We meet Prof. Paweł W. Woźniak in a winding old building near the Gußhaus at TU Wien, where the Human-Computer Interaction research group is currently housed on a temporary basis. The actual premises are being renovated at the moment, so we interview Prof. Woźniak amidst packed boxes, tools, technical prototypes, and 3D printers. At the Faculty of Informatics, Human-Computer Interaction (HCI) is a distinct research unit, at the intersection of computer science, engineering, and design, which Prof. Paweł W. Woźniak has been leading at TU Wien since April 2024. His team develops and researches user-friendly computer systems that promote well-being, health, and quality of life. We discuss technology acceptance, user research methods, the handling of qualitative data and user-centred prototype design – all against the backdrop of the current AI hype and the question of how technology can truly be made accessible and inclusive.

Taking a critical look at technologies

“So, the philosophy behind this whole field of research is that we are preparing people for future technologies. We ask ourselves what the future will look like, so that we are ready when a commercial product hits the market. Yes, acceptance, usability and also the negative consequences, that’s the focus. We take a critical look at today’s technologies and how they are designed for predefined user groups. And then we ask how they could be redesigned to be usable by everyone.” 

Prof. Woźniak’s team is conducting extensive research into health technologies such as fitness trackers, and into the impact that interaction design has on the use of health apps and users’ well-being. A recently published project explores how meaningful narratives can emerge from mere numbers and movement data, fostering ‘genuine’ reflection on the tracker data. Instead of rigid targets such as “12,000 steps”, his team uses language models to transform tracker data into vivid descriptions accompanied by visualisations. According to Woźniak, 60 user interviews clearly show that raw numbers provide little motivation; users want to draw their own conclusions rather than simply following instructions from an app. Instead, what is needed is a targeted presentation of one’s own movement and health data that provides clarity amidst a flood of information. 

Another project which turns its focus beyond usability towards accessibility is the ERC-funded ACCESSTECH led by Katta Spiel (Associate Professor). As HCI research has predominantly been shaped by non-disabled researchers, many technologies fail to reflect the lived experiences of disabled people – and are perceived by them as undesirable or even harmful. ACCESSTECH therefore critically examines whether traditional ‘accessibility’ is really what deaf people want – or whether true potential can only be realised through active collaboration with the deaf community.

Development and testing of prototypes 

“Interaction design is a tool of Human-Computer Interaction. This means that, in order to understand technologies, we must get directly involved in the design process – in principle, we first design the technologies, then implement them, then test them, and then redesign them accordingly. That is why we have students here from Computer Science as well as from the Angewandte who have studied design. The aim is to develop prototypes, which then serve as a proposal for a future technology that is superior in one or more aspects to what currently exists. We then conduct studies with people to evaluate these technologies.”

When it comes to researching usability and accessibility, there is a wide range of methods in HCI, depending on the maturity of a prototype and its intended area of application.To put simply, this means: ‘lab’ for doability, ‘field’ for reality, and ‘online’ for broad reach.

Paweł raves about the institute’s actual lab spaces, which feature, among other things, an observation room with a one-way mirror, eye-tracking cameras, and motion sensors. These laboratory experiments with early-stage prototypes are time-consuming, as are field studies in everyday life, for example, with wearables; results are only considered robust once the sample size exceeds 20 participants. Online questionnaire studies, on the other hand, represent a scalable method for large numbers of participants, who are recruited externally via crowdwork platforms according to demographic characteristics. Software such as Qualtrics serves as a central survey tool, as it enables JavaScript integration and web app players, allowing questionnaires to adapt to user interactions in real time and remain reusable. Though they are cost-effective, easy to implement, and delivering extensive quantitative data, online surveys are often associated with lower quality. According to Prof. Woźniak, with 1,000 respondents, around 20 per cent of the data will be discarded, as random response patterns and systematic clicking in the middle of scales distort the results.

Between data protection and reproducibility 

“What do you do about the reproducibility of qualitative data? Obviously, you can’t just make that public, particularly in our field, where we work with health and activity data, as well as personal secrets – people place a great deal of trust in us. It’s a bit unnerving for us too, sometimes, that we now have this in our data repository – the question is always: at what level of analysis can we publicly share what we’re doing? But we also have a third, important type of data that we produce: the prototypes themselves – […] code that can be shared in the TU Digital Library or on GitHub.”

Data sharing in the field of Human-Computer Interaction must be approached differently depending on the type of data, in order to strike the right balance between data protection and reproducibility. In qualitative studies such as the 60 fitness tracker interviews mentioned above, the application of the FAIR principles is limited: reproducibility is never fully achievable here, as the data is highly context-dependent and this context usually contains sensitive personal information that cannot be shared for data protection reasons; instead, coded content, thematic clusters, levels of analysis, and review steps are made available.

Quantitative data, such as that from online questionnaires, is already anonymised via external recruitment platforms (participants identified only by ID) and, following analysis, can be shared as anonymised CSV raw data, freely accessible in the TU Digital Library alongside the associated publication. According to Prof. Woźniak, five years ago, this practice still needed to be encouraged; today, it is standard practice among PhD students and significantly increases paper acceptance rates.

All video material from laboratory studies which often results in extensive amounts of data, is carefully stored locally, analysed anonymously and then deleted, whilst details on prototype development are shared generously: code on GitHub, instructions in YouTube videos, 3D models, electronics specifications and shopping lists are linked in the paper – so that researchers worldwide can replicate and further develop the work.

Summer of AI – winter of HCI

“The design principles remain constant over the long term; technology changes all the time, but people don’t change that quickly, which is why it’s important that we preserve the artefacts – the prototypes, the code, and the 3D models. The history of how we developed these technologies can inform future technologies greatly. That is why we talk about AI summers and HCI winters in the field. If you look back at the investments in AI over the last 60 years, it becomes clear that when hopes for AI and automation were too high and were not fulfilled, investment subsequently returned to interfaces and human-centred design.”

In conclusion, Paweł W. Woźniak paints a picture of the HCI field as one characterised by cyclical waves of hype, such as those seen in the imaginaries surrounding automation or the current AI hype, where inflated expectations lead to failure and resources subsequently flow back into user-centred interfaces. Naturally, his team works with local and web-based LLMs to explore their potential for Human-Computer Interactions. AI models are not currently suitable for research purposes such as the analysis of qualitative data – the context of interviews and field studies is too nuanced, the landscape of personal data too sensitive, and the models too opaque.

Prof. Paweł Woźniak explains that the institute trains people with entrepreneurial skills and conducts application-oriented research – yet at the same time, great importance is placed on highlighting alternatives that do not necessarily follow a business model. Some of these systems may be of little interest to the market but are all the more significant for society: they promote well-being rather than dependency, and their results are shared openly. Preserving artefacts helps us understand the technological development of future generations, as FAIR data practices help bridge the gap between technological hype and real-world application. According to Woźniak, Human-Computer Interaction is a long-term discipline that documents how the relationship between humans and computers has changed historically, and designs technology in such a way that it enhances human processes rather than replacing them.

Contact

Paweł W. Woźniak
Research Unit of Human Computer Interaction
TU Wien
pawel.wozniak@tuwien.ac.at

Center for Research Data Management
TU Wien
research.data@tuwien.ac.at