ABSTRACT. This paper examines how the emerging concept of non-technical debt (NTD), specifically Process, Social, and People Debt, can be utilised to understand and address recurring issues in game development. Drawing from Politowski et al.’s large-scale analysis of 200 game postmortems, we map the top ten industry challenges to NTD as described by Ahmed and Gustavsson. Politowski’s analysis showed that many issues, such as unclear vision, misaligned teams, and stress, stem from human and organisational decisions rather than technical limitations. While technical debt is well known, the growing recognition of NTD remains underexplored in game development. We argue that applying an NTD lens during development, rather than after the fact in a post-mortem, can help teams avoid costly issues, particularly in creative and high-pressure environments such as game production.

ABSTRACT. This paper analyzes the phenomenon of customer churn in ERP system implementations, with the aim of identifying effective strategies for reducing its occurrence in the ERP industry. The study is based on real-world data from 67 implementation projects conducted by an ERP service provider. It focuses on the stages of the implementation cycle where customer cooperation is most frequently discontinued and classifies the primary reasons for project abandonment. Based on the findings, a set of targeted strategies is proposed to mitigate churn risk. These recommendations are intended to improve implementation success rates and foster long-term customer engagement.

ABSTRACT. This paper presents the GenAI–PM Integration Model, a conceptual framework for embedding Generative Artificial Intelligence (GenAI) within Agile project management across the Software Development Life Cycle (SDLC). The framework addresses critical challenges posed by increased complexity, data management, and stakeholder collaboration in contemporary projects. It aligns specific GenAI capabilities to Agile project phases, demonstrating how AI can enhance decision-making, streamline processes, improve resource allocation, and foster higher-quality outcomes. By emphasizing co-creative interactions between AI systems and human teams, this model highlights the transformative potential of integrating GenAI into Agile methodologies, contributing to more adaptive and responsive information systems development practices.

ABSTRACT. The article presents the design and deployment of a production-grade MLOps infrastructure that integrates edge computing with cloud-based resources for efficient machine learning operations. The system connects vehicle-mounted sensors (cameras and LiDAR) with a centralized data lake and a Kubernetes-based cloud environment. A fully automated pipeline, developed using Apache NiFi, manages continuous data acquisition, preprocessing, metadata registration, and model lifecycle orchestration. Data collected on edge devices are stored on a shared NAS and seamlessly transferred to the cloud for large-scale training and evaluation. The system allows for safe access, flexibility, and processing based on metadata, making it easy to scale and repeat machine learning workflows that meet current MLOps standards. The system has been validated in real-world deployments, confirming its applicability to safety-critical scenarios.

ABSTRACT. This study examines gender equity among IT freelancers across selected European countries, focusing on differences in digital labor market participation and earnings between male and female freelancers. Using a non-invasive data scraping approach we obtained a rich dataset about 16 524 freelancers from Poland, Romania, Serbia, Spain, and Ukraine, we analyzed both labor market participation and earnings gap differences across three sectors: IT services, writing related services, and consulting services. The results show that there are gender differences in earnings among IT freelancers, but this is not a phenomenon observed in every sector. Possible future research directions are discussed, including expanded sectoral and geographical analyses and exploration of skill-specific influences on gender-based earning differences in digital labor markets.

ABSTRACT. In the problem of teaching project management, two approaches are used - unsupervised, where the student independently explores the knowledge, and supervised, where it is done under the oversight of a teacher. However, unsupervised approaches struggle with lower material assimilation and explanation of more complex concepts, while the teacher's availability bottlenecks supervised approaches' scalability. This paper proposes a hybrid solution integrating Generative AI (GenAI) to combine the scalability of unsupervised learning with the effectiveness of supervised methods. Firstly, a game-based simulation of a Software Development Project Management scenario was conducted with 67 students to understand the problem better. It highlighted students' high engagement and confirmed the need for feedback. Secondly, an initial validation using manually prepared queries showed promising potential in GenAI providing feedback and eliminating the need for teachers' engagement. However, it underlined the need to curate the answer scope and maintain the context of the simulation progress for such solutions to be effective. To answer those problems, two approaches are proposed: one leveraging a specialised Small Language Model and another employing a Large Language Model with Retrieval-Augmented Generation.

ABSTRACT. The paper presents research on the increase in mathematical knowledge of students using a specially implemented, deeply immersive and highly engaging virtual escape room based on mathematical puzzles. It was built under the supervision of the authors for use in a CAVE-type installation in the Immersive 3D Visualization Lab at the Gdańsk University of Technology. The virtual escape room comprises an introductory room followed by three themed rooms with 13 puzzles total that involve mathematical thinking. To assess the educational impact of the escape room, an experiment was conducted with students starting their first year of technical studies. The experiment consisted of solving mathematical puzzles in the escape room by cooperative teams of 2-5 students, preceded and followed by satisfaction surveys and knowledge tests. Each group also participated in classroom lesson, some before visiting the escape room and others after. The greatest increase in knowledge was achieved for classroom lessons, followed by the escape room.

ABSTRACT. The development of artificial intelligence (AI) is causing a lot of emotions and discussions around the world, both in the context of the possibilities it brings and the threats that may result from its irresponsible implementation. Trust in AI and the threats associated with it are important topics in the debate about the future of technology. This study examined how the scientific community in social sciences uses AI in scientific and teaching work. The survey obtained responses from 151 respondents who indicated their experiences with the use of AI by answering quantitative and qualitative questions. The study also determined the threats perceived by the academic community and the degree of trust in the answers obtained within the framework of AI technology.

ABSTRACT. Large language models (LLMs) are transforming programming education by enabling automated generation and evaluation of coding exercises. While previous studies have evaluated LLMs’ capabilities in one of these tasks, none have explored their effectiveness in solving programming exercises generated by other LLMs. This paper fills that gap by examining how state-of-the-art LLMs—ChatGPT, DeepSeek, Qwen, and Gemini—perform when solving exercises generated by different LLMs. Our study introduces a novel evaluation methodology featuring a structured prompt engineering strategy for generating and executing programming exercises in three widely used programming languages: Python, Java, and JavaScript. The results have both practical and theoretical value. Practically, they help identify which models are more effective at generating and solving exercises produced by LLMs. Theoretically, the study contributes to understanding the role of LLMs as collaborators in creating educational programming content.

ABSTRACT. This paper presents a Design Science Research (DSR) approach to addressing visualization bottlenecks in environmental Information Systems Development (ISD). By developing a CUDA-based atmospheric effects framework utilizing the Material Point Method (MPM) and Marching Cubes algorithms, we demonstrate how GPU acceleration transforms ISD methodologies for data-intensive decision support systems (DSS). This research contributes to digital transformation of environmental monitoring platforms by enabling real-time processing of complex simulation data that traditionally require significant computational resources. The prototype demonstrates scalable performance handling up to 6.5 million particles while enabling configuration-driven customization that allows information systems developers to integrate sophisticated environmental visualization without specialized graphics expertise. This approach democratizes atmospheric data visualization for environmental monitoring systems. Empirical results demonstrate real-time visualization capabilities suitable for operational deployment.

ABSTRACT. This study explores the Sustainable Investment Gap (SIG) in 13 European countries, defined as the gap between awareness of ESG-labeled financial products and actual ownership by individual investors. Drawing on data from the 2023 OECD/INFE, the analysis integrates indicators of digital financial literacy (DFL), online financial activity, and the size of the gap. A preliminary exploratory typology of five country profiles was developed, and contextual digital interventions were proposed. The results indicate that higher levels of DFL co-occur with a smaller investment gap and greater online engagement, but digital readiness alone is not enough to close the gap. The proposed typology provides a framework for designing digital tools to support sustainable investment decisions. In further stages of the study, it is planned to test selected interventions in Poland using experimental methods in a simulated FinTech environment.

ABSTRACT. Modern information systems, such as enterprise resource planning (ERP) systems, are vital for real-time decision-making, providing instant access to essential data. The rise of file-based storage engines provides valuable alternatives in information system architecture, especially for resource-constrained organizations. This study conducts comprehensive performance benchmarks across diverse database architectures, columnar file-based, in-memory, and traditional RDBMS, using standardized ERP workloads that simulate transaction processing, analytical reporting, and mixed operations with varying data volumes and concurrency levels.

ABSTRACT. The primary objective of this work is the automatic analysis of road marking degradation, to support maintenance decision making. The evaluation is conducted in two main aspects: the visual condition, assessed using a camera, and the retroreflective quality, which ensures nighttime visibility and is evaluated using LiDAR data. We employ the latest YOLOv12 neural network for road marking detection, enabling real-time analysis during vehicle movement. Following detection, LiDAR data is recorded and used to analyze the quality of the reflected beams. The primary parameter considered in this analysis is the intensity of the reflected signal (ranging from 0 to 255). Based on this parameter, road markings are classified into two categories: good condition and those requiring maintenance. The proposed approach provides an automated and effective tool for road infrastructure monitoring within intelligent transportation systems.

ABSTRACT. The purpose of the article is to identify, elaborate and systematize the role of technical standards in the formation of Smart Cities in Poland. The method used was a critical analysis of legal documents and a case study of Warsaw and Krakow, supported by induction and deduction. The results of the study show that the voluntary application of ISO 37120/37122 standards causes fragmentation of implementations and limits the advancement of Polish cities in the rankings of smart metropolises. The author argues that legal sanctioning of key Smart City standards and a mechanism for their regular updating could accelerate digital transformation and increase interoperability of urban infrastructure.

ABSTRACT. Mountain rescue operations face uncertainty, difficult terrain, and limited communication. This paper introduces a modular platform combining agent-based simulation in Unity with a microservice decision engine for real-time risk assessment. The system models weather, avalanches, tourist and animal movement, and processes sensor data using logic-based inference and machine learning. Early results confirm sub-second latency and high accuracy in hazard detection. Beyond technical capabilities, the platform proposes a cloud-based, data-driven business model designed for scalable deployment in rescue services, insurance, and tourism management.

ABSTRACT. This article presents a business model of a smart apiculture ecosystem. The domain of apiculture is considered to have a crucial impact on other branches of agriculture. The aim of the research is to provide a business model for a specific smart environment. This model is developed through a structured analysis of existing apiculture solutions, focusing on their technological components, value propositions, and target user segments. A comparative analysis is conducted using the Lean Canvas methodology to identify common strengths and shortcomings. Based on the findings, a new business model is proposed that integrates solar-powered hives, sensor-based monitoring, and real-time data analytics. The model is designed to support both professional and hobbyist beekeepers, as well as educational and research institutions, contributing to the broader adoption of smart agriculture technologies.

ABSTRACT. This paper presents a literature review on modularity and creativity in terms of design variants for generative adversarial networks for image creation. The objective is to lay the foundation for providing a suitable tool to support product design, as this area is considered a potential beneficiary of this concept. Based on the literature, a new model will be developed as an IT artifact in future research. Current approaches that allow the user to control certain features of GAN outputs are explored and commonly used metrics are investigated. Finally, limitations and future research directions are reflected upon.

ABSTRACT. Feature selection has great importance for simplifying machine learning and improving computational efficiency, especially when working with high-dimensional datasets. The rise of Large Language Models (LLMs) offers new opportunities in selecting predictive features. This paper aims to evaluate the potential of LLMs for feature selection tasks and examine whether a hybrid approach can lead to improved predictive performance. Using the DeepSeek-R1 model on publicly available datasets, the results show that LLM-driven feature selection holds significant promise. Furthermore, the performance of hybrid approaches highlights the value of LLMs as a complementary tool to traditional feature selection methods. Across the experiments, the hybrid approach either achieved the highest performance or ranked among the top-performing methods.

ABSTRACT. We present an approach for identifying persistently misclassified images in real-world thyroid ultrasound data. Using 484 images of thyroid nodules, we evaluated four different convolutional neural network architectures. Persistent misclassification is defined as images repeatedly misclassified across models and cross-validation folds. These cases are validated by an experienced radiologist and subjected to Grad-CAM analysis. Results confirm that images, that have negative impact on model results, often exhibit atypical or ambiguous features. We emphasize that persistent misclassification is an important source of diagnostic error, independent of model choice. Recognizing misleading cases is crucial for dataset quality, model robustness and the trustworthiness of AI systems in clinical applications. This work highlights the need for incorporation data validation strategies alongside standard performance metrics in the development of deep learning models.

ABSTRACT. This paper presents a novel approach to the classification of distributed data, which integratesthe cooperation of local decision tables within coalitions with rule induction and decision templates.The method aims to preserve model transparency while taking into account the diversityof data sources. Experiments were conducted on three datasets, comparing the performance offour rule induction algorithms: exhaustive search algorithm, genetic algorithm, covering algorithm,and LEM2. The best classification results were obtained for the exhaustive and geneticalgorithms, while the covering and LEM2 methods performed significantly worse. The proposedapproach achieves results comparable to the baseline method, which does not incorporate thecoalition mechanism, while offering higher interpretability. In addition, the proposed solutionwas compared with the Authors’ earlier approaches based on decision tree classifiers.

ABSTRACT. This study addresses the vulnerability of AI systems to adversarial attacks by extending the DeepFool algorithm. The paper proposes four new approaches and evaluates them according to a set of criteria. The methods are inspired by various optimisation algorithms. One of the proposed improvements adds the independent refinement stage, which reduces the final perturbation without extra gradient computations. Experimental results show that the appropriately modified algorithm reaches the decision boundary in fewer steps and with fewer gradient evaluations, while the refinement stage further decreases the magnitude of the perturbation. The combined approach can improve attack efficiency and reduce detectability, suggesting the potential for a wider application of advanced optimisation techniques in adversarial example generation.

ABSTRACT. The use of deep learning methods has gained momentum in the domain of survival analysis. Different models have been proposed to handle time-to-event data. Neural networks are used to find complex relationships between features, improving the predictive capabilities of deep learning models. When conducting experiments, one might want to reduce the number of methods that need to be examined because of the computational resources required for model training. Establishing families of deep learning methods that behave in a similar way might be beneficial for such scenarios. In this paper, we establish a way to measure differences between deep learning discrete-time survival analysis models. The proposed method is based on SHAP values. We conducted experiments for three datasets and five discrete-time survival analysis models. We proposed a special kind of plot that helps visualize the impact of features on the model outputs over time intervals. Based on the obtained results, we performed Friedman and Wilcoxon tests to examine statistically significant differences between the models.

ABSTRACT. While agile practices are widely adopted at the team level, extending agility to portfolio management remains a significant challenge, particularly in enabling continuous value realization within complex and volatile business environments. This paper investigates how agility and value delivery within an Information Systems (IS) portfolio can be enhanced by applying the principles of flow and the Theory of Constraints. Through an exploratory single-case study within an Indian fintech organization, we identify a cyclical framework for improvement encompassing flow visualization, constraint identification, constraint-aware resourcing, work-in-progress (WIP) limitations, accelerated learning, and dynamic reprioritization. Moreover, the study highlights foundational enablers like flow-optimized funding, metrics, and governance that collectively support agile portfolio capabilities. The findings indicate that applying flow principles and constraints management at the portfolio level helps to reveal bottlenecks, align resources strategically, enable rapid adaptation, and support system-wide optimization. The proposed conceptual framework advances the understanding of IS portfolio agility and offers actionable guidance for practitioners seeking to enhance the effectiveness and adaptability of portfolio management in dynamic enterprise contexts.

ABSTRACT. As a consequence of ChatGPT’s public release in 2022, software developers and learners of the profession were suddenly provided with a completely new and potentially extremely powerful tool to support them in designing and implementing their applications and answering occurring topic related questions. While, previously, community driven question and answer platforms like Stack Overflow were somewhat unique in their value proposition by providing (the chance for) answers directly geared towards the problems users encountered, they now have an alternative that additionally provides rapid response times. To determine how the emergence of ChatGPT impacted Stack Overflow, a literature review was conducted, exploring its effect on the users’ participation behavior and their perception. Further, the corresponding implications are discussed, especially highlighting the responsibility of higher education institutions in this context.

ABSTRACT. Building upon the analogy of Technical Debt, Process Debt refers to issues arising from inefficient or obsolete processes, which can substantially restrict an organization's effectiveness in delivering software. Process Debt creates additional tasks, such as rework, clarification, and workaround efforts, which can significantly increase the workload experienced by developers. A heightened workload may lead to stress and burnout. This study empirically examines the quantitative impact of five types of Process Debt on workload among Agile Software Development teams. Survey data from 191 participants in two large organizations revealed significant correlations between all Process Debt types and increased workload. Multiple regression analysis further identified Synchronization Debt, Roles Debt, and Infrastructure Debt as key predictors, highlighting their critical roles in intensifying workload pressures. These findings underscore the importance of proactively addressing specific areas of Process Debt, enabling organizations to enhance process efficiency, reduce developer overload, and maintain sustainable productivity.

ABSTRACT. Empiricism and continuous improvement remain the backbone of Scrum and other Agile methods. However, building a practical understanding of such a process during a one-day Scrum workshop or a university course remains an ongoing challenge. This paper introduces an extended version of a novel drawing-based game designed to demonstrate the use of client feedback in continuous improvement process and simulating work within a Scrum framework by means of an analogy involving the drawing of icons as an alternative to randomly generated results. In the performed study, 107 participants in subgroups took part in a one-hour-long game session and solved a Scrum knowledge test before and after each game. Results show noticeable improvement in Scrum knowledge among participants. Additionally, the drawing aspect of the game has been found engaging and significantly enhanced the construction of analogies between gameplay and real-world software development processes during the game and further work with the participants. This research adds to the existing knowledge on Scrum coaching and teaching, providing a simple-to-set-up game allowing for the simulation of the use of empiricism and continuous improvement without the need for complex, expensive tools or environments.

ABSTRACT. Context: Educational videos are common in online learning environments, but the effectiveness of instructor presence remains debated. Some theories suggest it enhances social engagement and motivation, while others argue it increases cognitive load. Objective: This study investigates the effects of different instructor presentation formats (intermittent, continuous, and absent) on learning outcomes, visual attention, and cognitive load in Mandarin Chinese vocabulary learning using eye tracking. We aimed to provide objective evidence on how these formats influence attention and cognitive load, providing practical implications for designing effective educational videos. Method: Using a matched-groups design, 120 participants were randomly assigned to one of three conditions based on their initial performance in an online learning session. During the second session, participants watched instructional videos while their eye movements were tracked, with learning outcomes measured through speaking tests. Results: No significant learning outcome differences were found across conditions. Eye-tracking showed learners in all conditions prioritized Pinyin (written phonetics) while largely ignoring the instructor. Conclusion: These findings challenge assumptions about instructor presence and demonstrate learners’ strategic visual attention regulation. Given equivalent outcomes across conditions, an instructor absent approach is preferable as it simplifies video design and reduces production complexity.

ABSTRACT. The growing number of online resources on information technology has left many learners feeling overwhelmed by the large number of career options and the paths to achieve them. This abundance of choices highlights the need for personalized career guidance and clear course recommendations to help learners focus on their specific goals. Existing recommendation systems fail to provide transparency and clear explanations for their suggestions. To bridge this gap, we present XCRS: Explainable Course Recommendation System, which recommends both career roles and associated courses in information technology with explainability at its core. XCRS utilizes large language model embeddings from Google, OpenAI, MistralAI, VoyageAI, and Cohere to deliver personalized recommendations tailored to users’ knowledge, past preferences, and future learning interests. Our contributions are two-fold: i) a pipeline to construct an explainable recommendation system for career pathways in information technology, ii) a replication package that includes the implementation, a public dataset of information technology courses, and the design for empirical evaluation. Our evaluation suggests that the overall system has been perceived as useful by the intended users, while there is no statistically significant difference in the performance of the large language models used.

ABSTRACT. With research proving the value of intensive tutoring for learners to improve their results but university educators often being limited in time and numbers, harnessing large language models (LLM) to bridge this gap and provide more support appears to be an auspicious solution. One promising avenue for this is students’ training in writing, since producing coherent texts is a strength of LLMs. To explore how university students and teachers perceive the quality of ChatGPT-generated reviews, a study was conducted in the frame of a university course on scientific writing for IT students, identifying the potential benefits but also the weaknesses of using ChatGPT as a reviewer for university students’ scientific texts.

ABSTRACT. Learning management systems (LMSs) have become a common practice in education delivery. The LMS market is growing rapidly, and the number of vendors offering diverse solutions is also growing. This raises the problem of selecting the optimal LMS. The choice of LMS is made using various criteria, both those related to the system itself and its technical parameters, as well as related qualitative criteria. The article addresses the problem of indicating the most optimal LMS, considering technical and cost parameters and user ratings. The most popular LMSs listed in the top LMS rankings were assessed.

ABSTRACT. This study investigates the extent to which Industry 5.0-related content, technologies, and practical components are integrated into higher education institutions in Poland. The findings reveal significant disparities based on institutional type and academic profile. Technical and public universities show the highest levels of readiness for digital transformation, while pedagogical, vocational, and artistic institutions (particularly private ones) demonstrate lower integration and limited access to relevant specializations. Using multiple correspondence analysis (MCA) and hierarchical clustering (HCPC), the research identifies latent structures and clusters of institutions with distinct digital adaptation profiles. Chi-square and Cramér’s V tests further confirm significant associations between institutional characteristics and Industry 5.0 readiness. The results highlight systemic inequalities and emphasize the need for differentiated policy interventions to support less advanced institutions in bridging the digital transformation gap.

ABSTRACT. As artificial intelligence (AI) and work automation become increasingly integrated into modern workplaces, understanding employees’ attitudes toward these changes is vital. This study examines how perceived competencies – both linguistic/technical (LC) and non-linguistic (NLC) – relate to Polish workers’ attitudes toward AI and automation. Using data from a representative sample of 1,067 employed adults, structural equation modeling (SEM) revealed that LC significantly predicted more positive attitudes, while NLC showed no significant effect. Multi-group analysis indicated gender differences in the strength of these associations. However, measurement invariance was not confirmed, suggesting different interpretations of competence items across gender. These findings underscore the importance of considering both competence profiles and sociocultural contexts when assessing workers' responses to technological change.

ABSTRACT. Standardized usability questionnaires are a fast and relatively effortless way of assessing usability of software products. Despite their long use, so far, little attention has been paid to the effect of sample size and the level of respondents’ acquaintance with the evaluated software on the measurement. This paper addresses this gap and uses SUS and mTAM measurements of ChatGPT to illustrate how the deviation from the mean usability score decreases with increasing sample size, and to confirm the significant effect of usage frequency and knowledge of the evaluated software and its alternatives on the measurement results. It also exposes no demographic bias due to participation of respondents of different gender, country of stay, and academic major.

ABSTRACT. Technology development necessitates the growth of competencies that enable employees to utilize it in their professional work effectively. Organizations such as the World Economic Forum (WEF) identify market trends and highlight the skills most sought after by employers, both generally and within specific sectors. This article presents the results of a study conducted at the turn of 2024/2025 on a group of Polish employees from various industries (N=288). The study aimed to assess employees' self-diagnosis regarding key competencies identified in WEF reports and to examine their correlation with using AI tools in their work.

ABSTRACT. Recommendation systems are essential for delivering personalized content across e-commerce and streaming services. However, traditional methods often fail in cold-start scenarios where new items lack prior interactions. Recent advances in large language models (LLMs) offer a promising alternative. In this paper, we adopt the retrieve-and-recommend framework and propose to fine-tune the LLM jointly on warm- and cold-start next-item recommendation tasks, thus, mitigating the need for separate models for both item types. We computationally compare zero-shot prompting, in-context learning, and fine-tuning using the same LLM backbone, and benchmark them against strong PLM-based baselines. Our findings provide practical insights into the trade-offs between accuracy and computational cost of these methods for next-item recommendation. To enhance reproducibility, we release the source code under https://github.com/HayaHalimeh/LLMs-For-Next-Item-Recommendation.git.

ABSTRACT. Data binarization involves converting a continuous data attribute into a finite set of binary attributes while minimizing information loss. It plays a crucial role in feature engineering in the data mining analysis. Data binarization simplifies data, improves model training quality, enhances model performance and interpretability of results, helping in understanding complex patterns. In this paper we present an original data binarization framework, called angle-based data binarization, that converts continuous attributes into discrete binary attributes. The proposed framework allows not only to simplify machine learning models, but can also lead to the improvement of the accuracy of a number of well-known traditional machine learning methods. We present results of an extensive series of experiments which evaluate the efficiency of the proposed method in the area of data classification. Using popular classification algorithms, we compared classification quality achieved on source datasets with classification quality achieved on their binarized versions. We also discuss binary attribute pruning, based on elimination of attributes with poor discriminative power.

ABSTRACT. Non-intrusive monitoring of electrical loads (NILM) implemented by the state analysis method critically depends on the selection of appropriate features to identify devices. The commonly used expert selection is not optimal, and computational methods of feature selection require the establishment of an optimisation criterion that will ensure a satisfactory level of NILM system performance. An important element of the discussed method is the selection of the classifier and its matching with the selection method to construct a softsensor. In this work, four feature selection methods (Boruta, ReliefF, mRMR, the author's method) and four classifiers (decision trees, random forests, artificial neural networks and a hybrid classifier) were implemented and tested. Software was implemented for the softsensor architectures tested, enabling the verification of optimal configurations for NILM. The research confirmed that the selection of features using optimisation methods and the use of a softsensor allow for better support in the decision-making process.

ABSTRACT. Chatbots powered by large language models (LLMs) are increasingly prevalent in various domains. Nonetheless, they face challenges such as hallucinations and losing context during extended conversations. This study tackles these issues by proposing a multi-agent strategy for chat architecture where multiple LLMs focus on distinct tasks to enhance the quality of their output. The suggested solution involves a supervisor agent working in conjunction with a document search and review module. We assess the performance of information systems with chatbots designed to respond to sustainability questions in English and handle technical documentation for plant equipment in Polish. A comprehensive analysis of commercial and open-source models revealed that Qwen2.5 v14b’s performance is comparable to that of the Gemini family models.

ABSTRACT. Agile Software Development (ASD) methodologies are often viewed as restraining IT innovation and causing technical debt. Recently, agile mindset leaders have been introduced as a remedy to solve this clash, describing them as those who secure dynamic capabilities. In other words, ASD alone, without an open, agile mindset, can serve as a blocker rather than a supporter of innovativeness in IT. To confirm this thesis, this study compares the impact of agile and non-agile mindset team leaders on developing dynamic capabilities in the IT sector to verify how critical an agile mindset is in IT. The Structural Equation Modeling (SEM) model was developed based on a sample composed of 474 IT employees to investigate it. Results showed that the sensing capability in the non-agile leader model is ineffective. Therefore, the innovations are hindered. The leaders with an agile mindset foster greater organizational agility and reduce the need to confront resistance to innovation. Therefore, agile mindset team leaders, in contrast to non-agile mindset team leaders, are those who can secure organizational agility in IT. So, securing ASD team leaders with an agile mindset is critical to win organizational agility in IT.

ABSTRACT. Dual-track development offers a promising strategy for integrating User Experience (UX) into cross-functional agile software development. This case study explores the practical implementation and application of dual-track development within a real-world context, aiming to identify effective practices, challenges, and impacts on workflows. A development team was supported throughout its adoption of dual-track development. Findings indicate positive influences on user-centricity, stakeholder integration, and concept work efficiency. Specifically, establishing a dedicated discovery track and systematically involving stakeholders proved beneficial. This study provides empirical evidence for dual-track development’s advantages in integrating UX within agile, cross-functional teams.

ABSTRACT. This study explores how agile software development supports user involvement in the context of relational digital transformation. Relational digital transformation offers an alternative to top-down approaches by viewing technologies as relational constructs embedded within evolving organizational practices. Achieving such transformation requires attention to the relationships that shape technologies, emphasizing sustained collaboration between users and development teams. While software plays a central role, traditional digital transformation often neglects relational dynamics. Agile software development—with its iterative process, adaptability to user needs, and emphasis on user involvement—emerges as a viable approach to implement relational digital transformation. This paper uses a qualitative case study to examine both the user involvement practices employed and users’ perspectives on these practices. The findings identify agile software development practices that can effectively support user involvement, while also revealing limitations, particularly the challenges of maintaining long-term participation and embedding user involvement beyond initial deployments.

ABSTRACT. In the era of growing digital threats, real-time detection of fake text messages has become a key challenge to ensure user integrity. The literature review examines the latest developments in the field, focusing on the use of advanced machine learning techniques and deep learning. To have a full understanding of the most recent developments, common techniques and challenges, the review covers both—methodological and practical aspects of implementing real-time systems, and discusses other aspects related to efficiency, scalability and data management. The study underscores the critical role of advanced analytical techniques in combating the rapid dissemination of misinformation in contemporary digital environments.

ABSTRACT. The strong development of artificial intelligence (AI) and its growing application in education are generating debate about its potential and challenges in higher education. This study aims to understand how students' experiences with differentiated active learning methods affect their perceptions of AI's role in academic education. A quantitative survey was conducted using a standardized survey questionnaire, including students from Poland, Romania, Greece and Croatia. The study used a quantitative approach, analysis was conducted in three analytical stages using (1) descriptive statistics, (2) hierarchical cluster analysis using Ward's method, and (3) a detailed description each of the identified groups, which allowed a comparison of their attitudes and experiences in the context of using artificial intelligence in education. The results show that students who have had experience with diverse teaching methods show more enthusiasm for AI, while those who prefer traditional methods are more cautious. Four distinct groups of students were identified who differ in their attitudes toward using AI in learning. The study underscores the importance of incorporating diverse teaching methods and educational technologies to support future competencies.

ABSTRACT. The paper analyzes teachers’ innovativeness in the context of using information and communication technologies (ICT). The research objective was to identify the most innovative groups of teachers and to examine whether their professional advancement and/or types of subjects taught influence the quality of ICT use. As a theoretical perspective, the authors have combined the Diffusion of Innovation (DOI) theory with the Technological Pedagogical Content Knowledge (TPACK) model. Data analysis indicated that the teachers with the second degree of professional advancement are more likely to use ICT tools as compared to the teachers with a higher and lower degree of advancement. Additionally, the teachers of science subjects use ICT tools more often than the teachers of humanities. The study also highlights the need for further research considering teachers’ demographic characteristics, as well as the drivers and barriers of facilitating the use of ICT tools in primary and secondary education.

ABSTRACT. This paper presents a novel method for applying gamification in university course design to enhance student engagement and allow flexible development paths. Inspired by RPG video game mechanics, the method incorporates six structural elements—main missions, side quests, boss fights, character development, specialization paths, and downloadable content (DLC). The approach was evaluated through surveys and grade analysis from 9 years and 526 students. Results show increased student satisfaction, better grade distributions, and stronger alignment with individual learning paths. The average final grade significantly increased (p < 0.001) from 4.22 (CI = 0.09) to 4.40 (CI = 0.05) thus confirming that the method had a positive impact on student performance.

ABSTRACT. Companies all over the world benefit from Industry 4.0, to fully harness its potential they need a skilled workforce. Hence, the goal of the paper is to recognize whether students, who represent the future workforce, are ready for the new technological conditions and challenges that these bring. Education at the academic level should follow and, in some cases, precede business conditions to ensure that students have the skills required in the labor market after graduating, both in hard skills and soft skills areas.

ABSTRACT. The aim of the article is to analyze the popularity and usefulness of artificial intelligence (AI) tools among higher education students and to examine how AI influences their academic, professional, and personal lives. The study was conducted on a sample of 104 students in Poland in 2025. The results indicate that chatbots and AI search engines are the most frequently used tools, and students primarily perceive AI as support in the learning process, work organization, and information acquisition. AI serves a complementary role rather than replacing traditional methods of education and work. Students show openness to the automation of routine tasks while emphasizing AI’s limitations in areas requiring interpersonal skills. The study provides up-to-date data on the integration of AI in students’ daily lives and may serve as a starting point for further research on the role of AI in education.

ABSTRACT. This study aimed to compare the effectiveness of the deep ANN embedding technique with traditional article selection methods in systematic literature reviews. The embedding model utilizes a natural language problem description to find semantically similar publications. Consequently, this technique is accessible to users without experience in data exploration. Traditional methods are represented by precise keyword queries in Scopus and Excel-based selection. Keywords used in these methods are extracted from the description by the GPT-4o model with a temperature set to zero, ensuring repeatability. The obtained results were evaluated using bibliometric metrics, which facilitate the assessment of similarities among filtered publications and enhance understanding of their connections. The findings demonstrated the superiority of the embedding model, achieving higher thematic coherence and more shared references and keywords. This approach improves the identification of relevant publications and significantly contributes to automating systematic literature reviews, which is desired in many scientific disciplines.

ABSTRACT. In today’s competitive business landscape, efficient supplier negotiations are crucial for reducing procurement costs and securing favorable terms. We present Negotia, a neuro-symbolic, purchaser-centric approach that automates concurrent supplier negotiations via AI-generated emails and contracts. Following the neuro-symbolic paradigm, our approach combines neural AI methods with symbolic knowledge representations to maximize comprehension and automation while reducing hallucinations. Negotia streamlines the process by enabling simultaneous negotiations with multiple pre-sourced suppliers, significantly reducing time and effort. Additionally, the system enables supplier sourcing and incorporates customized purchasing strategies, further enhancing procurement efficiency. Key benefits of Negotia include cost savings, improved process efficiency, and alignment with sustainability and risk management goals. We employ the System Usability Scale and the Technology Acceptance Model 2 to evaluate usability and user acceptance, focusing on perceived usefulness and ease of use. The study also includes an expert evaluation of participants’ email and contract drafts, assessing key attributes such as correctness, naturalness, completeness, benevolence, expertise, and credibility. The results reveal high usability (88.75) and strong user intention to adopt Negotia (R² of 0.888), with participants achieving high-quality outputs. Consequently, the system demonstrates the potential for substantial annual savings for large-scale procurement operations, ultimately driving digital transformation in procurement.

ABSTRACT. This exploratory study examines young people's perceptions of metaverse and its potential: how they understand the concept, the benefits and risks they associate with it, and their thoughts on how metaverse will shape the present and the future. The data was collected through interviews with bachelor’s students as part of a technology innovation course, including a metaverse experiment. Thematic analysis shows young people emphasize social aspects, including interaction and connection with real people. They see metaverse impacting the future of work and discuss education and learning alongside leisure activities. The biggest concerns revolve around information security and safety. We highlight the importance of educational and academic discussions around the limitations and possibilities of metaverse and the importance of inviting the young generation into such discussions.

ABSTRACT. This paper introduces a novel reference point-based multi-criteria decision analysis (MCDA) framework for Virtual Private Server (VPS) selection tailored to small and medium-sized enterprises (SMEs) undergoing digital transformation. Unlike traditional MCDA methods that often recommend excessively powerful or costly solutions, the proposed approach allows decision-makers to define a target solution based on specific software and operational requirements. The empirical study demonstrated that this method provided VPS rankings more aligned with actual SME needs, optimizing resource allocation without over-provisioning. The results showed stable, reliable recommendations, offering a cost-effective and practical approach for VPS selection in digital transformation.

ABSTRACT. Artificial intelligence (AI) is transforming various industries, including the field of project management. AI-based tools can optimize processes, support decision-making and automate routine, repetitive tasks, leading to increased efficiency and effectiveness. However, organizations vary in their awareness and implementation of AI in the domain of project management, which depends on factors such as industry, company size and level of digital maturity. This study aims to assess the level of familiarity and extent of AI technology implementation in IT project management organizations. The analysis covers key areas of AI use, perceived benefits, challenges and risks of implementing this technology. The results of the survey will contribute to a better understanding of the current situation and provide guidance on best practice for future AI implementation in the project management domain.