MyPW is an innovative, data-driven smart campus framework developed at Warsaw University of
Technology to modernize university operations and enhance the academic experience. The system
integrates data from various sources, including IoT sensors, cameras, infrastructure monitors, and user
inputs collected via mobile apps, kiosks, and web platforms. This comprehensive data collection enables
MyPW to support campus operations, research, and crisis management while fostering collaboration
within the academic community. At its core, MyPW relies on technologies such as machine learning and
Large Language Models (LLMs) to process and analyze data in real-time. These technologies facilitate
event detection, resource optimization, and personalized user interactions. The system’s scalable
microservices architecture ensures adaptability and long-term sustainability, while secure data repositories
protect sensitive information and support ongoing research. The MyPW system is designed to serve four
main stakeholder groups: students and staff, administrators, campus security, and research teams. Sensory
data from IoT networks—such as video surveillance and heat meters—are processed using deep learning
algorithms to detect movement patterns, incidents, and communication flows. Simultaneously, user-
generated queries are handled through a multimodal platform, where LLMs help verify events by
integrating user reports with sensory data. All collected data is anonymized and stored in cloud-based
repositories, allowing for AI-driven analysis and the development of new features tailored to user needs.
MyPW’s approach aligns with the Smart City framework, emphasizing security, inclusivity, and
sustainability. By combining human-centred design with AI-driven insights, MyPW aims to create a
functional, user-friendly, and sustainable campus environment that supports collaboration, efficient
management, and achieving sustainable development goals.

In 2024, in response to the transformative impact of generative AI on higher education, Merito
Universities Development Centre launched a pioneering cohort online course designed to equip
faculty teachers with the necessary skills and insights to start using gen. AI technologies in their
teaching. Created after extensive research, this course aims to transition faculty from AI
sceptics to informed advocates. The aim is for them to be capable not only of integrating AI tools
effectively in their teaching but also of understanding the need for crucial changes in the
students’ learning processes, engagement activities, and assessment methods.
The cohort course unfolds over five modules and ten weeks, including more than sixteen plus
live online meetings, addressing key aspects of AI in teaching. It emphasizes practical
applications and the development of AI-informed teaching materials, AI subject policies, AI
ethics (and what they mean for both students and teachers). All these elements (alongside a
final project) ensure that academic teachers discover hands-on that working with AI is far from
intuitive.
The course's success is based on applicable guidelines for instant implementation. These
include a shortlist of practical AI tools, a proactive approach to integrating AI in preparing
materials, and maintaining transparent communication with students regarding AI — an
obligatory ingredient for a successful teacher-student relationship.
The course's structure and timeline create an environment of shared learning and continuous
professional development—which continues after each course ends. Merito’s 100+ “AI
Ambassadors in Didactics” (now recognized with certificates from our universities) are already
creating ripples of change in their local environments due to the materials and insights gained
during this course. My talk is about sharing outcomes and lessons learned from 3 editions of this
course.

Psychometric measurement faces numerous challenges, including outliers, differential item functioning, and complex response patterns that require sophisticated analytical techniques capable of capturing intricate data interdependencies. In this study, we present a novel application of the ReMasker masked autoencoding framework, extending the range of analytical techniques available for psychometric research by leveraging neural networks' pattern recognition capabilities. Our approach employs a transformer-based autoencoder architecture to learn the underlying structure of questionnaire responses, providing a complementary method to traditional psychometric analyses. We demonstrate this framework's utility using personality questionnaire datasets and validate our findings through simulation studies. The ReMasker model's reconstruction capabilities enable multi-level analysis of response patterns from item, variable, and whole-answer-set perspectives, yielding insights into data structure and response consistency. Particularly valuable is the examination of reconstruction loss between model outputs and original responses, which offers a data-driven method for understanding patterns that conventional statistical techniques may not capture. Our findings suggest that autoencoder-based approaches represent a valuable addition to the psychometric toolkit, providing researchers with enhanced capabilities for understanding complex response patterns and underlying psychological constructs. These proposed solutions represent an initial step toward integrating deep learning methodologies into psychometric practice as a complementary analytical approach.

“Seeing Minds in Machines”: Prevalence of Phenomenal-Consciousness Attribution to Large Language Models in Poland

Rapid adoption of Large Language Models (LLMs) by the lay public has led to a large number of people attributing some degree of internal experience to those systems. Building on mind-perception literature, which highlights experience as the core dimension driving attributions of mind, recent US data show that 67 % of the 300 adult participants credited ChatGPT with at least minimal phenomenal consciousness (Colombatto & Fleming, 2024). Such beliefs may inflate trust, blur moral boundaries, and erode the ontological basis of human identity (Akbulut et al., 2024). We will report the first Polish estimates for phenomenal-consciousness attribution on an online sample of 600 adults. Participants complete a single slider item gauging the extent to which they believe ChatGPT “has subjective experiences such as feelings and sensations,” mirroring the American protocol. The presentation interprets the results in light of the documented societal risks identified by leading AI research labs and the three-factor psychological theory of anthropomorphization. The results will offer preliminary national estimates for policymakers and the research community at large.

Background: Traditional qualitative research in social sciences faces significant constraints including high resource demands, potential interviewer bias, and geographical limitations that restrict the scope and scale of inquiry.

Method: This paper introduces MimiTalk, an AI-powered automated interview platform leveraging Large Language Models and a dual-agent constitutional architecture designed for ethical and effective qualitative data collection. We conducted comprehensive evaluations involving 20 diverse participants to assess platform performance using quantitative metrics including information entropy, natural language processing analysis, and semantic coherence measures, alongside qualitative feedback assessment.

Results: Results demonstrated that AI-led interviews achieved information richness comparable to human interviews, with high semantic coherence scores (mean similarity: 0.817) and positive user experiences characterized by reduced interview anxiety and enhanced disclosure comfort. A crucial comparative study directly contrasted AI-led interviews with traditional human-led interviews conducted among academic researchers across various disciplines, revealing that AI-led interviews successfully elicited conceptual frameworks and technical insights remarkably consistent with human interviewer results.

Conclusion: This research positions MimiTalk as a robust complementary tool that enhances the efficiency, scalability, and accessibility of qualitative research while maintaining authenticity and depth, paving the way for advanced AI-assisted research frameworks.

Keywords: Artificial Intelligence, Large Language Models, Automated Interviews, Qualitative Research, Research Methodology, Human-AI Interaction

Fictional works have long produced a wide range of depictions of autonomous humanoid robots.
However, until recently, real-world robots were capable of only very limited autonomy. The
recent explosive development of large language models—followed by advances in multimodal
and action-based AI systems—has dramatically changed this landscape. These breakthroughs
now enable humanoid robots to exhibit an unprecedented degree of flexibility in both the actions
they perform and the decisions they make. As a result, robots are increasingly shifting from
being mere tools under the direct control of human operators to becoming genuine partners in
social interaction. Whether a robot is perceived as a tool or as a social partner has a profound
impact on the psychological and neural mechanisms that guide human behavior in human–robot
interactions. In this talk, we will present our recent work exploring this topic using the iCub
humanoid robot. We will discuss key experimental findings, highlight how humans interpret and
respond to robot autonomy, and outline several promising directions for future research. These
directions may prove critical for the practical deployment of autonomous robots in everyday
social, educational, and therapeutic settings.

Aleksandra Landowska1, Aislinn D Gomez Bergin1, Ayodeji O. Abioye2, Jayati Deshmukh3,
Andriana Boudouraki1, Maria Waheed1, Athina Georgara3, Dominic Price1, Tuyen Nguyen3,
Shuang Ao3, Lokesh Singh3, Yi Dong4, Rafael Mestre3, Joel E. Fischer1, Sarvapali D.
Ramchurn3

1 School of Computer Science, University of Nottingham, Nottingham, UK 2 School of
Computing and Communications, The Open University, Milton Keynes, UK 3 School of
Electronics and Computer Science, University of Southampton, Southampton, UK 4 University
of Liverpool, Liverpool, UK

This presentation introduces the REACT (REsponsible and Adaptive robots in Complex
seTtings) project, a multidisciplinary initiative funded by Responsible AI UK. The project aims
to develop responsible and adaptive multimodal multi-human multi-robot (MHMR) systems
for complex social environments, such as care homes. Recognising the challenges of an ageing
population and pressures on healthcare services, REACT seeks solutions that go beyond
automation. The goal is not to replace humans but to enhance human-AI teaming, supporting
collaboration and shared decision-making.
REACT focuses on building context-aware, emotionally responsive, and ethically grounded
robotic systems. It employs a range of sensors, including neuro, physiological, behavioural,
and environmental data, to enable dynamic communication between humans and robots. By
integrating neurophysiological and behavioural data with advanced AI, robots can adapt to
individual needs, preferences and situations. The project is designed for multiple stakeholders,
including care workers and residents.
A co-design methodology involving older adults, caregivers, healthcare professionals, and
ethicists ensures that the technology remains relevant, responsible, usable, and ethically
grounded. Early outcomes include adaptive robotic prototypes, multimodal datasets, realistic
care scenarios, and ethical frameworks. These aim to enhance care quality and build trust in
technology while promoting dignity and transparency.
This presentation outlines the project's goals, work packages, and next steps. It invites
feedback, audience engagement, and collaboration, fostering inspiration

This study presents the development and validation of a web-based application designed to
support the participatory design (co-design) of assistive robots intended for elderly care. The
main objective was to create a tool that facilitates the involvement of diverse users in the
design process of socially assistive technologies.
The application enables users to define their preferences regarding robot features, generate
visual representations of robots using the DALL·E model, and evaluate the generated
images based on personal perceptions. The system includes a login mechanism with
session management, allowing both unauthenticated users and registered administrators to
participate. An export function supports data collection for further statistical analysis.
Structurally, the application consists of three main modules: (i) a user preference form, (ii) an
AI-based image generation interface, and (iii) an evaluation module where users rate the
generated images. The backend was deployed on a remote server, enabling broader
accessibility and facilitating research participation.
To evaluate the tool’s effectiveness, a user study involving 36 participants was conducted.
Participants interacted with the system, resulting in the collection of a large dataset
subjected to Two-Way ANOVA analysis. Out of 702 statistical tests, 21 revealed significant
differences, mainly related to participants’ education level and marital status. These findings
underscore the relevance of demographic factors in technology acceptance and perception.
The results validate the application as a promising platform for inclusive robot design. It can
be further expanded and implemented in elderly care settings to enhance the development
of technologies tailored to varied user needs, supporting human-centred approaches in
assistive robotics.

What happens when a robot doesn’t politely mirror our opinions but openly says “I disagree”? As robots and conversational AI move into roles once reserved for human advisors, teachers and everyday companions, even small shifts in dialogue can strongly influence how people feel and behave.

In our study, 241 participants interacted with the humanoid robot Pepper, which either agreed, disagreed or remained neutral during a short conversation. The effects were clear: disagreement consistently lowered trust, reduced perceived expertise and made people less willing to share personal data. Neutrality, interestingly, worked just as well as agreement, suggesting that robots don’t need to flatter us—they simply shouldn’t contradict us without good reason. Only charitable donation decisions remained unaffected.

These findings matter because social robots are becoming soft interfaces to information, services and data. Their conversational style can subtly shape human decisions, even in casual contexts. Designing robots that communicate ethically, predictably and with social awareness is essential for technologies that support rather than unsettle their users.

Paradox of Data Commons: Governance Lessons from a Health Data
Intermediary Multi-Case Study

The future of digitalization depends on implementing the right model for data
governance, balancing rights, values, and societal goals. This research investigates a
matched triple of health Data Intermediaries (DIs) – public authority Findata, data
cooperative Midata, and non-governmental foundation Donate Your Data – to
understand how different DI models impact data governance goals and the rules-in-use.
I find that the choice of the DI model influences data governance and thus the pursued
goals. But the findings also reveal a paradox, where DIs couple and balance
extractivism with privacy, collectivism with self-determination, and trust with economic
gain. Embracing this paradox instead of resolving it is what leads to successful
governance. This study contributes to the theoretical development of data governance
studies by offering insights that can support policymakers in designing effective
institutions.

Keywords: data governance, commons, data intermediary, health data, privacy, case
study

Wearing Care: From Human Data to Dignity in AI-Supported Telecare Systems
for Older Adults

Institute of Applied Computer Science, Lodz University of Technology

Keywords: human-computer interaction, wearable technology, remote health
monitoring , predictive analytics, elder care

In the face of rapidly ageing populations and healthcare workforce shortages,
wearable-based telecare systems enhanced by Artificial Intelligence (AI) and
Generative AI (GenAI) open new frontiers for proactive, continuous, and personalized
elder care. This talk presents insights from the design, implementation, and
evaluation of intelligent wearable-driven telecare systems that go beyond passive
monitoring towards active, adaptive, and relational support and how such systems
can support older adults in maintaining autonomy, health, and dignity at home, and
on their own terms.
We present findings from a study, based on in-depth interviews conducted with
patients, remote caregivers, and support staff, of a large-scale (over 2000 older
users) remote health monitoring (RHM) program implemented in an urban region of
Poland. These findings reveal that technology-supported care is most effective when
it is relational, transparent, and responsive—not merely reactive.
We also reflect on two complementary research and development projects. AsSIST
(completed): a personalized telecare system integrating a smart wearable wristband
with non-invasive sensors (e.g., vitals, glucose, hemoglobin) and a 24/7 platform
supporting independent living and healthy routines. CardioSense (ongoing): an AI-
enhanced system for early prediction of cardiovascular events (heart attacks, strokes,
heart failure) using continuous biosignal analysis and decision-support for healthcare
professionals.
Both systems demonstrate how wearables and AI can work together to detect risk,
enable timely intervention, and support communication within care networks. We also
explore the emerging role of GenAI in facilitating intuitive, contextual
interaction—offering older adults a sense of control rather than surveillance.
The talk offers an interdisciplinary perspective on designing remote care systems that
are technologically advanced, ethically grounded, and human-centered—blending
HCI, healthcare innovation, and social insights to shape caring futures through
wearable intelligence.

This project builds on the assumption that a deeper understanding of neurobiological
mechanisms can significantly improve diagnostic precision and therapeutic outcomes for
anxiety and depression. By integrating genetic data, neuroimaging findings, and behavioral
indicators with deep learning models, I explore innovative interdisciplinary approaches to
mental health diagnostics.
Based on recent genome-wide association studies (GWAS), scientists identified
anxiety-associated variants in genes such as TMEM132D, PDE4B, and NTRK2, which
regulate stress responses and synaptic plasticity. Rodent studies confirm functional links
between these genes and anxiety-like behaviors, suggesting clear biological pathways that
could be a part of diagnostic tools.
Complementing this, they presented a machine learning-based diagnostic support system
using a battery of six cognitive-behavioral tasks. A Random Forest classifier achieved up to
80.5% accuracy in distinguishing anxiety from depression, based solely on implicit
behavioral data. This method offers an objective, non-invasive tool that avoids the biases of
self-report questionnaires, enhancing early diagnosis and informing individualized
interventions.
The proposed framework combines Explainable AI (XAI), cognitive performance profiles, and
biological markers to expand and potentially create a system for more personalized and
precise mental health care. The project highlights how neuroscientific findings can be
translated into actionable diagnostic tools, with potential applications in both clinical and
R&D contexts.

Keywords: Translational Neuroscience, Explainable AI (XAI), Anxiety, Depression, Deep
Learning, Neuroimaging

As artificial intelligence becomes increasingly embedded in everyday life, the issue of trust between humans and artificial agents is gaining critical importance. Trust is not only a facilitator of effective and safe cooperation between humans and AI systems, but also a buffer against potential misuse and overreliance. This presentation delves into the multidimensional nature of trust in AI, examining how various factors—such as system transparency, behavioral consistency, perceived competence, interface appearance, and social context—shape users’ attitudes and behaviors toward artificial agents.

In addition to theoretical exploration, the presentation draws on empirical findings from two studies conducted by the author. The first study involved a month-long interaction between participants and a chatbot, assessing changes in trust levels before and after the engagement, accompanied by a qualitative analysis of chat transcripts. The second study utilized a questionnaire distributed among students from various Polish universities, investigating how the field of study and frequency of AI tool usage influence trust levels and the tendency to anthropomorphize AI systems.

The goal of this presentation is to offer actionable insights for designing human-AI interactions that foster an optimal level of trust—not too low to hinder engagement, nor too high to encourage blind reliance. By identifying psychological mechanisms and design principles that shape trust, we can better align AI systems with users’ expectations and needs, ultimately supporting more balanced and meaningful collaboration.

Keywords: chatbot, human-computer interaction, trust in artificial intelligence

Session  IV: LLM & Gen AI

This study investigates whether large language models (LLMs) developed in different coun-
tries better capture the cultural values of their home markets, and whether prompting these

models in local languages improves cultural alignment. As LLMs become central tools for
generating synthetic consumer data in global marketing research, concerns grow over their
ability to authentically reflect cultural norms.

We systematically compare four cutting-edge LLMs—GPT-4, GPT-4o (OpenAI, US), Mis-
tral (France), and Yi (China)—across five cultural contexts (China, France, US, UK, and a

global English-speaking average). Each model was prompted 100 times in English, French,
and Chinese using 18 validated items from the GLOBE cultural framework, covering four
key dimensions: Uncertainty Avoidance, Gender Egalitarianism, In-Group Collectivism, and
Power Distance.

Results show that while LLMs accurately reproduce broad patterns in cross-cultural rank-
ings, they systematically deviate from benchmark values. Surprisingly, locally-developed

models did not consistently perform better in representing their home cultures. Yi signifi-
cantly underestimated Chinese Power Distance and Collectivism, while Mistral did not out-
perform US models for French values. Prompt language influenced results: English prompts

yielded more stable outputs but often reduced cultural variation; Chinese prompts intro-
duced larger deviations, especially in Gender Egalitarianism and Uncertainty Avoidance.

All models showed a consistent downward bias in Power Distance and upward bias in Gender
Egalitarianism, raising concerns about potential distortions in AI-generated market insights.

These findings highlight risks in relying solely on LLMs for cross-cultural consumer simula-
tion and call for careful model selection, prompt calibration, and human oversight.

We offer a replicable benchmarking approach using GLOBE dimensions to assess cultural
alignment in synthetic data. Our study contributes to the growing literature on AI in
marketing by clarifying the limits of current LLMs in culturally sensitive applications and
suggesting strategies for responsible, culturally aware deployment.

Rapid adoption of Large Language Models (LLMs) by the lay public has led to a large number of people attributing some degree of internal experience to those systems. Building on mind-perception literature, which highlights experience as the core dimension driving attributions of mind, recent US data show that 67 % of the 300 adult participants credited ChatGPT with at least minimal phenomenal consciousness (Colombatto & Fleming, 2024). Such beliefs may inflate trust, blur moral boundaries, and erode the ontological basis of human identity (Akbulut et al., 2024). We will report the first Polish estimates for phenomenal-consciousness attribution on an online sample of 600 adults. Participants complete a single slider item gauging the extent to which they believe ChatGPT “has subjective experiences such as feelings and sensations,” mirroring the American protocol. The presentation interprets the results in light of the documented societal risks identified by leading AI research labs and the three-factor psychological theory of anthropomorphization. The results will offer preliminary national estimates for policymakers and the research community at large.

Intelligent Realities: From Digital Twins to Industrial Virtual Worlds

As industries evolve beyond automation, the fusion of Artificial Intelligence (AI) and Virtual Worlds is redefining how we design, operate, and learn. Intelligent Realities combine digital twins, immersive environments, and real-time data to create dynamic ecosystems where humans, machines, and algorithms continuously co-adapt and co-evolve.       

In these Industrial Virtual Worlds, AI serves as both an enabler and a collaborator — augmenting human expertise through predictive insights, situational awareness, and adaptive simulation. Rich simulation libraries, gamified learning modules, and behavioral analytics transform training and operations into interactive, measurable experiences.

Sectors such as Oil & Gas, Defense, Metallurgy, Mining, and Aerospace are increasingly leveraging these intelligent environments to simulate complex workflows, conduct remote training, and test mission-critical strategies without physical risk. By integrating cognitive AI agents and immersive XR technologies, organizations can visualize entire industrial ecosystems — from design to deployment — in a fully interactive, data-driven space.

We will explore how Virtual Worlds, generative AI, and multi-agent simulations converge to build the next generation of industrial systems. Ultimately, Intelligent Realities mark a clear shift from automation to cognition — redefining how industries design, learn, and grow in the age of intelligent, immersive transformation.

Sketching Soft Skills: Early Testing of a VR Escape Room Using Paper-Based Scenarios

Virtual reality (VR) is increasingly being used in education and training each year, particularly to develop soft skills through immersive experiences. Despite this, designing practical VR applications remains a challenge due to the high cost, complexity and limited flexibility of high-fidelity (hi-fi) prototyping. Low-fidelity (low-fi) approaches, such as paper scenarios or clickable mockups, offer a faster and more accessible method for early usability testing. Nevertheless, despite their potential, the diagnostic value of low-fidelity prototypes in the context of immersive VR remains under-researched.

This study aims to assess whether and to what extent low-fidelity prototypes can be used to identify usability problems and elicit measurable psychophysiological responses in users. We conducted an exploratory usability study with 62 participants who interacted with VR escape room scenarios created on paper. The tasks were designed to develop soft skills. During the sessions, participants wore Embrace Plus smartwatches, which recorded physiological and behavioural data, including electrodermal activity (EDA), heart rate (HR), pulse rate variability (PRV), respiratory rate, skin temperature and motion indicators.

The collected data was synchronised with scenario events to analyse user responses in real time. Qualitative observations and user feedback were also collected to supplement the physiological observations. The results of our experiment indicate that even without complete immersion, the low-fi prototypes were able to induce stress, confusion or engagement, as evidenced by changes in EDA and HR. Additionally, several usability issues were identified, including unclear instructions and cognitive overload.

These results support the use of low-fidelity prototyping as a valuable method for early-stage testing of VR applications. The combination of subjective reports and objective biometric data provides a richer understanding of the user experience, enabling rapid iteration before investing in costly VR application development. This study contributes to expanding the methodological toolkit for human-centred design in VR environments.

AI on the Bench: Training Judicial Decision-Making Through Immersive Courtroom Simulations

This talk, led by Dariusz Szawurski-Radetz (Head of the VR/AI Lab at Poland’s National School of Judiciary and Public Prosecution), presents an innovative immersive VR solutions for judicial training. The systems (like ISSVR developed and introduced by the Polish company Cinematic VR) use decision-tree-based 180° or 360° courtroom scenarios to engage judges and prosecutors in realistic case simulations. Participants with VR headsets confront complex situations – for example, evaluating racial or minority legal dilemmas or AI-generated forensic evidence – by choosing actions at each step and witnessing either correct or incorrect outcomes, each accompanied by concise legal explanations.

These interactive VR “films” immerse users in the nuances of bias, privacy, and due process. Scenarios include both correct and incorrect endings, helping trainees see the consequences of each decision in a low-stakes setting. Judges who have tried the system praise its realism and safety: one observed it “offers a more realistic courtroom environment than traditional mock trials”. Newer judges find it “excellent” for building courtroom skills (without the stress of peer observation), and experienced judges use it to rehearse unusual or tech-driven cases. Many see VR as a powerful complement to conventional training methods.

This human-centered approach helps judges “live through” abstract legal dilemmas and internalize lessons. By translating complex AI-related challenges into concrete VR based experiences, the National School promotes innovation in legal education and afirms judicial preparedness. It exemplifies how emerging technology can prepare decision-makers to uphold rights, fairness, and human-centered values in the age of AI, AR, and VR.

Beyond Code and Canvas: Practice-Based Insights for Artist-Engineer Teams

Franciszek Sobiech, Magdalena Wróbel-Lachowska, Laurent Babout
Lodz University of Technology, Institute of Applied Computer Science

Contemporary computational interactive art projects increasingly rely on seamless
integration of human–computer interaction (HCI) principles and artistic practice, yet
fundamental tensions may persist between engineering rigor and creative freedom. As both
participant and observer, drawing on an autoethnographic, practice-based research
methodology, this study examines three co-creative processes conducted between 2023 and
2025, each bringing together an artist and HCI researcher to develop participatory
installations. These projects incorporated multimedia elements such as paintings, video,
sound, and sculpture, which offered a broad range of possibilities for interaction. Specific
installations include an interface-controlled video mapping over paintings, a room installation
incorporating sensor-controlled sound and video and an interactive kinetic sculpture with an
AI sound generation feature. Through the practice-based process, I identify recurring
challenges in communication and decision-making. Key findings reveal that making use of
frequent prototyping sessions in line with an iterative mindset significantly enhances
collaborative fluency. Moreover, a significant challenge when starting such a project is based
on a lack of shared language and awareness of technical possibilities. Another observation
is the need to consider an artist’s perspective when advancing explainable AI projects. This
work-in-progress project offers preliminary insights that collectively advance HCI theory by
foregrounding the artist’s unique needs and offer a step towards a collaboration model for
interdisciplinary teams.

Keywords: practice-based research; interactive art; human–computer interaction; creative collaboration

AI Is Here – Now What? A Human-Centered Guide to Organizational Adoption

Artificial Intelligence is no longer a futuristic concept—it is already transforming how we work, communicate, and make decisions. But while technological advancements accelerate, many organizations and individuals are struggling to keep up. Fear, resistance, and uncertainty often slow down the adoption of AI or lead to misuse. This raises a fundamental question: how can we adopt AI in ways that are responsible, effective, and inclusive?

This presentation proposes a human-centered approach to AI adoption that prioritizes people over technology. We will explore how to introduce AI into organizations by building digital confidence, earning trust, and reshaping workflows and mindsets. Rather than focusing solely on tools or technical integration, the session emphasizes the importance of communication, behavioral change, and psychological safety in AI-driven transformations.

Participants will gain practical strategies to engage teams, reduce friction, and promote ethical use of AI technologies. Whether you are leading a digital transformation or just beginning your AI journey, the session offers actionable insights to make adoption smoother and more impactful.

Drawing on interdisciplinary knowledge from behavioral psychology, change management, and real-world organizational case studies, the session will show how to create environments where AI is not only accepted but embraced. Attendees will leave equipped with a clearer understanding of how to align technological change with human values, ultimately enabling more sustainable and people-centered innovation.

Keywords: Artificial Intelligence, Human-Centered Design, Organizational Change, Digital Transformation, Ethics, Behavioral Psychology, AI Adoption, Trust Building, Change Management

The Junior Gap: AI, design, and the missing first job

In UX/UI, we’re seeing a growing “junior gap” – fewer entry-level job offers, rising expectations, and AI tools handling more and more of the work once done by humans. What does this mean for people learning design? What skills help them compete – not just with experienced professionals, but with intelligent software? And how can education help bridge this gap? I’ll give my perspective as a lead product designer in a large company and a lecturer: I’ll explore what’s really changing in the design job market, what skills remain uniquely human, and how junior designers can navigate a world where AI tools are already part of the team. We’ll look for practical answers — beyond hype and fear.