Conference Agenda

GlowAR is a hybrid AR installation that combines a physical, flower-shaped 3D structure with an iPad application and spatial projection. The predefined structure serves as a scaffold that provides novice users with orientation without fully determining their creative choices. Using the iPad, users can select and recolor individual components and augment them with animations, virtual elements, or playful interactions. GlowAR is currently an interactive design and configuration environment based on predefined content, but it is intended to evolve into a low-threshold authoring tool. The projection makes changes immediately visible on the physical object, allowing nearby onlookers to follow the creation process.

This work-in-progress presents aLife (audio for life), an open and pedagogically oriented XR-based platform that aims to mitigate the conceptual difficulties involved in understanding sound as a spatial, dynamic, and dramaturgical component of audiovisual narratives. The project is developed by researchers and educators with long-standing experience in teaching sound technology, sonic creativity, and spatial audio in multimedia and interactive contexts. Its motivation emerges from sustained pedagogical practice and from the authors’ intention to contribute to the development and understanding of spatial sound within the academic and scientific community.

Students' prior knowledge and goals are becoming increasingly heterogeneous, yet curricula and courses are rarely designed to accommodate this. GraphIT is an open, prototypical web platform that models curricula and courses as a dependency graph: to each topic its suggested prerequisites are linked. Built on Wikibase and queried through SPARQL, the platform stays extensible in schema, content, and interface. We have used it to document and organize multiple courses in a media informatics degree program. Custom visualizations surface central topics, paths between topics, and clusters of related material. In this demo we focus on how educators and students interact with the graph and derive insights.

Virtual reality (VR) cycling exergames can make indoor exercise more engaging by combining physical activity with immersive environments. While mismatches between real and virtual motion are often treated as undesirable, carefully controlled discrepancies can also create new interaction possibilities. We present an interactive demo making this design space tangible for real and virtual inclinations in VR cycling. Attendees ride a tilting indoor bike and experience combinations of physical bike tilt and visually simulated uphill and downhill slopes. This first experience illustrates how visual slope cues and bodily feedback jointly shape perceived inclination, and when mismatches may still feel congruent. In the second experience, attendees try a VR cycling exergame prototype in which slope changes are integrated into gameplay. Players avoid obstacles, deactivate barriers, and collect points while comparing motion-based cycling input with a button-based mode. The demo enables attendees to feel how controlled inclination cues can enrich VR cycling exergames.

Digital mobile health interventions need to decide not only what support to provide, but also when to intervene. Poorly timed prompts can be missed, ignored, or perceived as disruptive even when the intervention content is useful.

We demonstrate PULSE, a Personalized User-Level Scheduling Engine implemented as a Model Context Protocol (MCP) server, through ScrollWise-JITAI, a just-in-time adaptive chatbot-delivered micro-coaching intervention for doomscrolling and problematic social media use.

PULSE estimates receptivity within a criticality window and returns candidate timestamps, an estimated engagement probability, the selected evidence level, and a rationale. The demonstration highlights progressive personalization for the cold-start problem: recommendations can rely on historical population data, data from the current intervention, a tailored cohort of similar users, or a personalized model once enough individual interaction data are available. Conference attendees can manipulate synthetic user profiles, time windows, data sparsity, and unavailability constraints, inspect the resulting distributions, and observe how a ScrollWise micro-dialog is scheduled and logged. The demo shows how receptivity-aware timing can become a transparent and reusable infrastructure component for just-in-time adaptive interventions or ecological momentary assessments.

We present two Immersive Virtual Reality Learning Environments (IVRLEs) featuring virtual twins of real-world engineering laboratories: the Virtual Construction Design Lab (CD Lab) and the Virtual High-Voltage Lab (HV Lab). Both developed and in-class evaluated in close collaboration with the teaching staff at the Berlin University of Applied Sciences and Technology (BHT), these IVRLEs leverage VR's immersive and multimodal nature to enhance STEM education. They facilitate the visualization of invisible phenomena, simulate high-risk scenarios, and enable location-independent learning. We demonstrate our IVRLEs and share experiences at the Mensch und Computer (MuC) 2026 Demos track to support future research and inform the design of prospective virtual learning environments.

This paper presents SketchBeyond, a mobile web application that enables users to transform simple hand-drawn sketches into 3D-printed objects and Augmented Reality (AR) experiences. To realize this process we leveraged artificial intelligence (AI) 3D generation models. With the system we aim to provide a personal, practical and enjoyable experience to 3D modeling, AR, and 3D printing.

Traditional music production tools often require technical knowledge, which can discourage novice users from exploring sound design freely. We present SpaceSound, an immersive virtual reality installation that enables users to create ambient soundscapes by interacting with planets and satellites in a virtual environment. The environment in SpaceSound represents outer space and consists of six planets with satellites orbiting around them. Each planet serves as a loop-based sound source, representing instruments such as drums, pads, bass, keys, plucks, and an arpeggiator, while satellites act as sound modifiers, providing effects such as filtering, chorus, reverb, tremolo, and distortion. Users can activate, combine, and modify sound sources through direct spatial interaction and receive immediate auditory feedback and optional contextual explanations. To support creative exploration without overwhelming users, the system restricts the combinations of these sounds to ensure that they remain harmonically coherent. SpaceSound demonstrates how immersive interaction and metaphor-based design can make sound synthesis concepts more accessible while supporting calm, collaborative, and expressive soundscape creation.

As an interdisciplinary field, Human-Computer Interaction (HCI) is shaped by diverse theoretical traditions that influence what researchers notice, value, and critique in interaction scenarios. Yet these theoretical lenses often remain implicit, difficult to articulate, and knowledge about different theoretical traditions is scattered. To make HCI researchers and students aware of their theoretical lenses and initiate reflection, we designed the Theory of Human–Computer Interaction Glasses, an augmented reality demonstration that allows visitors to explore an interaction scene through six different theoretical lenses (or „glasses“). Through the theory glasses, visitors can experience how each HCI theory highlights distinct phenomena and offers different explanations. By actively switching glasses, visitors are invited to compare theoretical lenses and reflect on their own theoretical positioning. Ultimately, we hope that this demonstration improves understanding across research traditions and reduces friction to transform intertheoretical interactions.

We demo iihks!, an interactive XR installation in which visitors co-generate a personalised, painting-styled portrait from a selfie, a chosen painter, and a poem, and then watch their live surroundings repainted in the same style on a framed portrait screen, on their phone, and inside a VR headset. The user is able to step into her own painting. The properties that make such a system pleasant and easy to use (a smooth interface, an emotionally legible result, and no perceptible wait) are exactly those aesthetic theory attributes to kitsch. Built on the same generative pipeline as our earlier BildKlangPoet [2], which on its own closes the aesthetic relation the way any consumer selfie-to-portrait app does, iihks! reopens it with two practical moves: a single visible control (the carousel of the visitor’s own images) that drives every display and never disappears, and a persistent community gallery, backed by a statistics page, that sorts every confirmed image by visual similarity and, during the print or e-mail export, zooms out to the whole accumulated mass before returning to the visitor’s own picture. That export flow stages the point directly: the result that arrived as an individual production is shown, and measured, as one point in a dense cluster of near-identical defaults, testing whether reflexivity can be engineered into the same frictionless surface that produces the kitsch.

We present an early demonstrator of an AI-assisted workflow companion for multimodal eye-tracking analysis. The demonstrator explores how an assistant can help researchers formulate analysis goals, inspect available data, propose processing steps, and document these steps as transparent nodes in a processing tree. The work is situated in the context of the BiSigma Analyzer, a modular desktop application for eye-tracking and physiological data analysis. The demonstrator does not aim to provide fully autonomous analysis or a production-ready agent runtime. Instead, it focuses on the interaction concept: natural-language analysis requests are transformed into inspectable workflow proposals, shown as draft nodes in the processing tree, and executed only after user review. This makes the system a design probe for discussing transparency, control, and reproducibility in AI-assisted scientific software.

While custom haptic proxies significantly enhance immersion in Virtual Reality (VR), the native tracking of arbitrary physical objects remains a persistent challenge. Researchers typically rely on expensive, external, or bulky tracking add-ons because commercial Head-Mounted Displays (HMDs) cannot natively recognize custom shapes. We address this limitation by introducing an open-source artifact that replicates the casing and internal mounting layout of the Meta Quest 3 Touch controller. Our modular model allows designers to seamlessly merge native tracking architecture with any 3D-printable object, embedding the original internal tracking electronics directly into custom proxies without requiring external hardware. This demonstration showcases the applicability of the open-source model by letting visitors experience the workflow for designing custom controllers at an interactive station, view live 3D printing, and explore the physical benefits of natively tracked, haptically diverse proxies within a VR application.

The conversion and transformation of data and models is a standard task in information technology. In most cases, this is accomplished using software developed specifically for this purpose. A better approach in terms of maintainability and extensibility is to use languages designed specifically for this application to describe such transformations. This demo illustrates the practical implementation of the Object Management Group’s Query/View/Transformation standard. Using the QVT relational language defined therein, data and model transformation rules can be described graphically and declaratively and then executed. This no-code solution enables even those without programming experience to define and use data and model transformations.

To foster the inclusion of children with acute or chronic illnesses, mobile telepresence robots as avatars in schools represent a promising approach. However, through their sensors and interactions, these robots may threaten various dimensions of privacy for different stakeholders. These concerns range from the physical privacy of classmates at school to the informational privacy of the ill child participating from home. Such privacy risks and the associated concerns can reduce acceptance of the technology or even lead to its rejection, despite its potential benefits for the affected child. To address these challenges, we have designed and implemented in our BMFTR PRIVATAR project user-friendly solutions that enable stakeholders to protect their privacy according to their individual preferences through different interaction mechanisms. Our demonstrator provides meaningful privacy control that extends beyond standard consent-based data protection. This empowers all stakeholders to engage more confidently with a technology that is both highly promising and still relatively unexplored in everyday settings.

We propose a three-station demo of We, Brute Force, an 8-minute VR 180° stereoscopic dance miniature on the impossibility of explaining the human to a non-human reader. The demo extends the artwork [10] by two interactional elements: a feedback card set through which visitors map their own perceptual response back into the equine-plausible movement notation used in the work, and an AI-powered video avatar of the author(s) that hosts an interactive post-experience talk about the artistic concept of this work. Together, these stations stage a transformation of interaction itself: from a one-directional address of the human toward non-human intelligences into a bilateral, embodied exchange in which visitors’ own perceptual responses re-enter the work as material — made perceptible at the scale of breath, weight, and the displaced gaze.

Event accessibility information is often incomplete, generic, or difficult to interpret, limiting disabled people’s ability to assess participation opportunities. We combined qualitative interviews (𝑁 = 12), a comparative analysis of accessibility checklists (𝑁 = 13), iterative prototype design, and an in-the-field summative evaluation (𝑁 = 2). The resulting role-based platform integrates multiple accessibility checklists, enables organizers to document detailed information about venues and transport, and presents it to visitors in a structured, visual, and context-sensitive format that supports independent decision-making.

LLMs increasingly influence how people access information, form opinions, and interact with digital services. Their role in public discourse and the media ecosystem raises questions traditionally addressed by media regulators, including nondiscrimination, free political opinion formation, and compliance with societal and democratic norms. However, the models’ internal mechanisms remain inaccessible, requiring black-box testing to evaluate normative alignment. As a preliminary step, a structured normative catalogue was developed in collaboration with a SMA. Building on prior work that identified regulator-relevant evaluation criteria for Large Language Models, high-level values drawn from Article 5 of the German Constitution, the Medienstaatsvertrag (State Media Treaty), the Digital Services Act, the AI Act, and international AI ethics guidelines were translated into technically testable criteria. This demonstrator shows the first two implemented criteria: detection of discrimination and free political opinion formation.

We present a full-stack explainable AI system, aiding researchers in reading academic papers. Users need to extract relevant information

to understand complex content. Thus, they can query the system in natural language about uploaded documents or use customized

prompts.We focused on the incorporation of three design goals: Truthfulness, transparency and privacy. Truthfulness is addressed

by implementing a retrieval-augmented generation (RAG) pipeline, grounding the system answers in provided documents and avoiding

hallucinations. Transparency is provided by incorporating several XAI components, which are developed based on users’ informational

needs. Lastly, privacy is preserved by hosting the system and language models on local university structure. Lessons learned in system

development include the extraction of informational needs in fast, iterative user testings, system revalidation in dynamic environments

and balancing trade-offs in the front-end. Future development could tackle different stages of the research process, such as academic

writing and add more features, such as the connection to local data bases.

Diagnostic reasoning is an iterative cognitive process of synthesizing clinical data into a diagnosis. However, teaching this skill systematically is challenging, often requiring highly individualized learning strategies. To address this, we developed AskRudi, an LLM-based trainer for medical students. The platform integrates a structured teaching framework grounded in human-centered design and evidence-based pedagogy.

Given the task of labeling and cleaning point cloud–based 3D data, existing solutions are typically either fully manual or fully automated. Manual labeling in 3D requires substantial time investment, while automated approaches often lack robustness and still demand post-processing and supervision to ensure quality results. We argue that both domains may benefit from an integrated approach and therefore propose an immersive VR tool for instance segmentation and labeling of 3D point clouds with model-assisted automation. The tool combines two state-of-the-art machine learning methods for instance segmentation with standard functionalities for manual cleaning and labeling.This hybrid design enables partial automation to reduce labeling effort while still allowing supervised corrections, ultimately creating a semi-automated system.

Human-robot interaction has evolved from side-by-side operation toward more collaborative human-robot teams, making natural communication increasingly important. This demo presents a system that augments the humanoid robot Reachy with natural, bi-directional voice interaction powered by a Large Language Model (GPT-Realtime-2). Unlike a previous experiment in which the robot communicated solely through a rigid script, the LLM now autonomously manages physical movements, state transitions, and interaction phases based on conversational context. The demonstration scenario is a collaborative tower-building task in which the robot and a participant jointly place three wooden blocks, with participants free to speak to the robot naturally rather than following a fixed script. The cognitive processing runs on an external machine to ensure fluent arm movement despite the computational demands of LLM integration. The system illustrates how natural voice interaction may positively influence variables such as trust and anthropomorphism in human-robot interaction, offering visitors a glimpse into future modalities of interacting with robots.

Participatory Design (PD) projects involve numerous interconnected aspects, including stakeholders, contextual factors, methods, partnerships, and outcomes, making them challenging to plan and reflect upon. Existing frameworks and guidelines typically present these aspects as linear textual representations, thereby limiting exploration of their relationships. To address this challenge, we present the Reflection Framework for Participatory Design (RFPD), an interactive web-based framework that supports the exploration and reflection of complex participatory design processes. Building on an adaptation of the Community-Based Participatory Research (CBPR) framework, the RFPD combines a synchronized graph visualization with structured content to enable dynamic, non-linear navigation between interconnected PD aspects. The framework further supports project reflection through integrated reflection questions and privacy-preserving local export of documented results. The presented system is fully functional and can already be applied throughout participatory design projects to support planning, documentation, learning, and reflection. In this paper, we present the conceptual design, interaction concept, and technical implementation of the RFPD prototype, developed for PD in informal care.

Videospiele bieten ein großes Potenzial zur Förderung von motorischen und kognitiven Fähigkeiten, jedoch stehen vor allem ältere Personen mit wenig bis keiner Spielerfahrung vor Hürden beim Spielen. Häufige Herausforderungen sind komplexe Spielmechaniken sowie das vorausgesetzte implizite Wissen von etablierten Spielkonventionen, über das ältere Erwachsenen ohne vorherige Spielerfahrung meist nicht verfügen. In diesem Paper stellen wir das Videospiel "MaFa", ein zugängliches, storybasiertes 2D-Spiel vor, das speziell auf die Fähigkeiten und Bedürfnisse von älteren Personen mit wenig bis keiner Videospielerfahrung entwickelt wurde. Das Spieldesign setzt auf eine reduzierte Spielkomplexität, einer intuitiven Führung der Spielenden sowie entschleunigten Interaktionsmechaniken, wodurch Aufgaben innerhalb der Spielwelt ohne Zeitdruck gelöst werden können. Eine Vorstudie mit einem ersten Prototypen gibt Hinweise darauf, dass die gewählten Gestaltungsprinzipien die Barrierefreiheit optimieren und insbesondere für Erstnutzer*innen einen positiven Einstieg ermöglichen können.

Conducting human-robot interaction (HRI) research with physical robots presents some practical barriers: robots are costly, require dedicated lab space, and demand safety considerations, often placing HRI research out of reach. Augmented reality (AR)-based robot simulations promise a way around these barriers, offering more inexpensive hardware, eliminating physical risk, and enabling more flexible, distributed study designs. Yet AR's usefulness hinges on whether interactions with virtual robots elicit human responses comparable to those with physical robots. Recent work suggests this may be the case for certain aspects. Though evidence remains limited and further research is needed in that regard. This requires software tools enabling realistic AR-based HRI experiments. We present an early version of such an system that leverages the same control stack as physical robots to produce realistic behaviour. Already used in two studies, it forms the foundation for our ongoing investigation into AR's validity as a substitute for physical HRI.

The Virtual Lab combines the core functions of a collaborative whiteboard with a creative element. Using a set of isometric illustrations of everyday objects, teachers and students can work together to set up, structure and continuously develop a visual workspace. As a browser-based tool that requires no specialized equipment, it remains easily accessible, whilst the spatial design of virtual working environments supports group organization and collaborative working in a hybrid teaching context.

Understanding depth perception in Virtual Reality is challenging because multiple visual depth cues contribute simultaneously to the perception of three-dimensional space. While these cues are well described in literature, their individual influence is difficult to experience directly. We present an interactive Virtual Reality application that enables users to explore classical depth cues by selectively activating, deactivating and combining them in real time. The application provides a guided learning experience in which individual depth cues are explained and immediately visualized inside immersive VR. A modular framework allows each cue to be controlled independently while maintaining a consistent shared virtual environment. The demonstration runs on a Meta Quest 3 headset and invites visitors to interactively investigate how individual cues influence spatial perception and performance during specific tasks. The system is intended as an educational tool for teaching depth perception as well as a flexible framework for future perceptual experiments.

Adherence to the intended use of a data collection protocol or digital health intervention is essential for both intervention effectiveness and data quality. We present ADHERE-MCP, a reusable model context protocol (MCP) server that predicts and prevents imminent non-adherence in mobile data collection and intervention studies. The service receives the latest adherence status and the next scheduled interaction session from a mobile app, reconstructs a participant-specific interaction history, estimates non-adherence risk, provides a non-adherence rationale, selects a prevention strategy based on established behavior change techniques, and logs the decision. Our demo uses ScrollWise, a five-day chatbot micro-coaching intervention against doomscrolling and problematic social media use, to let attendees experience what happens when a user misses scheduled coaching sessions. We further outline a scenario-based feasibility and usefulness check with simulated user interactions. The demo reframes non-use as an actionable interaction state and illustrates how prediction algorithms and prevention policies can be swapped for different studies.

AnimalVR is a Virtual Reality demo designed to bridge the gap between abstract product labels and the actual living conditions of livestock through immersive embodiment. Visitors inhabit the body of a laying hen, experiencing daily life across two distinct scenarios: conventional barn housing and organic free-range systems. By allowing visitors to directly explore and compare these environments from a first-person perspective, AnimalVR translates abstract labeling schemes into concrete, lived experiences. Unlike existing campaigns that rely on emotional shock or abstract knowledge transfer, AnimalVR utilizes a neutral, fact-based design that encourages reflection on animal welfare and sustainable consumption.

This paper presents NoviceAI, a web-based teachable agent that inverts the AI-as-tutor paradigm by positioning a Large Language Model (LLM) as a programming novice that gets taught by students. Based on the Protege Effect, which states that learners achieve deeper understanding when teaching others, NoviceAI prompts students to explain introductory programming (CS1) concepts through dialogue. The system features an adaptive comprehension model, adjusting its simulated knowledge level based on student explanation quality. Strong explanations, evaluated through an LLM-based reflection process, cause the agent to "learn" and increase its knowledge level, leading to more sophisticated questions. Weak student explanations reduce the agent's knowledge level and trigger more basic questioning.

Supported work contexts such as sheltered workshops often require work instructions that can be adapted to individual abilities, tasks, and needs for supporting people with disabilities. A digital worker assistance system (WAS) can provide this worker guidance through step-by-step instructions and situated feedback during manual work. The system's practical use, however, also depends on whether instructions can be created, adapted, and maintained without extensive technical expertise.

This demo presents SIDEKICK, a modular assistive workstation system for configurable digital work instructions. SIDEKICK combines a worker-facing step-by-step instruction interface with physical input and feedback components, such as buttons, ultrasonic sensors, and LED-based visual cues. Its current redesign separates instruction authoring, runtime guidance, and hardware communication. This enables the system to be more easily extendable with hardware components, making instructions more adaptable to work processes.

The demonstration focuses on three core processes, explorable by participants: creating and adapting work instructions, pairing hardware devices to instruction events, and setting up and running task instructions at the workstation. Through the demo, we aim to discuss implementation approaches for systems to create assistive work instructions, while remaining configurable, with a focus on its usability in the sheltered workshop context. This addresses challenges impacting workshops' inclusivity and efficiency, respectively.