIR -- a multi-agent conversational coach. Four large language model (LLM) agents, coordinated by a deterministic orchestrator, collect information from the user in natural language, place the recommended tasks into the free time slots using Allen's interval algebra, and validate them against preferences and temporal constraints. We report an in silico evaluation on the Progressive Healthy Lifestyle Challenge of \textsc{CalendarBench}, which simulates thirty persons from three contrasting personas over eight weeks with a growing weekly batch of recommended tasks, scored with a multi-objective human-centric scheduling metric. We observed a higher scheduling gain for the multi-agent coach compared to single-agent LLMs, reinforcement-learning, and heuristic planners, with the handcrafted schedules of a human coach ahead of all automated schedulers. In addition, we implemented GWAIHIR as a platform-independent service and deployed it in GameBus platform with chat and calendar user interfaces, where it returns the proposed time slots while keeping the person in control. This work presents a proof of concept that demonstrates technical feasibility and paves the way for studies on human participants to explore acceptability and real-world effectiveness.
Short Papers
Paper Nr: 15
Title:

Generative Human Motion Mimicking through Feature Extraction in Denoising Diffusion Settings
Authors:

Alexander Okupnik, Daniel Bisig, Kyriakos Flouris and Johannes Schneider
Abstract: Generative motion models offer new opportunities for embodied human–AI co-creation, yet most dance-generation systems either synthesize single-person motion from music or learn two-person coordination from duet datasets. We study a complementary setting: generating partner-like dance motion from a single incoming performer sequence. The proposed framework guides a diffusion-based motion model by decomposing the reference motion representation into an interpolationsmoothed component and a residual component, referred to as lowand high-frequency components following the ILVR terminology. In the implementation used for the reported experiments, smoothing is applied along the feature axis of the motion representation; the smoothed reference component provides feature-level guidance, while the generated residual preserves variation. The method combines an established motion continuation framework with Iterative Latent Variable Refinement (ILVR)-inspired reference refinement, yielding inference-time control over the degree of motion mimicry. We evaluate the approach in an offline study using generated 3D motion clips. Quantitatively, we report Fréchet distance and diversity scores using both kinetic and manually defined motion features. Perceptually, we conduct a user study in which participants rate generated motions with respect to domain competence, originality, value, surprise, perceived reference responsiveness, and the balance between following and variation. The descriptive results indicate that intermediate refinement strengths provide the most favorable perceptual trade-off, suggesting that feature-smoothed reference guidance can support controllable partner-like motion generation while preserving creative variation.
Paper Nr: 54
Title:

Appropriate Reliance as a Multifaceted Phenomenon: Characterizing Reliance Impact Factors in XAI-Assisted Decision-Making
Authors:

José Cezar de Souza Filho, Rafik Belloum and Káthia Marçal de Oliveira
Abstract: Explainable AI (XAI) aims to improve decision-making by providing explanations to help users understand AI predictions. While XAI holds promise, users often demonstrate inappropriate reliance on these systems, either over-relying on incorrect predictions or under-relying on accurate ones. Although recent studies have proposed strategies to mitigate reliance issues, evidence remains fragmented and lacks a structured view of what should inform Explainable User Interface (XUI) design to support appropriate reliance in human-AI collaboration. In this study, we address this challenge by conducting a systematic literature review with a grounded theory-based analysis. As a result, we identified a set of factors impacting appropriate reliance. Our findings show that appropriate reliance is a multifaceted phenomenon shaped by the interplay of contextual, human, and system-related factors, which are often studied in isolation. These findings provide a structured foundation for reliance impact factors in XAI-assisted decision-making that can guide future research in human-centered XAI.
Paper Nr: 60
Title:

Music Recommendation Based on Continuous Valence–Arousal Estimation from Lyrics, Audio, and User's Physiological Signals
Authors:

Masaki Omata and Shinnosuke Watanabe
Abstract: Music recommendation tailored to a listener's emotion is a promising way to enhance the listening experience, yet conventional approaches often rely on discrete emotion categories and on either lyrics or audio alone. In this study, we propose a music recommendation method that represents emotion as continuous values in the two-dimensional valence–arousal space. The emotion of each song is estimated from its lyrics and audio using separate regression models for valence and arousal, and the two estimates are integrated by a weighted average; we found that the optimal weight ratio of lyrics to audio was 2:8 for valence and 5:5 for arousal. The user's emotional state is estimated from physiological signals—electroencephalogram, blood volume pulse, electrodermal activity, and skin temperature—using a time-series regression model, and a song is recommended by a nearest-neighbor search in the valence–arousal space. In a within-subjects experiment (N = 8), the proposed valence–arousal recommendation was compared with conventional emotion-category recommendation, Self-Assessment Manikin (SAM) recommendation, and random recommendation, and the presented songs were ranked and analyzed using the normalized-rank approach. The proposed method achieved significantly higher emotional congruence than the other methods.
Paper Nr: 83
Title:

LLM-Based Semantic Echo Cancellation for Human-AI Dialogue Systems: A Preliminary Proof-of-Concept
Authors:

Mario Radke, Bakir Hadžić, Akihiro Yorita, Naoyuki Kubota and Matthias Rätsch
Abstract: Overlapping speech in human-AI conversational systems creates mixed transcripts that hinder downstream processing. While no prior work has systematically evaluated semantic post-processing for speaker separation in this context, conversational AI systems possess complete knowledge of their own generated output, enabling semantic filtering unavailable in human-human dialogue. While not competing with hardware-level Acoustic Echo Cancellation (AEC), this study provides the first quantitative proof-of-concept evaluation of LLM-based semantic post-processing to separate user speech from chatbot content in Speech-to-Text outputs. We evaluate performance across three signal-tointerference ratios (+6 dB, 0 dB, −6 dB) using 150 manually annotated test instances, achieving 34–86% success rates, with performance showing a strong, statistically valid dependence on signal-to-interference ratios (p < 0.001). The semantic filter operates downstream of any STT system and maintains viable performance even under adverse acoustic conditions. These baseline metrics establish foundational performance characterization for future hybrid acoustic-semantic approaches and demonstrate the feasibility of semantic post-processing as a complementary approach to traditional diarization methods.
Paper Nr: 57
Title:

A Matrix-Based Visualization Tool for Learning Backtracking Algorithms
Authors:

Ying Zhu
Abstract: Backtracking is a fundamental algorithm for combinatorial search problems such as subsets, permutations, combination sum, N-Queens, and constraint satisfaction. Although it is central to data structures and algorithms courses and technical interviews, many students struggle to understand it because it combines recursion, iteration, pruning, and undoing previous choices. Learners often have difficulty forming a clear mental model of how the search path evolves and how similar backtracking problems differ in subtle ways. To address these challenges, we present an interactive visualization system for learning backtracking. The system uses a synchronized multi-view interface that updates dynamically as users step through source code. Its key idea is to represent backtracking as a matrix-based search process: each row corresponds to a recursion depth, and each column represents a candidate choice. The visualization highlights active choices, current path elements, skipped options, base cases, and backtracking steps, while also making recursion, return flow, and iteration flow explicit. The system makes three contributions. First, it provides a compact visual model for understanding backtracking execution. Second, it offers a consistent matrix-based representation for comparing related backtracking problems that share similar recursive structures but differ in rules such as candidate reuse, duplicate handling, and index advancement. Third, it coordinates source-code highlighting, matrix state, recursion depth, partial output, and explanatory callouts, making the hidden dynamics of backtracking visible to learners.
Paper Nr: 85
Title:

Towards Personality-Aware Fashion Recommendations: A Comparative Study of Personality Integration Strategies
Authors:

Me Me Myat Noe, Suncica Hadzidedic and Mengyisong Zhao
Abstract: Fashion recommendation is challenging because style preferences are subjective and partially expressed in interaction histories, making behavioural signals alone insufficient to capture individual taste. While personality-aware recommender systems (RSs) have shown promise in music and movie domains, their use in fashion remains underexplored. This study investigates how personality traits influence fashion recommendation by comparing three RS architectures—a post-filtering Factorisation Machine (FM) reranker, integrated FM, and DeepVBPR-against a FashionCLIP cosine similarity baseline. Personality traits are inferred from Amazon Reviews 2023 user text using a fine-tuned RoBERTa model. Personality integration consistently increased intra-list diversity across all architectures. Integrated FM with hard negative sampling achieved the highest diversity but showed increasing popularity bias with larger candidate pools, while DeepVBPR most consistently recommended items less popular than users' own histories. None of the personality-aware models surpassed the baseline in accuracy. These findings suggest personality primarily affects recommendation characteristics rather than relevance, highlighting an accuracy--diversity trade-off shaped by RS architecture and negative sampling.
Paper Nr: 90
Title:

Designing Phoneme-Level Feedback for Child Speech Remediation under Imperfect Recognition
Authors:

Vidushee Shekhar
Abstract: Children with phonological difficulties need feedback at the level of individual speech sounds, but automatic speech recognition (ASR) is least reliable exactly where those children sit: young, variable, and often non-standard in their productions. Most pronunciation tools hide this uncertainty behind a single correctness verdict. We argue the opposite is required: the design question for a remediation interface is not how to present a verdict, but how to behave when the recogniser may be wrong. This paper contributes an interaction design for phoneme-level literacy remediation, together with an empirical characterisation of the recognition constraint that design must tolerate. We specify what evidence a forced-alignment pipeline exposes (phone identity, temporal boundaries, and a per-phone confidence score) and how that evidence maps onto what a child, a teacher, and a therapist each see, including a confidence-gated correction loop that withholds corrective feedback rather than risk a false correction. The design is presented as design intent; it has not been deployed with children, and no learning-outcome claims are made. To quantify the constraint, we compare four acoustic models under a unified Kaldi pipeline: a monophone baseline, a speaker-adaptive triphone system (Tri3b), a Time Delay Neural Network (TDNN), and a Factorised TDNN (TDNN-F). All are trained on a 3,000-utterance subset of adult LibriSpeech train-clean-100 and evaluated on child speech from the JIBO Kids corpus; no child speech is seen in training, so results describe adult-to-child domain transfer. TDNN-F reaches the lowest phone error rate (34.2%) while Tri3b gives the closest boundary agreement (58 ms), motivating a hybrid design. Because JIBO Kids provides word-level transcription only, these figures measure divergence from a canonical lexicon, not verified pronunciation error.

Area 3 - Human Factors for Interactive Systems, Research, and Applications

Full Papers
Paper Nr: 28
Title:

Browser-Based Touch Behavioral Biometrics for Mobile E-Commerce
Authors:

MTeresa Gonzalez-Garcia, Martin Gonzalez-Rodriguez, Daniel Fernandez-Lanvin and Javier De Andres
Abstract: Behavioral biometric authentication based on mobile touch interaction has traditionally relied on native applications and sensor-enhanced environments. However, modern browser security and privacy restrictions limit access to many device-level signals previously available in web contexts. This work investigates whether user discrimination remains feasible in mobile browser-based e-commerce environments using only interaction data obtainable through standard web APIs. The proposed approach formulates authentication as a one-class classification problem and models user behavior using Mahalanobis-distance metrics with different covariance estimation strategies. Interaction features were extracted exclusively from browser-level Pointer Events generated through touchscreen interaction, without relying on auxiliary sensors or privileged device information. Experimental results show that meaningful behavioral discrimination can still be achieved under these constraints. Robust and regularized covariance estimators, including Ledoit-Wolf and Minimum Covariance Determinant, achieve identical best Equal Error Rate (EER) values of 9.77%, while Ledoit-Wolf attains a higher Area Under the ROC Curve (AUC-ROC) of 0.9865, indicating improved overall separability. These findings demonstrate the potential of browser-native behavioral authentication for mobile e-commerce.
Paper Nr: 35
Title:

Is This Correct? How Perceived Inaccuracies Shape Museum Visitors' Engagement with a Virtual Expert

Authors:

Marcus Winter and Phil Blume
Abstract: This paper examines museum visitors' perceptions of inaccuracies in the responses of a conversational agent, cast as a virtual expert on the history and architecture of a heritage building, and how they shape their engagement and learning. It discusses the curatorial need addressed by the agent, its technical design and implementation, and its empirical evaluation with visitors to the building. Findings show that visitors engaged extensively with the agent and appreciated the direct access to expertise, despite expressing a preference for human interaction. Most participants reported that the agent helped them to learn about the building's history and architecture. While many qualified this by expressing doubts about the provided information, perceived inaccuracies also provided opportunities for deeper engagement and independent inquiry. The highly rated user experience, potential for learning, and open, low-energy, privacy-preserving technical design strongly align with curatorial goals. The paper's main contribution, empirical findings on how perceived inaccuracies affect visitor's engagement and learning, is relevant for practitioners and researchers designing conversational agents in heritage contexts.
Paper Nr: 38
Title:

Visual Analytics across Scales for Infectious Disease Analysis and Decision Making
Authors:

Apoorva Karagappa, Jonas Gilg, Moritz Zeumer, Pawandeep Kaur Betz and Andreas Gerndt
Abstract: Actionable decision-making in epidemiological scenarios requires more than the analysis of aggregated case numbers. It involves reasoning across multiple interdependent dimensions, including the temporal evolution of infections, the spatial distribution of outbreaks, population heterogeneity, mobility behavior, and localized interaction pathways. Existing epidemiological dashboards and model-based visual analytics systems provide valuable support for monitoring, forecasting, and scenario comparison, but they often treat macroscopic epidemic trends and microscopic mobility dynamics as separate analytical perspectives. This limits the ability of analysts to move from descriptive observations of disease spread toward localized explanatory reasoning and intervention-oriented decision support. In this paper, we present a cross-scale visual analytics framework for interaction-driven spatio-temporal epidemiological analysis. The framework follows a general-to-specific analytical logic, enabling users to move from macroscopic population-level outbreak patterns toward stratified, mobility-based investigation. We instantiate the framework through two visual analytics components. ESID supports macroscopic analysis of regional epidemic trends, forecasts, infection states, and intervention scenarios. PANDEMOS extends this workflow with microscopic analysis of agent-based mobility data, including trips, activity contexts, transport modes, origin–destination relationships, and trip chain structures. Together, these components support a macro-to-micro reasoning process in which regional outbreak signals can be progressively connected to localized mobility and interaction patterns.
Paper Nr: 42
Title:

Bridging Generative AI and Sociotechnical Practice for the UN 2030 Agenda: A SAwD-Based Analysis of SDG Extended Reports
Authors:

Flávio Henrique Alves, Maria Cecilia Calani Baranauskas and Alexandre L'Erario
Abstract: This article proposes a Socially Aware Design (SAwD) process to support the automatic extraction of information from SDG Extended Reports, using Large Language Models (LLMs) integrated into a Retrieval-Augmented Generation (RAG) pipeline. The approach follows an evidence-first principle, in which each extracted element must be supported by direct textual evidence from the analyzed document. To this end, the process operationalizes three SAwD artifacts: the Diagram of Interested Parties, the Evaluation Frame, and the Semiotic Ladder, through a multi-agent flow responsible for indexing, retrieval, synthesis, and validation. As a result, a structured dataset was produced from 17 SDG Extended Reports, enabling traceable and evidence-based extraction consistent with the provided sources. The results indicate that integrating LLMs into SAwD enables a more systematic, auditable, and socially contextualized analysis of SDG Extended Reports. Furthermore, the structured information generated can support semioparticipatory workshops, stimulating critical discussion among different interested parties across the Operation, Contribution, Source, Market, and Community layers, and strengthening meaning-making and decision-making processes aligned with SAwD principles. Thus, the proposed method brings large language models closer to sociotechnical design practices applied to monitoring and improving sustainable development, and can therefore help advance the goals of the United Nations (UN) 2030 Agenda.
Paper Nr: 44
Title:

An Open Physiological Dataset for Stress and Cognitive Load Assessment
Authors:

Sarra Baghdadi, Olivier Augereau, Nico Pallamin, Marc Hoarau and Pierre De Loor
Abstract: Cognitive load and stress significantly influence human performance, learning, and decision making. Excessive mental workload can impair efficiency, increase error rates, and contribute to fatigue, making its assessment a critical challenge in fields such as human computer interaction, neuroscience, education, and ergonomics. While subjective measures remain widely used, physiological signals offer an objective and continuous means of monitoring cognitive states, particularly in dynamic environments where self reports are impractical. This paper presents a multimodal physiological dataset specifically designed for the assessment of cognitive load and stress. The dataset combines electrodermal activity (EDA), photoplethysmography (PPG), electroencephalography (EEG), pupillometry, and eye-tracking measurements, together with stress ratings obtained through a Visual Analog Scale (VAS) and cognitive load ratings obtained through the Instantaneous Self Assessment (ISA) scale. By integrating complementary physiological modalities, the dataset enables comprehensive investigations of the relationships between cognitive load, stress and physiological responses. We detail the experimental protocol, data acquisition procedures, and labeling methodology, including VAS and the use of ISA for cognitive load evaluation. Furthermore, we describe the dataset sharing policy, under which a subset of the dataset will be made openly available to the research community, while access to sensitive data will be subject to controlled sharing procedures. We also discuss potential applications of the dataset for developing and evaluating machine learning models for cognitive state estimation, as well as for advancing research in neuroscience, ergonomics, and human computer interaction.
Paper Nr: 58
Title:

Cognitive Value Recontextualization for Reflective Moral Alignment in Human–AI Decision Support: An Empirical Study
Authors:

Waseem M. Samkari and Thomas C. Eskridge
Abstract: Human–AI decision support in morally charged, high-stakes settings is almost always evaluated on outcome quality, while whether a recommended action aligns with the decision-maker’s own moral commit ments is rarely measured. Standard preference-learning pipelines deepen this gap by treating behavioral inconsistency as noise to smooth or drift to silently re-estimate. We present Value-Recontextualizing Decision Support (VRDS), a domain-agnostic paradigm that treats moral in- consistency as a diagnostic signal, realized through two interaction mechanisms: Cognitive Value Recontextualization (CVR), which re-presents an apparent value violation in a morally intensified frame, and Adaptive Preference Alignment (APA), which inserts an explicit confirmation step before any value-profile update. In a between-subjects 2×2 wildfire- crisis study (N = 52), the full CVR+APA condition produced the highest Value Consistency Index (M = 0.875); CVR was the dominant alignment driver and APA contributed an additive improvement. Objective task performance did not differ significantly across conditions, and a post-hoc equivalence test placed the full-system–baseline difference within a small operational margin (δ = 0.05). Perceived value consistency, not task performance, predicted reflective decision quality. The results suggest that interaction design can support value-consistent decision-making without a measurable performance cost, reframing alignment and performance as separable design dimensions for human-centered AI.
Paper Nr: 78
Title:

Trust in AI-Enabled Decision Support Systems for Critical Decision-Making
Authors:

Priit Värno, Chris Kristjan Kivaste and Sonia Claudia da Costa Sousa
Abstract: This article examines how trust is formed and calibrated in AI enabled decision support systems for critical military decision making under uncertainty, deception, and compressed decision cycles. It argues that AI’s operational value depends less on isolated performance gains than on socio-technical integration into command and control workflows that reshapes practices, authority, and accountability. Building on trust research, the paper distinguishes trust from reliance and emphasizes trust calibration, aligning user expectations with actual system capabilities, to avoid both overreliance and undue neglect in high stakes contexts. Using a sequential exploratory mixed methods design in the Estonian Defence Forces, Study 1 analyses 20 semi structured interviews to identify trust relevant barriers and facilitators and to elicit preferred AI outputs and explanation needs. Study 2 converts these themes into time constrained vignette scenarios, experimentally comparing recommendations with versus without explanations and embedding a critical system failure to observe trust erosion and recovery across repeated interactions, supported by repeated trust and workload measures. Results indicate trust is shaped by interacting system, user, interaction, and battlefield factors. Explainability supports early sensemaking and post failure recalibration when concise and operationally relevant, but can add burden and heighten disappointment after failures.
Paper Nr: 79
Title:

Modality Matters in AI Mock Interviews: Experience, Physiology and Performance with a Three-LLM Interview Panel
Authors:

Kanishka Samrat Kolakaluri, Tam V. Nguyen, Trung-Nghia Le, Minh-Triet Tran and Khanh-Duy Le
Abstract: Large language models have made automated interview practice broadly accessible, yet most deployed systems present the interviewer through a single channel, usually text. Whether the delivery modality itself changes how practice feels and how candidates perform remains poorly understood. We built InterviewAI, a mock-interview system in which a panel of three distinct LLM interviewers (Claude, GPT-4o and Gemini), each assigned a hidden competency focus, questions a candidate about their own résumé under a competency-gated termination policy modelled on real hiring panels. The same panel can be delivered as a text chat, a voice-only interlocutor, or a lip-synced 3D avatar, and the system continuously records answer scores, webcam-based gaze behaviour, and heart activity estimated by remote photoplethysmography, with all biometric processing on-device. In a counterbalanced within-subjects study, thirty-eight participants completed a full interview in each modality. Spoken modalities were rated far more natural than text (χ2(2) = 10.38, p = .006, Kendall’s W = .74), but the embodied interviewer was also experienced as the most stressful (p = .048), while heart-rate elevation was largest in the voice-only condition. Panel-assigned answer scores were descriptively lowest for voice and comparable between text and avatar. Usability was rated excellent (SUS M = 91.6), and modality preference split almost evenly. We discuss what these dissociations between experienced stress, physiological arousal and performance imply for the design of AI interview training, and release the system, which runs end-to-end on commodity hardware without API dependencies.
Paper Nr: 82
Title:

Differences in AI Anxiety: Exploratory Factor Analysis Suggests Similar AI Anxiety Structure in German- and English-Speaking Samples
Authors:

Nick Sheth and Klemens Weigl
Abstract: The mass use of AI applications is transforming human-AI interaction while uncovering and intensifying AI anxieties. To measure these anxieties in German-speaking populations, this study translates and psychometrically evaluates the Artificial Intelligence Anxiety Scale (AIAS) [31]. In contrast to a concurrently German published version by Hajek et al. [9] exploratory factor analysis (EFA) is employed to test whether a distinct anxiety structure emerges from the data. A quota sample (N = 225) with balanced age and gender representation completed an online survey. The full scale (α = .95) and subscales (α = .85-.96) demonstrated good to excellent reliability. EFA identified four factors: learning, AI configuration, replacement, and socio-technical blindness, largely consistent with the original English structure. Five items required removal, suggesting cultural and linguistic specificity. Limitations include the sample’s composition and the factor characteristics. The German AIAS provides a reliable instrument for human-AI interaction research and targeted anxiety interventions.
Paper Nr: 89
Title:

Identifying Students with Giftedness/High Ability as a Sociotechnical Process of Collective Intelligence: A Systematic Literature Review
Authors:

André Ferreira Zeferino, Claudia L. R. Motta, Cibele R. da C. Oliveira and Daniel Schneider
Abstract: This study presents a Systematic Literature Review (SLR) that investigates the challenges and perspectives in the identification process of students with Giftedness/High Ability (G/HA). Starting from the problem of under-identification, the research analyzes how scientific literature addresses the topic, identifying that the phenomenon transcends the limitations of isolated assessment instruments. The work supports the hypothesis that the identification of G/HA students constitutes a sociotechnical challenge of Collective Intelligence (CI), in which the fragmentation of knowledge among school actors hinders effective decision-making. The analysis, based on 12 studies selected between 1998 and 2026, points to the need to move beyond models centered on individual assessments in favor of collaborative architectures that integrate, validate, and share knowledge distributed across the school community. Ultimately, this review demonstrates that overcoming school-level under-identification requires a paradigm shift-moving from static, isolated assessments toward continuous, curated knowledge flows supported by CI frameworks.
Paper Nr: 101
Title:

From Hostility to Deliberation: A Sociotechnical Model for Mitigating Unproductive Debates in AI-Mediated Social Media
Authors:

Guilherme Moura, Luiz Filipe Brandi, Ramon Chaves, Emilly Lopes, Krsna Pinheiro Gusmão, Jano Moreira De Souza and Daniel Schneider
Abstract: Civil society's participation in digital public debate is undermined by the operational logic of mainstream platforms, whose algorithms prioritize engagement over the quality of discussions, fostering affective polarization and unproductive debates. Given the absence of design mechanisms capable of mitigating these dynamics, we propose a sociotechnical model, informed by Knowledge Management principles, to support the identification, organization, and mitigation of unproductive debates. This qualitative, documentary research is based on a comparative analysis of the Reddit, Discord, Mastodon, Bluesky, and Kialo platforms. From this synthesis, articulated with Marc Angenot's theoretical framework, we derive a model comprising five mitigation dimensions cut across by a Generative AI mediation layer, under the principle of human--AI symbiosis. An illustrative application to a real-world case demonstrates the model's potential to restore the conditions for deliberation and collective knowledge creation, offering guidance to platform designers and managers.
Paper Nr: 103
Title:

Self-Custody in an Unsupported Ecosystem: Usability and Security Challenges among Novice and Experienced Crypto Wallet Users in Bangladesh
Authors:

Md Ishmam Tasin, Md Sadek Ferdous, S M Taiabul Haque, Wolfgang Prinz and Farida Chowdhury
Abstract: Self-custodial wallets are crucial tools in blockchain systems. They are used for securely sending, receiving, and storing funds. However, users often encounter usability challenges in these wallets, which can lead to security vulnerabilities. Although these wallets are increasingly used across the Global South, populations in this region face unique socio-technical realities, including limited institutional support for blockchain education, and heavy reliance on informal markets and Mobile Financial Services, that remain largely unexamined in existing usability literature. Our study investigates the usability and security challenges faced by users with diverse levels of cryptocurrency experience in Bangladesh, one representative population of the Global South. We conducted two complementary studies, a task-based usability study with 10 non-users (Study 1) to identify first-contact usability barriers across core wallet features, and semi-structured interviews with 20 experienced cryptocurrency users (Study 2) to examine how usability challenges, security perceptions, and trust persist or evolve with continued use. Our findings provide insights into challenges around Secret Recovery Phrase storage and recovery, network management, credential comprehension across wallet passwords, private keys, and recovery phrases, transaction fee transparency, and advanced concepts such as Swapping and Bridging, alongside challenges distinct to the population we are studying, including reliance on informal peer networks and centralized exchanges to acquire cryptocurrency, and expectations of transaction feedback shaped by everyday use of Mobile Financial Services. Building on these findings, we propose design recommendations alongside a set of culturally grounded approaches for wallet designers seeking to build more usable, secure, and socially responsive self-custodial wallets for underserved global south context.
Paper Nr: 106
Title:

Political Deepfakes across the Electoral Cycle: A Systematic Review–Based Political Cycle and Human Interaction Framework for Democratic Resilience
Authors:

Evandro Claro Freire, André Ferreira Zeferino, Cibele Oliveira, Claudia Lage Motta and Daniel Schneider
Abstract: Generative artificial intelligence has expanded the production and circulation of political deepfakes, while the literature remains fragmented across technical, cognitive, social, and regulatory perspectives. This systematic review analyzes human interaction with political deepfakes and proposes an integrated framework. Searches in the ACM Digital Library and Scopus, covering the period from 2017 to 2026, retrieved 2,562 records. Following processing, deduplication, AI-assisted full-text screening, and human validation, 70 studies were included. The results show that the effects of political deepfakes depend on the synthetic artifact, political context, human factors, social interaction, platform ecosystems, and institutional governance throughout the political-electoral cycle. Evidence remains concentrated on exposure, perception, detection, and immediate effects, whereas cumulative institutional, governance, and long-term democratic consequences remain comparatively underexplored. Based on these findings, a framework is proposed, integrating six political-electoral stages, a recursive cycle of human interaction, the principles of Human-Centered Design, six cross-cutting analytical dimensions, and one stage-specific analytical dimension for each phase of the cycle, supporting democratic resilience.
Short Papers
Paper Nr: 16
Title:

Virtual Reality as a Lens to Examine User Attitudes toward Autonomous Driving in Realistic Traffic
Authors:

Thomas Miller, Enes Yigitbas and Robin Heveling
Abstract: The fast advancement of Autonomous Vehicles (AVs) has exceeded the understanding of how users perceive and accept this technology. Previous research on user acceptance often relies on theoretical or survey-based studies, while experimental work tends to focus on the user’s own vehicle, missing complex traffic interactions. This is a limitation, as real-world traffic is becoming more complex, involving many different vehicles and unpredictable interactions. Using Virtual Reality (VR), users can safely experience and interact with realistic traffic scenarios. This work introduces a VR-based driving simulator that integrates realistic and immersive traffic environments. The simulator was used to conduct a user study in which 27 participants experienced various driving scenarios, allowing an in-depth evaluation of the user acceptance of AVs. The study demonstrates that immersive VR is an effective method to investigate the user acceptance of AVs, as participants reported high usability, presence, and manageable workload. The study further indicates, based on the SCTAM and VRTAM analyses, that governmental approval, trust, and practical functionality are the most influential factors of participants’ acceptance.
Paper Nr: 17
Title:

From Gamification to Ritualization: A Novel Way for Reflecting Daily Lifestyle
Authors:

Risa Kimura, Tatsuo Nakajima and Ichiro Satoh
Abstract: This paper presents insights on the design of smart artifacts aimed at enhancing everyday life. First, we advocate examining how the quality of everyday life can be enhanced by shifting from gamification toward ritualization. Rather than focusing solely on explicit behavioral goals, it proposes reconsidering how design can introduce ritualized practices that enrich daily experience. To support this discussion, we analyze carefully selected case studies that are well suited to everyday contexts, with the aim of extracting key considerations for incorporating ritualization into design. Through the analysis of these findings, the paper offers novel perspectives for the development of new designs, supported by a concrete example to elucidate the future directions. The primary contribution of this study lies in its proposition of a new design perspective for smart artifacts intended to enhance daily life. It emphasizes the importance of ritualized behaviors and the more-than-human based artifacts associated with them, thus providing a fresh perspective for future design endeavors.
Paper Nr: 20
Title:

Towards Audience-Driven Real-Time Data Storytelling
Authors:

Angelica Lo Duca
Abstract: This paper introduces the concept of Audience-Driven Real-Time Data Storytelling (ADRDS), a framework that utilizes participatory and social sensing to enable audiences to actively shape the presentation of data stories. At its core, ADRDS employs a pre-authored decision tree, where the presenter defines the data story through a branching three-act arc (setup, conflict, and resolution). Each node contains a question, branching rules, thresholds for coverage, consensus, and confidence intervals, as well as one or more potential calls to action (CTAs). Audience members connect to the ADRDS framework via a URL on their mobile devices and answer questions in real time. ADRDS evaluates responses against the decision tree in real-time, following strong branches when thresholds are satisfied or activating fallback paths when signals are inconclusive. ADRDS then selects the final action consistent with the traversed path and the audience’s demonstrated readiness, subject to fairness safeguards. ADRDS transforms presentations of data stories into audience-shaped decision processes, where live data actively drives the narrative. This work paves the way towards a new form of data storytelling, where the final data story depends on the audience's background and interests.
Paper Nr: 21
Title:

Learning about Islam through Virtual Reality: Immersive vs. Non-Immersive Environments
Authors:

Ugur Enes Ates and Enes Yigitbas
Abstract: Misconceptions and prejudice toward Islam remain widespread, partly due to limited opportunities for non-Muslims to engage directly with Islamic communities or visit mosques. Guided mosque tours can reduce prejudice and foster empathy, but access is often constrained by geographic, cultural, or personal barriers. This study develops and evaluates a virtual reality (VR) mosque tour as an immersive alternative for interreligious learning and compares it with a non-immersive educational video. In a between-subjects experiment (N = 24), pre-/post-tests and self-report measures assessed knowledge acquisition, engagement, presence, usability, and cognitive workload. Both VR and video significantly increased factual knowledge, with no statistically significant difference in knowledge gain between conditions. However, VR participants reported higher engagement, although this difference was not statistically significant after correction for multiple comparisons. The VR condition also received high presence and usability ratings and was associated with low perceived workload. These results do not indicate a knowledge advantage for VR. Because the conditions also differed in interactivity, duration, supervision, and testing environment, the observed differences cannot be attributed exclusively to immersion. The findings suggest that immersive tours can provide an engaging and comfortable complement to traditional formats when real-life encounters are not feasible.
Paper Nr: 22
Title:

Expert-Derived Parameters and Requirements for Adaptive Control Elements Using Magnetic Shape Memory Alloys
Authors:

Lukas Fuchs, Marco Hutter, Peter Schmid, Thomas Maier and Bernd Gundelsweiler
Abstract: Modern machines are more often leased or rented leading to the user being confronted with a new user interface often and machines additionally have an ever-increasing number of functions. This challenge is solved for many machines with adaptive graphical user interfaces on touchscreens. However, mobile machinery is controlled mostly with physical control elements, which are not as adaptive yet. Magnetic Shape Memory (MSM) alloys are a promising technology to make physical control elements adaptive. However, so far it is unclear which functionality should be adaptive, which technology readiness different MSM properties have and how MSM can be implemented in an adaptive control element. Therefore, three expert interview series with twelve human machine interface (HMI) experts of original equipment manufacturer (OEM), four of control element suppliers and nine MSM experts have been conducted. The experts see most promising functionality in adaptive force-displacement curves, haptic feedback and integrated sensing. The technology readiness level is for all the main MSM properties apart from integrated sensing in the deployment stage (TRL 6 7). The knowledge derived from the interview will be used to make a founded decision for an MSM-based adaptive control element being developed, tested and evaluated with a user study.
Paper Nr: 23
Title:

From Loss to Restoration: A Progressive Attenuation Chatbot for Grief Support
Authors:

Jiaxun Lin, Xiaoyi Fu and Jiahao Feng
Abstract: With the growing extent of the use of artificial intelligence (AI) in mental health services, chatbots are becoming good tools for grief counseling. However, the present works are mainly focused on the evaluation of clinical effects, and rarely explores how to systemically translate the theories about bereavement to interaction design approaches. This paper is the first application of a grief-support chatbot which systematically utilizes the dual-process model (DPM) as the computational strategy of the gradual reduction of the interaction intensity. Longitudinal empirical results yield two main findings. First, the emotional index ((M_t)) behaves roughly as the DPM predicts. In the acute phase (days 1–10, all 27 participants), (M_t)) stayed high and stable (mean ≈ 82, SD = 4.1). During the integration phase (days 11–20), it fluctuated more deliberately – standard deviation rose to 12.3; one-way ANOVA returned (p=0.058), close to significance. The algorithm aligns broadly with the oscillation mechanism. Second, keyword analysis reveals participants remained predominantly loss-oriented throughout intervention, with few forward-looking expressions; this pattern verifies the chatbot avoids prematurely forcing restoration-focused adjustment. Together, these findings operationalize grief fluctuations into measurable interaction parameters and introduce a "gradual reduction" design grounded in anti-dependency principles. Given the study's exploratory nature and limited sample reaching the final phase (only one participant completed acceptance-phase data collection), these findings are preliminary indicators of the attenuation mechanism's potential. Findings are constrained by short intervention duration, small sample size and scarce acceptance-phase data; further large-scale long-term controlled trials are needed to validate the attenuation mechanism.
Paper Nr: 27
Title:

An Evaluation of Open-Source Browser-Based Webcam Eye-Tracking under Real-World Conditions
Authors:

Gonzalo Alonso-Fernandez, Sara Vecino, Martin Gonzalez-Rodriguez, Javier De Andrés and Daniel Fernandez-Lanvin
Abstract: This study evaluates two open-source webcam-based eye-tracking libraries, WebGazer.js and EyeGesturesLite, against the Tobii Pro Nano infrared tracker as a reference standard across 45 participants performing 6 tasks. A preprocessing pipeline combining temporal alignment, spatial affine correction, and extreme value removal was applied before analysis. Euclidean distance measurements showed WebGazer.js achieved a median gaze error of 4.23°, with no statistically significant deviation from Tobii on either axis, indicating the absence of systematic spatial bias. EyeGesturesLite reached a median error of 7.84° and exhibited significant bias on both axes. For fixation detection using the Dispersion-Threshold Identification (I-DT) algorithm, WebGazer.js fixation centroids deviated 4.41° from the Tobii reference on average, roughly half the 8.87° deviation of EyeGesturesLite. These results quantify the accuracy cost of choosing zero-cost, browser-deployable alternatives over dedicated hardware and provide actionable guidance. WebGazer.js constitutes a viable low-cost option for usability studies where areas of interest (AOIs) are large enough to absorb a 4-5° margin, while neither library is appropriate for fine-grained fixation analysis or precision AOI mapping.
Paper Nr: 45
Title:

When Gameplay Becomes Stressful: A Preliminary Exploration of Sensor-Free Stress Modelling in Virtual Reality Shooter Games
Authors:

Susanna Brambilla, Giovanni Morlacchi, Giuseppe Boccignone, Alberto N. Borghese and Laura Anna Ripamonti
Abstract: Can a model trained only on low-pressure gameplay detect behavioural changes associated with stress without physiological sensors or stress labels during training? We investigate this question in a controlled virtual reality (VR) first-person shooter using an LSTM-based variational autoencoder (LSTM-VAE). The model was trained on tutorial sequences and evaluated during a subsequent survival phase using gaze, head and controller kinematics, or button inputs. Reconstruction error was converted into a continuous baseline-deviation score and compared with retrospective DANTE stress annotations from N = 38 participants. The strongest condition, HMD+MC/Single, showed a statistically reliable but small association with self-reported stress (r¯ = .214, approximately 4.6% shared variance), whereas binary separability remained close to chance. These findings do not support a deployable stress detector or establish that the recurrent variational architecture is superior to simpler alternatives. They instead provide preliminary evidence that native VR telemetry contains a small, consistent stress-related signal. Predictive processing and free energy motivate the baseline-deviation formulation, while reconstruction error is the operational quantity evaluated in the experiments.
Paper Nr: 51
Title:

Hofstede’s Cultural Dimensions, Semiotics and Search User Interface Design
Authors:

Karen Chessum, Ingo Frommholz and Haiming Liu
Abstract: The cultural background of an information seeker potentially influences their preferred user interface (UI) design. This study looks at how Geert Hofstede’s 6 cultural dimensions together with semiotics have been applied to search engine (SE) user interface designs. A study by 25 participants from various cultural backgrounds has been analysed to establish if Hofstede’s cultural model can be used to convey meaning and meaning making through acts of signification to users. The key findings from this from our small-scale study have shown that Hofstede's cross-cultural dimensions can be used to convey meaning and meaning making through semiotics i.e. differential decoding interface signs. We acknowledge that Hofstede's cultural model has been the subject of ongoing critique: nevertheless, we demonstrate herein, the validity of Hofstede's model in providing a sound referent framework for interface signs. Given the small sample size, we freely acknowledge that further follow up studies using larger subject groups will be needed to fully support and validate the findings of our study.
Paper Nr: 52
Title:

Cognitive Debt in the IDE: AI Coding Assistants and the Engagement Threshold Model
Authors:

Mashail N. Alkhomsan
Abstract: AI coding assistants have achieved rapid adoption in professional software development. The main discourse frames this shift in terms of productivity. Yet a growing body of empirical evidence reveals a more complex picture: perceived productivity gains frequently diverge from measured outcomes, sustained reliance on AI-generated code introduces downstream maintenance costs, and developers across experience levels report meaningful changes in how independently they approach problems. This paper advances a conceptual framework for human–AI interaction in software development. The Engagement Threshold Model (ETM), which argues that the cognitive consequences of AI assistance are determined not by how much a developer uses AI but by the character of that use. Specifically, whether their posture toward AI output is active and interrogative or passive and accepting. A cross-sectional survey of 48 practicing developers provides empirical grounding for the model. The findings consistently support the ETM's central prediction: developers who interrogate AI output report substantially lower perceived skill atrophy than those who accept it passively, regardless of how heavily they use AI tools. Critical evaluation of AI output emerges as the mechanism through which posture exerts its effect; metacognitive awareness operates as a protective buffer. Implications for tool design, code review practice, and developer education are discussed.
Paper Nr: 56
Title:

Bridging the Gap: Intuitive LLM-Based Interfaces for Enhanced Cognitive Care
Authors:

Marco Ajovalasit, Tejas Sanjay Bhanage, Ziwei Ding, Zhiqiang Huang, Fabio Salice and Sara Comai
Abstract: As population aging intensifies, caregivers supporting older adults with cognitive decline, especially those living alone, face increasing challenges in monitoring behaviour, interpreting data, and maintaining emotional connection. This study presents CareSense Narrator, a human-centred system that integrates non-intrusive environmental sensors with large language models (LLMs) to transform raw data into contextualised care narratives. Designed for collaborative care, it helps caregivers understand individuals’ levels of independence, prevent critical situations, and enable personalised interventions. The system processes behavioural and physiological data through four stages: indicator prioritisation, semantic interpretation, recommendation. generation, and knowledge delivery. It emphasises empathetic communication through adaptive language, visual elements, and voice interaction, enhancing both clarity and emotional resonance. This work highlights a novel integration of LLMs with human-centred design, improving decision-making efficiency and care confidence. It also provides early evidence of the value of contextual narratives, proposing a new paradigm for care technologies that combine intelligence with empathy and human connection.
Paper Nr: 66
Title:

Exploring Visuo-Passive Haptic Garnishes to Enhance VR Drinking Experience
Authors:

Phi-Vu Vo-Diep, Chi-Thanh Vi, Hoai-Nam Do and Khanh-Duy Le
Abstract: The immersive visualization capabilities of virtual reality (VR) have been increasingly explored to enhance beverage consumption experiences. Prior research has demonstrated that combining virtual manipulation of beverage or environmental colors with olfactory cues can influence users’ taste perception of plain water, thereby mitigating the effects of beverage unavailability due to cost, geographical constraints, or seasonal limitations. To further advance this aroma-enhanced VR drinking approach, we conducted a within-subject user study involving 14 participants to investigate the effects of incorporating visual and passive haptic garnish cues in the form of virtual fruit slices. While scent significantly affected taste and liking, we did not detect statistically significant effects of the combined visuo-haptic garnish manipulation on taste, liking, or presence. However, some participants described the visual garnish as contributing to realism and aiding flavor identification. These qualitative observations suggest a potential visual role for garnish that warrants further investigation.
Paper Nr: 80
Title:

Working with Unclear Automation: Conditional Trust and Cautious Coordination in Mixed Fleet Logistics
Authors:

Veronika Žigraiová, Aparajita Chowdhury and Aino Ahtinen
Abstract: Automation is transforming logistics and warehouse work, where human workers and automated systems increasingly share physical workspaces. In these mixed fleet environments, trust is not only an attitude toward automation, but part of practical coordination: workers must interpret system behaviour, anticipate movement, and decide when to rely, wait, avoid, or intervene. This paper examines how trust, uncertainty, and control were described by industry specialists and adult logistics students in relation to warehouse automation. Drawing on scenario-grounded semi-structured interviews with 11 participants, we analyse how participants made sense of unclear automation behaviour and described maintaining safety and coordination when trust was incomplete. The findings show that uncertainty emerged from ambiguous signals, unexplained pauses, limited interpretability, and difficulty anticipating system intent. Rather than leading simply to distrust, these moments prompted cautious coordination strategies, including slowing down, keeping distance, monitoring the system, and wanting intervention possibilities. The analysis also distinguishes experience- based and expectation-based accounts: industry specialists grounded trust in workflow and implementation concerns, while logistics students emphasized anticipated risks, training needs, and limited exposure. The paper contributes a focused account of conditional trust and cautious coordination in warehouse automation, highlighting the need for actionable communication, perceived control, and informed human intervention.
Paper Nr: 81
Title:

Planning Assistance System Transparency for Search and Rescue Missions
Authors:

Jérémy Monguin, Rafik Belloum, Marie-Pierre Pacaux-Lemoine and Emmanuelle Grislin-Le Strugeon
Abstract: Human-robot cooperation in dynamic and high-risk environments increasingly relies on intelligent systems to support task allocation and coordination. While existing cooperation frameworks offer useful models to structure agent roles and autonomy levels, they provide limited support for the operator's understanding of why a plan with a certain task allocation is proposed, why alternatives are discarded, and how to compare competing options. This limitation is particularly critical in contexts such as search and rescue, where operators must make rapid yet informed decisions under uncertainty. We argue that effective and well-calibrated human-AI cooperation requires moving beyond task allocation alone, toward explanation-aware decision support. Building on the Know-How-to-Operate and Know-How-to-Cooperate model, automation transparency, and reliance-oriented explainable interfaces, we introduce CoSight, a decision support system for multi-robot navigation involving one operator, two robots, and a Planning Assistance System. Our approach distinguishes three progressive levels of information (descriptive, justificatory, and comparative) grounded in the SAT model, and establishes a theoretical link between these levels and the cooperative competencies of the assistance system, a connection not yet explicitly operationalized in prior work. We also argue that this information should be presented adaptively based on the operator's behavior and cognitive profile. This methodological paper presents the conceptual framework, system design, and experimental protocol of CoSight. The protocol has not yet been evaluated with participants; its forthcoming empirical validation will investigate how progressive transparency affects appropriate reliance on intelligent coordination systems.
Paper Nr: 87
Title:

Who Bears the Burden of Decentralization? A Socio-Technical Responsibility Allocation Framework for Alternative Social Media
Authors:

Krsna Pinheiro de Souza Gusmão, Daniel Schneider, Jano Moreira de Souza, Luiz Filipe Brandi and Guilherme Moura
Abstract: The structural capture and algorithmic monopoly of Centralized Social Media have encouraged technology studiers and builders to develop decentralized alternatives, and users to migrate their communities toward those new networks. However, while backend decentralized software architectures are scaling successfully, mainstream user adoption remains bottlenecked by high system friction. Centralized platforms remove interface complexity by abstracting data routing and moderation tasks away from the user, whereas decentralized systems open up this stack, frequently transferring the operational weight directly onto human labor. This work introduces a Socio-Technical Responsibility Allocation Framework to evaluate alternative digital social networks through a human-computer interaction (HCI) lens. Synthesizing recent empirical literature, we track how essential system tasks are distributed across a four-tier landscape: the high cognitive overhead and network literacy costs forced onto the end-user; the asymmetric moderation tool deficits and operational burnout borne by community volunteers; the boundary parameters codified directly by the native protocol architecture; and the collaborative verification mechanics provided by the automation technology - specifically decentralized federated learning and cryptographic reputation engines. The synthesis highlights a recurring trade-off: greater formal autonomy can increase user and volunteer workload. We propose design directions for redistributing this work while retaining human oversight; their effects require empirical validation.
Paper Nr: 88
Title:

Human-Centered AI in the Global South: A Brief Survey of Concepts, Methods, and Domains
Authors:

Rodrigo Adamski, António Correia, Simone C. dos Santos and Daniel Schneider
Abstract: The adoption of Human-Centered AI (HCAI) principles in the Global South has generated a growing but fragmented body of literature cutting across computer science, social computing, science and technology studies, and development informatics. This survey synthesises findings from 103 peer-reviewed papers published between 2019 and 2025. Through inductive thematic analysis we identify five empirically grounded framings through which HCAI is conceptualised in Global South contexts: Postcolonial/Decolonial, Culturally Adaptive, Operational/Implicit, Epistemic Justice, and Governance & Institutional Accountability. We map the methodological repertoire employed across the corpus, identify eight primary application domains, and derive a structured research agenda comprising eight priority gaps. The survey suggests that HCAI in the Global South may constitute a distinct theoretical and applied terrain — shaped by infrastructure constraints, legal pluralism, epistemic asymmetries, and state capacity deficits — that standard HCAI frameworks have yet to adequately address.
Paper Nr: 94
Title:

Developing GLEE: A Practice-Oriented Framework for Game-Based Learning Experience and Evaluation
Authors:

Laura Herrewijn, Sven Charleer, Yasmine Wauthier, Ellen De Bruyne, Veerle Van Assche and Gert Vanthournout
Abstract: Serious games must provide an engaging player experience while achieving their intended learning outcomes, yet development teams often lack practical ways to connect design intentions, game user experience, and learning evidence. This paper presents the Game-based Learning Experience and Evaluation (GLEE) framework: a conceptual and methodological framework for designing, evaluating, and improving game-based learning experiences. GLEE was developed through a focused conceptual review, interviews with sixteen serious game professionals from nine organizations, and a pragmatic review of measurement instruments. The resulting framework distinguishes Game User Experience from Learning Processes and Outcomes and organizes relevant constructs into four analytical groupings: Design Qualities, Experiential Responses, Subjective Perceptions, and Objective Outcomes. It provides a modular approach for selecting measures and combines post-play self-report with a game-specific pretest--posttest procedure. The contribution of this paper lies in articulating the framework, its conceptual foundations, components, and proposed evaluation procedure. GLEE is being tested and iteratively refined through a series of case studies. Three completed applications across different games, learner groups, and educational settings provide preliminary support for the feasibility of its modular measurement approach. Detailed empirical evaluations of the individual games fall outside the scope of this paper and will be reported separately. Future work will focus on comparative validation across contexts and on implementing and evaluating GLEE as a digital tool for research and practice.
Paper Nr: 96
Title:

Perceived versus Actual Usability of Computer Game User Interfaces
Authors:

Piotr Tokarski, Karol Łazaruk, Monika Korzeniewska, Karolina Rybak, Małgorzata Plechawska-Wójcik and Mariusz Dzieńkowski
Abstract: The usability of game user interfaces directly influences player performance and experience, yet it is usually assessed solely through self-report. This study combined eye tracking with two subjective instruments, the System Usability Scale and a forced-choice comparative survey, to compare the perceived and the actual usability of four commercial games from different genres: Fortnite (battle royale), Genshin Impact (story-driven action role-playing), Infinity Nikki (adventure with dress-up elements), and The Sims 4 (life simulation). Twenty participants located common interface elements while their eye movements were recorded. The interfaces differed significantly on almost every objective measure. The largest differences concerned the organization of visual scanning, whereas differences in target localization were the most consequential, determining whether an element was found at all. Perceived and actual usability nevertheless diverged. Infinity Nikki supported search best yet was ranked only third, Fortnite was judged best despite intermediate performance, and only The Sims 4 was rated as poorly as it performed. At the level of individual players, judgments matched measured performance in barely half of the cases and tracked how quickly an element first drew the gaze rather than how efficiently the task was completed. This mismatch was systematic rather than random, and the two subjective instruments were internally consistent, indicating a coherent impression that did not follow task performance. Self-report and behavioral measures should therefore be treated as complementary rather than interchangeable when evaluating game interfaces.
Paper Nr: 11
Title:

Gamified Offline STEM Learning in Low-Connectivity Contexts: Motivation, Anxiety, and Design Implications for LMSs
Authors:

Owen Luyando Chilwalo and Josephina Mikka-Muntuumo
Abstract: Gamification is the use of game design elements in non-game contexts. Science, Technology, Engineering, and Mathematics (STEM) education in many schools is constrained by unstable internet connectivity, power cuts (load shedding), and low learner engagement. Popular Learning Management Systems (LMSs) often assume stable connectivity and are difficult to use in such settings. We designed, developed, and evaluated a gamified LMS with offline access for Grade 10–11 STEM learning in a low-connectivity school context in Namibia. The system is implemented as a Progressive Web Application (PWA) with local storage and background synchronisation and includes interactive lessons, quizzes, points, badges, progress indicators, a leaderboard, and an embedded browser-based coding environment. In an exploratory qualitative field study with two STEM teachers and eleven learners, gamification was associated with increased learner motivation and participation, but some learners reported anxiety when performance was visible to peers. Offline-first design reduced stress related to data, internet, and electricity availability. These findings are promising indications from a small-scale study rather than strong evidence of impact. The paper contributes a contextual account of how offline-first architecture and gamification interact with learner emotions in a low-resource classroom, and derives design implications for STEM learning platforms in low-connectivity environments.
Paper Nr: 18
Title:

Exploring the Effects of Controlled Pain Stimuli on User Immersion and Enjoyment in VR Gaming
Authors:

Jannes Kanwischer and Enes Yigitbas
Abstract: This work investigates the potential impact of simulated pain experiences on user immersion and enjoyment within a virtual reality (VR) gaming environment. Although research is continuously examining various sensory inputs, the role of pain, a fundamental component of human physiology and a key mechanism for real-world adaptation, remains largely underexplored in VR. Therefore, the omission of pain stimuli in VR may overlook opportunities to deepen the user experience, potentially limiting immersion and emotional impact. To address this gap in research, we introduce controlled pain stimuli through electrical muscle stimulation (EMS) as part of a VR game experience. We had nine participants engaging with a controlled VR environment under three experimental conditions: no pain, moderate pain, and strong pain. The pain stimuli were administered using an EMS vest, with stimulus intensity individually calibrated for every participant prior to each experience. In this exploratory within-participant study (n=9), we descriptively examined how the three conditions related to self-reported immersion and enjoyment-related responses. Within this sample, immersion scores tended to be higher in the pain conditions. Moderate pain was most often selected as the most enjoyable condition, whereas strong pain was most often selected as both the most immersive and the least enjoyable.
Paper Nr: 36
Title:

Artificial Intelligence Driven Emotion-Aware Public Speaking Coach
Authors:

Sofia Condesso, Nuno Leite and Artur Ferreira
Abstract: Emotion recognition enables the use of emotion-aware systems. However, the design of feedback mechanisms translating recognition outputs into actionable, grounded coaching is still in its early stages. In this paper, we investigate how persona conditioning and prompt complexity influence emotion-aware coaching feedback generated from structured emotion timelines in two types of public-speaking: academic and pitch presentations. We present a local-first framework in which users upload presentation recordings, processed through speech and text emotion recognition pipelines. This information is then provided to the Coach Agent powered by a Large Language Model (LLM) via persona-conditioned prompts to generate coaching feedback. Two coaching personas (Academic Advisor and Pitch Mentor) are evaluated with Simple and Detailed prompting strategies, with different instructional constraints. Responses are assessed through a LLM-as-a-judge protocol, using a 5-point Likert scale. The experimental results show that persona conditioning affects feedback style and evaluation scores. Pitch Mentor performs best in both types of presentation being less dependent on complex prompting than the Academic Advisor. Moreover, the Detailed Prompting strategy often benefits the Academic Advisor, whereas Simple Prompting strategy achieves better results for the Pitch Mentor. The code repository can be found at GitHub, https://github.com/sofiafernandescd/EmoReA.
Paper Nr: 37
Title:

DiversityLens: A Tool for Automated Demographic Analysis and Visualization of Social Datasets
Authors:

Adrian Auer, Veli Ates, Mehmed Yüksel and Erik Kamsties
Abstract: DiversityLens is an automated auditing tool for face datasets that makes upstream demographic composition analysis practical and reproducible. The tool recursively scans an image or video dataset directory, applies face detection using two interchangeable backends, and uses DeepFace to estimate age, gender, and race for every detected face. Audit outputs include a structured CSV record and an interactive HTML dashboard. We evaluate six publicly available and frequently used face datasets: UTKFace, All-Age-Faces, FairFace (training and validation splits), CelebA, FFHQ, and VVAD-LRS3. Each dataset is analysed with both the CPU-friendly cascade detector and a GPU-accelerated single-shot detector, yielding fourteen backend-aware audit runs. The results document significant demographic imbalances and compound intersectional underrepresentation of largest-to-smallest intersectional subgroup ratios of up to 1000:1 across widely used datasets. A ground-truth accuracy calibration on UTKFace quantifies the reliability of the estimated demographic labels, and an intersectional analysis highlights compound underrepresentation that single-attribute summaries can conceal. The results show that DiversityLens can produce consistent, backend-aware characterisations of dataset composition and reveal representation gaps across widely used face corpora. The tool is positioned as a practical instrument for upstream dataset auditing, particularly in Human-Robot Interaction workflows that rely on face-based perception models. DiversityLens is open-source and available at https://pypi.org/project/diversitylens/.
Paper Nr: 49
Title:

Characterising User Engagement in a Virtual Human Coach–Based Diabetes Prevention Web App: Interim Findings from the VA|PREVENTION Study
Authors:

João Balsa, Isa Brito Félix, Maria Beatriz Carmo, Matilde Silva, Diana Margarida Carlos, Raquel Quintãos, Tomás Fonseca, Cristiana Firmino, Helga Rafael Henriques and Mara Pereira Guerreiro
Abstract: Lifestyle change is central to Type 2 Diabetes (T2D) prevention. Digital be-haviour change interventions (DBCIs) offer a scalable alternative to traditional face-to-face approaches. A critical aspect is to ensure sufficient engagement to deliver the intended benefit. Engagement should therefore be explored using standardised behavioural indicators. We developed a diabetes prevention web application to support behaviour change, focusing on diet and physical activity. The web app comprises three main components: (i) a virtual human coach designed to interact with users daily and support behaviour change through predefined behaviour change techniques; (ii) educational menus on healthy eating, physical activity and sedentary time; (iii) dashboards to monitor users’ personal progress, track behavioural changes over time and get support from healthcare professionals and members of their social network. The effectiveness and implementation of the web app is currently being evaluated in a 10-month trial involving 180 individuals. We report interim findings on user engagement with the general web app and the embedded virtual human coach over the first 4-month period. We defined a set of engagement variables, aligned with the Intervention Engagement Ontology, to extract from usage data. In this study, we include variables related to the number of logins and the time spent interacting with the virtual coach. Machine learning methods for clustering and rule induction were then applied to this dataset to find user engagement patterns and learn user profiles. We found that there is a clear distinction between more engaged and less engaged participants, that allowed us identify rules that relate the level of engagement with some of the participants’ sociodemographic features.

Area 4 - Interaction Design

Full Papers
Paper Nr: 25
Title:

Gamified Interaction Design in Programming Education: Examining the Motivational Effects of Competition, Leaderboards and Collaboration
Authors:

Marisa Venter
Abstract: Gamification has emerged as a promising pedagogical strategy in programming education, yet the differential motivational effects of specific game mechanics such as competition, leaderboards and collaboration remain insufficiently understood from an interaction design perspective. This study examines how these three game mechanics influence the intrinsic motivation of first-year Internet Programming students at a South African university, using a concurrent mixed-methods design comprising a quantitative survey and qualitative group interviews with students categorized according to the HEXAD user typology. Structural equation modelling revealed that competition and leaderboards are significant positive predictors of intrinsic motivation, while collaboration was not significant in the full structural model, though it demonstrated a significant independent effect when examined in isolation, highlighting the importance of examining game mechanics simultaneously. Qualitatively, student preferences varied substantially by user type, with only philanthropists showing a clear preference for collaboration while the remaining four user types displayed balanced preferences, demonstrating that aggregate preference data can be misleading without user-type disaggregation. The study contributes a theoretically grounded mixed-methods analysis of competitive and collaborative game mechanics in programming education and offers practical interaction design guidelines for educators seeking to optimize gamification in this context.
Paper Nr: 31
Title:

Engaged but Distracted: Mind Wandering and Conversational Activity in Avatar - versus Chatbot-Based Career Counseling
Authors:

Laura Maas, Yvonne Kammerer and Jürgen Seitz
Abstract: Conversational AI systems increasingly offer different output modalities, such as text-based chatbots and voice-based avatars, yet their differential effects on user attention and cognition during task-oriented interactions remain poorly understood. This between-subjects study (N = 108) examined whether an interactive voice-based AI avatar and a text-based chatbot differed in focused attention, mind wandering, extraneous cognitive load, conversational activity, and comprehension during a structured career-orientation conversation. While the conditions did not significantly differ in focused attention, extraneous cognitive load, or comprehension, the avatar condition resulted in significantly higher self-reported mind wandering despite greater conversational activity. These findings suggest a dissociation between attentional and behavioral engagement that is further reflected in the absence of comprehension benefits in the avatar group compared to the chatbot group. Content analysis of session transcripts corroborated this pattern: avatar users showed a significantly higher proportion of off-topic conversation than chatbot users. Additional exploratory analyses indicated that mind wandering and extraneous cognitive load were elevated specifically among avatar users with high conversational activity. Notably, prior avatar experience was associated with lower mind wandering and higher focused attention in the avatar group, pointing to novelty as a plausible explanation for the observed effects.
Paper Nr: 32
Title:

Scenario-Based Expert Evaluation of Digital Nudges in Learning Management Systems
Authors:

Thorleif Harder
Abstract: Digital nudges in learning management systems (LMS) are increasingly used to support students in self-regulation, collaboration, and exam preparation. However, existing research has primarily measured behavioral effects after implementation, leaving design quality criteria such as comprehensibility, situational appropriateness, and legitimacy unexamined before deployment, when autonomy-critical flaws are difficult to revise. This study reports a formative, scenario-based expert evaluation of ten digital nudges for a Moodle plugin. Eight HCI experts assessed five nudge groups through asynchronous prototype annotation and an online questionnaire embedding scenarios and Likert items. While the nudge concepts received predominantly positive ratings (87 % of Likert ratings ≥ 4), the qualitative analysis consolidated 285 comments into 39 design measures, revealing concept-level acceptance despite implementation-level issues. These measures were synthesized into five design implications for autonomy-preserving nudges in LMS, spanning interaction clarity, data transparency, and cognitive-emotional fit. The findings suggest that scenario-based expert evaluation can reveal design weaknesses that remain invisible to behavioral outcome measures.
Paper Nr: 71
Title:

Victims of Bernburg: Reflective AR Storytelling in a Memorial Context
Authors:

Camilla Vang Østergaard, Chris Hall and Stefan Jänicke
Abstract: Memorials dedicated to victims of Nazi persecution face a unique challenge when introducing digital technologies into spaces of remembrance. While technologies such as Augmented Reality (AR) can enrich historical interpretation and provide access to archival material, they must be designed in ways that preserve the dignity of the victims, respect the memorial context, and support meaningful visitor reflection. This article presents the design, development, and formative evaluation of an AR-based mobile application for a memorial site associated with the Nazis' "euthanasia program," Aktion T4. It explores how location-based AR can be designed to provide visitors with additional historical context through witness-driven narratives reconstructed from digitized archival material and linked to the physical environment. The system was developed in Unity using image-based interactions to create an accessible digital experience for standard mobile devices. This study examines how the design of AR experiences can support visitor engagement in memorial environments by strengthening the connection between visitors and the lives of victims of Nazi persecution. The article provides insights into the design considerations associated with developing AR experiences for sensitive cultural heritage, including how technological possibilities can be balanced with historical accuracy, respectful representation of the victims, and opportunities for reflection and learning.
Short Papers
Paper Nr: 13
Title:

Dream Complex: The Integration of LLMs and AR for Mapping Subconscious Dreams into Non-Linear 3D Narrative Structures
Authors:

Ruoyun Dai
Abstract: Dream visualization represents a significant challenge in psychological research due to the ephemeral and non-logical nature of the subconscious. This paper presents a multimodal framework that integrates human-computer interaction and augmented reality to transform verbal dream records into immersive, navigable 3D environments. The system employs an intelligent generative pipeline, utilizing a three-dimensional coordinate system mapping emotional valence (x-axis), arousal amplitude (y-axis), and logical consistency (z-axis) to automate the spatial synthesis of narratives derived from user descriptions. By leveraging large language models to analyze verbal records for sentiment and logical consistency, the framework maps extracted entities into a non-linear spatial structure inspired by the literary theories of Italo Calvino. Participants interact with a physical installation through digital overlays, scanning specific markers to reconstruct the emotional fragments of the dream. Ultimately, this framework serves as an exploratory human-computer interaction prototype. While future iterations hold potential for clinical settings upon integration with biometric data, the current system functions primarily as a novel medium for personal psychological self-reflection and artistic visualization of subconscious narratives.
Paper Nr: 14
Title:

A Usability and Universal Design Investigation into Carousel Interaction in Mobile Contexts
Authors:

Hasti Shayganfard and Pietro Murano
Abstract: This paper presents details of an empirical study into the use of web page carousels in mobile user interfaces. Two web page user interfaces were designed to simulate an online clothing shop. One user interface was implemented with carousels and the other did not have carousels. Thirty participants were recruited for the study, which included one participant with muscular dystrophy. The data that was collected and statistically analyzed suggests that using carousels on web pages for mobile interaction fosters worse performance and user experience than an equivalent web site without carousels. An interview was conducted with the participant with muscular dystrophy and the results suggest that the carousels were easier to use for this particular participant. However, automatic carousels are not easier to use. The results suggest that a single solution might not fit all users. Based on the results, we suggest three possibilities that would require further investigation and larger sample sizes. The first is to implement a solution with the default of no carousels, but with an option to switch them on. The second option is to structure web pages differently, where there would be less long vertical scrolling and less interaction elements located at the extremities of the screen. The third would be to design and evaluate completely new controls that could be suitable for all users.
Paper Nr: 39
Title:

Observational Contour Drawing in Mixed Reality with HoloLens 2
Authors:

Hang Wu, Ruding Lou and Nordin Zakaria
Abstract: Observational contour drawing — the practice of representing perceived edges and spatial relationships from direct observation — has long been confined to 2D surfaces. This paper explores whether mixed reality (MR) can open a new dimension for this practice. We present an MR application for HoloLens 2 that enables bare-handed 3D drawing through pinch-gesture interaction, with a complete set of editing operations and an exploratory user study comparing 3D MR drawing to traditional 2D paper drawing. A study with 19 participants yielded a mean overall satisfaction score of 3.68/5 (SD = 0.48), providing preliminary validation of the core interaction model, and identified key areas for improvement: drawing precision, interface ergonomics, and advanced feature discoverability.
Paper Nr: 68
Title:

Designing Sustainable Interaction: The Role of Shape Design in Vehicle Interfaces
Authors:

Lisa-Marie Dyka, Julia Gritzbach, Daniel Holder, Lukas Fuchs and Wolfram Remlinger
Abstract: Sustainability is becoming an increasingly important focus in interaction design, particularly in vehicle interiors. Given that shape design plays a crucial role in user interaction, this study examines how the geometry of interfaces can contribute to the United Nations' Sustainable Development Goals (SDGs). Drawing on qualitative transdisciplinary expert interviews across design and multiple SDG domains, the study identifies key areas of improvement for sustainable design, illustrated through conceptual examples. The findings indicate that shape design enables the translation of abstract sustainability values into actionable design guidelines centered on human, environmental, and economic considerations. In this way specific shape characteristics can be linked to interaction qualities such as perception, accessibility, and safety. Building on this, the paper outlines a methodological approach for a sustainable and user centered design with applicability beyond vehicle interiors.
Paper Nr: 72
Title:

Multilayer Music Representation for Accessible and Interactive Digital Archives: The ``Il Pazzariello'' Platform
Authors:

Martina Gremignai and Luca Andrea Ludovico
Abstract: The digitization of cultural heritage has greatly increased access to archival collections, but access alone does not ensure understanding. This issue is especially relevant for musical heritage, where digitized sources often require specific competences and do not convey the temporal, sonic, and contextual dimensions of music. This paper addresses the role of human-computer interfaces in mediating such materials, with particular attention to heterogeneous archival sources and non-specialist audiences. The proposed approach relies on IEEE~1599, an XML-based standard for multilayer music representation, whose structure supports the synchronization of symbolic notation, graphical score images, audio recordings, metadata, and additional multimedia resources. The paper discusses this approach through the ``Il Pazzariello'' web platform, developed within a digital humanities initiative on historical musical sources from the Modena area and designed for middle-school students as a representative non-specialist audience. Beyond the specific case study, the contribution frames IEEE~1599-based interfaces as a replicable model for digital music archives, capable of connecting heterogeneous sources, supporting source-based historical learning, and promoting more accessible, open, and inclusive forms of engagement with musical heritage.
Paper Nr: 73
Title:

Analysis of the Influence of the Subgestalts on the Formal Design of Products
Authors:

Pascal Rommel, Lars Gadermann, Daniel Holder and Thomas Maier
Abstract: This paper presents a conjoint analysis to evaluate the impact of subgestalts on the formal design of a product and its perceived value. For this purpose, an online survey was conducted using eleven models, of which nine were defined through an orthogonal design plan, while two additional models were included as hold-out profiles. The results from n = 138 participants were evaluated using a conjoint analysis. The findings show that the subgestalt layout, with an importance of 37,040%, has approximately twice the influence on whether a gestalt is perceived as one unified whole compared to other subgestalts, whose importance are each around 20%. This demonstrates that the subgestalts shape, colour & surface and graphic also have a substantial impact on whether the gestalt is perceived as one unit. The evaluation of the perceived value shows a similar pattern. Layout has the highest importance, at 31,425%, while the other subgestalts are around 23% and are therefore much closer to one another. In this context, the part-worth utilities are particularly relevant, as they indicate a preference for an integral design of the subgestalts layout and colour, whereas an additive design is preferred for the subgestalts shape and graphic. These findings suggest that broader subgestalts, which more strongly define the overall frame of the gestalt, are perceived as more aesthetical, when they have a unified appearance. In contrast, more detail-oriented subgestalts tend to be perceived as more aesthetical when designed additively.
Paper Nr: 86
Title:

Affective Polarization De-Escalation: A Framework for Hyper-Personalized Digital Nudges in Synthetic Societies
Authors:

Wladimir A. P. das Neves, Angélica F. Dias, Simone C. dos Santos, Antonio Correia and Daniel Schneider
Abstract: The contemporary informational ecosystem, heavily shaped by algorithmic curation, systematically fosters filter bubbles and echo chambers, exacerbating political polarization. Interventions based on behavioral economics, such as Digital Nudges, face the "Interface vs. Structure Paradox": while visual interface nudges are ethically transparent, they often trigger the Boomerang Effect when colliding with the confirmation biases of radicalized users. Conversely, structural interventions (e.g., Random Dynamical Nudges) are highly effective at disrupting echo chambers but border on manipulative dark patterns, posing risks to the democratic public sphere. To overcome the methodological limitation of treating users homogeneously, this position paper proposes an experimental framework for hyper-personalized sociotechnical interventions. By integrating Large Language Models (Local LLMs) to ensure Privacy by Design, the adaptive system dynamically extracts the user's psychometric "latitude of diversity" to mathematically calibrate the injection of diverging content. Simultaneously, the framework uses advanced prompting techniques to generate customized metacognitive explanations (Boosting) within the interface, ensuring ethical transparency. We outline an in-silico experimental protocol using LLM-based Multi-Agent Systems (MAS) designed to validate the framework, aiming to bridge the methodological chasm by unifying computational network topology with Natural Language Processing (NLP) assessments of affective polarization.
Paper Nr: 93
Title:

Design for Sense of Place: A Reflective Design Analysis of Two Place-Based Artifacts
Authors:

Gul Sher Ali, Gaute Brandser, Daniel Jan Axlid and Sobah Abbas Petersen
Abstract: Research work on place-based digital technologies addresses sense of place as a design concern for helping people feel connected to outdoor public places. Yet evaluations tend to focus on individual experience and give less attention to how places become shared, plural and sometimes contested. We address this gap through a reflective design account of two place-based artifacts: an Augmented Reality application for place-anchored personal narrative and a Geographic Information Systems platform for gathering in-situ urban perception. We analyze the artifacts together as an annotated portfolio, analyzing each through two analytical instruments: four design resources for reflection in Human-Computer Interaction and a conceptual framework for technology and social wellbeing in built environments. Through a reflective analysis of the two artifacts, we explore how place-based technologies addresses sense of place. Our main contribution is the cross-artifact finding that participants in both evaluations reported they would withhold views diverging from the prevailing tone they perceived in others’ visible contributions. Then three design considerations follow from these findings. These design considerations are candidate intermediate-level knowledge that could support plural rather than convergent voice for sense of place in place-based technologies.
Paper Nr: 97
Title:

Beyond WCAG Compliance: Combining Accessibility Audits with Experiences of Users with Visual Impairment: A Case Study of a Municipal Website
Authors:

Karol Łazaruk, Piotr Tokarski, Marek Miłosz, Mariusz Dzieńkowski, Sergio Luján-Mora and Małgorzata Plechawska-Wójcik
Abstract: Public-sector websites in the European Union are legally required to conform to the harmonised European standard EN 301 549, which incorporates the Web Content Accessibility Guidelines (WCAG). Formal conformance, however, does not by itself guarantee a usable experience for people with disabilities. This paper reports an exploratory instrumental case study of the municipal portal of the City of Lublin, Poland (lublin.eu). The portal is a public-sector website subject to this statutory obligation. The audit combined two stages. In the first, eighteen independent evaluator teams applied twenty-six WCAG success criteria to a set of pages selected according to the official European monitoring methodology, using automated tools together with manual inspection. In the second, two users with visual impairment (low vision) explored the same portal, and the barriers they encountered were analysed qualitatively. The audit identified contrast as the most frequently failed criterion. The users independently confirmed contrast as their central difficulty but also encountered barriers the audit did not surface, and they valued features that were not addressed by any of the criteria. Some of these lay on pages outside the sample, some outside the audit protocol, and a few outside what the criteria themselves can express. The users also identified a design feature that helped them in practice but is not covered by the success criterion set. These findings add to the evidence that formal WCAG evaluation and feedback from users with disabilities are complementary: the audit locates technical defects, whereas users reveal which defects matter in practice and expose barriers a checklist-based audit can miss. Distinguishing the barriers the audit did not reveal from those it could not reveal shows which part of that gap a broader audit would close and which part it would not.
Paper Nr: 100
Title:

Beyond Paperless Receipts: Digital Receipts as Reflective Consumer Journey Infrastructure
Authors:

Thomas Neifer and Kerstin Salewski
Abstract: Digital receipts are often treated as paperless substitutes for proof-of-purchase documents. This paper develops a broader conceptualization of digital receipts as reflective consumer-journey infrastructure in grocery practices. From a consumer-journey perspective, grocery purchases remain relevant after checkout because products become part of household routines, budgeting, dietary reflection, storage, waste prevention, and sharing. We synthesize prior work on digital receipts, Integrated Household Information Systems, food-related HCI, reflective HCI, personal informatics, sustainable HCI, and human-data interaction to explain how receipt-based data can connect transaction records with product information, household context, interaction mechanisms, and user control. The paper first systematizes application fields of digital receipts and then reduces them to a basic model of receipt-based grocery practices. This model relates actors, data objects, consumer-journey phases, and interaction types. Building on the model, we propose a layered reflective digital receipt-based IHIS architecture consisting of receipt capture, product and food information, household context, reflection and interaction, and governance and user control. The paper contributes to HCI and consumer-facing information systems research by clarifying how digital receipts can support reflection while identifying limits related to data incompleteness, contextual uncertainty, moral pressure, privacy, and user control.
Paper Nr: 43
Title:

An Adaptive Companion for Low-Emission Zone Compliance: An HCI-Driven MaaS Integration

Authors:

Amal Hammami, Hend Marouane, Hassene Mnif and Mohamed Mosbah
Abstract: Urban drivers navigating Low Emission Zones (LEZs) face a critical yet under-addressed Human-Computer Interaction challenge: making complex, time-pressured compliance decisions with no real-time decision support, often resulting in anxiety, confusion, and environmentally counterproductive behavior. This paper presents the design, implementation, and preliminary evaluation of a human-centered adaptive mobile companion that transforms LEZ compliance from a punitive restriction into a guided, low-effort, and motivating experience. Grounded in established HCI principles, our interaction design embeds three core innovations into a unified Android MaaS application. First, we introduce a zero-burden Crit'Air classification service powered by a dynamic multi-learning engine (Random Forest, R² = 0.98) that predicts vehicle emission categories. Second, we design a proactive notification architecture leveraging Vehicle-to-Infrastructure (V2I) communication and Cooperative Intelligent Transport Systems (C-ITS) to deliver context-aware Decentralized Environmental Notification Message (DENM) alerts with a maximum of three sustainable mobility alternatives. This limited-choice design aims to reduce interaction complexity and potentially limit driver distraction and cognitive overload. Third, we propose an SDT-grounded gamification framework, comprising eco-points and loyalty badges, that incentivizes sustainable mobility choices through a sense of competence. The system was validated through network-level simulation in OMNeT++ in Bordeaux, France, and evaluated through a stationary task-based study. The results indicate successful task completion, perceived transparency of access decisions, and positive behavioral intention toward the proposed mobility options. This work contributes a reproducible, open-data-driven interaction design framework that demonstrates that LEZ compliance and user satisfaction can be achieved through human-centered adaptive design.
Paper Nr: 48
Title:

Analyzing Upper-Face Action Units as an Indicator of Perceived User Satisfaction to Support User Classification: A Preliminary Study
Authors:

Antonio Suarez-Crespo, Javier De Andres, Daniel Fernandez-Lanvin and Martin Gonzalez-Rodriguez
Abstract: This preliminary study explores the relationship between Upper-Face Action Units and user’s perceived satisfaction during interactions with an e-commerce website. Previous research suggests that upper face movements may function as cooperative signals of insufficient understanding or a need for clarification during face-to-face communication. To investigate whether these facial movements also play a significant role in human–computer interaction, we conducted an experiment combining Upper-Face AU detection with post-task usability surveys. Participants carried out some online shopping tasks while their facial expressions were recorded using a webcam. The videos were analyzed using Py-Feat to extract the AU data for each participant. Then, we applied basic linear regression to explore possible connections between specific AUs and the responses from the usability survey. The results showed some behavioral patterns through associations between Upper-Face AUs and user experience. Eyebrow-related AUs: AU01 (Inner Brow Raiser), AU02 (Outer Brow Raiser) and AU04 (Brow Lowered) showed marginal connections with questions related to the website ease of learning, content trustability, and the intention to revisit the site. AU05 (Upper Lid Raiser) and AU06 (Cheek Raiser) were the most prominently associated AUs in the study.
Paper Nr: 59
Title:

A Human-AI Collaborative System for Teaching System Design Inquiry
Authors:

Ying Zhu
Abstract: In the age of generative AI, software engineers start to focus more on system design than low-level coding. However, current computer science curricula do not provide enough structured instruction in system design. Existing system design resources focus primarily on solutions, teaching system architectures (e.g., microservices), design patterns, and technical mechanisms (e.g., caching). Less attention is given to the in- quiry process itself: how to ask questions to clarify requirements, uncover quality attributes, and reason about constraints and tradeoffs. To address this gap, we present SystemDesignInquiry, an interactive human-AI collaborative environment that teaches students how to ap- proach system design through structured questioning. For this purpose, we propose a System Design Question Matrix (SDQM), a two-dimensional pedagogical framework that maps six architectural layers (User, Inter- face, Function, Data, Structure, and Operation) against four reasoning modes (Purpose, Quality Attributes, Design Choices, and Failure and Scale). To improve user experience and reduce perceived delay, the environment separates conversational roleplay from inquiry assessment. A Respondent Large Language Model (LLM) answers student questions through an en- gineering persona, while a concurrent Classifier LLM maps each question to the SDQM and updates inquiry-coverage feedback in the background. A separate post-session LLM generates the final debrief. The system is designed to scaffolds structured questioning about requirements, trade- offs, and failure modes rather than memorization of standard architec- tures.
Paper Nr: 63
Title:

Toward an Educational Virtual Model for Heat Pump Training: A Virtual Reality Platform and Pedagogical Scenarios
Authors:

Rima Aridi, Paul Richard, Emmanuelle Richard and Thierry Lemenand
Abstract: Heat pumps (HPs) are becoming key technologies in the transition toward more sustainable energy systems. However, their ther- modynamic principles and operating mechanisms remain challenging for learners to understand. We argue that the use of virtual heat pump models combined with Virtual Reality (VR) techniques is highly relevant for supporting the learning of thermodynamic systems, as it enables im- mersive and embodied understanding of complex processes. However, we also contend that simpler, non-immersive interaction modalities can similarly lead to effective learning outcomes. In this paper, we propose virtual models of heat pump systems together with several scenarios designed to facilitate technical education. These scenarios include the identification of system components and their functions, the diagnosis of anomalies and malfunctions.
Paper Nr: 75
Title:

Visualization-Based Storytelling of the Changing Arctic Ocean
Authors:

Esben Bay Sørensen, Karl Attard and Stefan Jänicke
Abstract: The Arctic Ocean is changing rapidly due to human-induced climate change, which is altering the entire Arctic ecosystem. Many of these changes are difficult to communicate to public audiences because they involve processes that unfold below the water surface and across large spatial and temporal scales. We present a public-facing, visualization-based storytelling system that uses scrollytelling to guide non-expert users through selected Arctic Ocean processes. The system was developed through a participatory design process with Arctic marine scientists and combines narrative explanations with geospatial map views, temperature and sea-ice data, SVG-based illustrations, and interactive elements. As users move through the narrative, visual elements and visualizations are progressively revealed, highlighted, or transformed to explain relationships between sea ice, light availability, turbidity, coastal-zone habitats, and seafloor photosynthesis. We evaluated the system through a preliminary mixed-method study with 16 participants, combining questionnaire responses, written feedback, and interaction tracking. The results suggest that participants found the system engaging and helpful for understanding changes in the Arctic Ocean, while also identifying areas for improvement in interaction flow, attention guidance, and explanatory context.
Paper Nr: 104
Title:

Towards More-Than-Human Design: Navigating the Expanding Scope and Complexity of Design through Design Constraints, Design Frames, and Generative AI
Authors:

Tom Gross
Abstract: The design of interactive systems has been facing an ongoing evolution: the systems to be designed are changing, and so the methods and tools used to produce them during interaction research and interaction design needed to be adapted. Over the decades, the complexity of the design situation has drastically increased. At the same time, new technologies such as Generative Artificial Intelligence (GenAI) have the potential to profoundly alter the way we design, develop, and evaluate interactive systems. This paper characterises the challenges of this evolution from User-Centred Design to Human-Centred Design to More-Than-Human Design. It particularly analyses the increasing complexity this entails. It discusses approaches to address those challenges through design constraints and design frames. Finally, it derives requirements for future GenAI-enhanced methods and tools for interaction research and interaction design to tackle those challenges.