ABSTRACT. The study investigates the impact of integrating Artificial Intelligence (AI) in teaching seventh-grade mathematics, focusing on the chapter on integer numbers. It explores how AI can revolutionize tradi- tional teaching methods by providing innovative, engaging, and person- alized learning experiences. Through an experimental design involving control and experimental groups, the research aims to uncover the ben- efits and limitations of AI in education, offering insights and practical guidelines for its effective integration into teaching practices. The broader implications for education suggest that AI could be beneficial in various segments of the educational process, not only in terms of improving the understanding of specific subjects but also in preparing students for a job market increasingly reliant on AI. This perspective encourages a compre- hensive view of the educational process, beyond singular outcomes, and opens avenues for further research and application of evolving AI mod- els like large language models. The study concludes by highlighting the importance of continuing to explore new paths for education and other segments of human activity through the application of rapidly evolving AI technologies.

ABSTRACT. Croatian Academic and Research Network (CARNET) conducted a major nine-year EU-funded programme called “e-Schools: A Comprehensive Informatization of School Operation Processes and Teaching Processes Aimed at the Creation of Digitally Mature Schools for the 21st Century.” The programme was launched in March 2015 with a pilot project covering 10 percent of Croatian schools (151 schools). Based on the pilot’s experiences and results, CARNET initiated the second phase of the programme in September 2018, which included all elementary and high schools in Croatia funded from the state budget (1321 schools), lasting till end of 2023. With a total value of 212 million euros, the e-Schools programme contributed to boosting the performance of the primary and secondary education system so that Croatian students could prepare for further education and lifelong learning, i.e. for entering the labour market. This paper intends to use the concept of digital maturity of schools as an evaluation tool for assessing the programme’s impact. The Framework of Digital Maturity and the corresponding Instrument developed during the programme, provided data for assessing the programme’s main benchmark. The aim of the paper is to show how the level of the digital maturity of schools, as measured in the pilot phase and upgraded and measured in the second phase, represents the benchmark for evaluating and assessing the impact of the large-scale horizontal ICT investment in the education. More precisely, it will show how the major investment in digital infrastructure through establishing local computer networks, equipping school staff, classroom upgrades, e-services implementation, and development of digital educational materials, including teaching scenarios and staff training raised the digital maturity level of Croatian Schools. Unlike existing research, this paper will provide unique insights into the importance of the concept of digital maturity for the measurement of the programme impact.

ABSTRACT. Online discussion forums serve as dynamic environments where students and teachers collaboratively generate and utilize a wealth of content for knowledge sharing and assessment. The research involved 18 informatics graduates at the Faculty of Science at the University of Mostar, Bosnia and Herzegovina and two teachers who extracted and analyzed transcripts from an online forum, which was part of the online course "Evaluation of E-Learning Systems (EES)" held on Moodle during the winter semester of the academic year 2022/2023. The paper introduces a novel Natural Language Processing (NLP) approach to evaluating student contributions by contrasting their postings with corresponding instructional materials. Utilizing text similarity measurement, the research addresses key questions: Does the content extracted from individual student postings reflect student knowledge on a given topic? Do similarity scores align with human rankings of contribution relevance? Do students equally benefit from collaborative learning? The research evaluates the efficacy of five multilingual sentence embedding models and integrates human analysis to assess the relevance of students' contributions. Contributions of this study include the evaluation of multilingual sentence embedding models and a thorough examination of the human-perceived relevance of student contributions. The findings aim to enhance the understanding of whether this approach can effectively assess and validate the educational value of student discussions within online forums and contribute to the optimization of collaborative learning experiences.

ABSTRACT. The application of gamification in the learning process is developing approach for increasing learners’ motivation and engagement by introducing game elements into the educational environment. Gamified learning activities allow learners to acquire knowledge, improve skills and encourage positive traits that the game builds specifically for learning purposes. Gamification can enhance learners' engagement, increase their motivation toward learning, promote collaboration, provide instant feedback, and establish a positive learning environment. The effects of gamification are mostly positive, but there are also certain challenges or shortcomings, such as the development of such technologies in education is expensive, the question arises of the security of learners' personal data, the impossibility of determining the long-term effects of gamification, and how to adapt the elements of gamification to the different personalities of learners. Artificial intelligence (AI) allows gamification to be adapted to a learner in a real-time environment. AI has become very important for the learning environment, where the individual needs of each learner take centre stage. One of the most important goals of Artificial education is to provide individual learners with personalized learning suggestions on aids depending on their learning status or personal preferences. Tailoring content to unique learning preferences becomes a unique learning journey and experience for each individual learner.

ABSTRACT. The e-Schools Programme is an initiative of the Croatian Academic and Research Network - CARNET- to digitalize the Croatian primary and secondary education system, bridging the gap between traditional methods and modern digital solutions. Earlier solutions should have uniformly integrated digital technology in all schools in Croatia, often due to insufficient infrastructure and education of school staff. Lack of knowledge of new technologies limited digital solutions, such as the e-Class Register, the electronic class register for primary and secondary schools. The e-Schools Programme brought a comprehensive solution, offering not only the necessary equipment and infrastructure but also education and support for teachers in using new technologies, which directly impacted the increase in the use of the e-Class Register. The methodology used in this paper included a quantitative and qualitative analysis of the use of the e-Class Register, evaluating user satisfaction and the efficiency of administrative processes. The key results show a significant increase in using the e-Class Register, greater transparency, and organization in schools, highlighting the e-Schools Programme as a successful model for modernizing education.

ABSTRACT. eacher Competence Frameworks were explored from the perspective of the competences needed for teachers to understand and successfully use Arti- ficial Intelligence (AI) in education. A database search for Continuing Profes- sional Development (CPD) programs for Croatian school teachers was conducted to address the following research questions: What topics related to AI in educa- tion are represented in informal Continuous Professional Development (CPD) opportunities for Croatian school teachers at the national level? What categories of digital competences, based on DigCompEdu, are represented in these oppor- tunities? How many teachers in Croatia have taken advantage of the available opportunities for CPD related to AI in education? The results and discussions revealed that the proportion of activities specifically related to AI in education is notably low, with certain areas and topics receiving more attention than others. Further research should be conducted to establish the reasons for gaps and ine- qualities in the offerings of professional development programs for teachers in the field of AI in education. The overall offerings of AI-related professional de- velopment programs for teachers should be broadened. The topics should be more diverse, and the DigCompEdu areas should be more equally represented.

ABSTRACT. Student behavior in the online environment includes learning activities and knowledge testing using e-learning platforms. We researched to determine the behavior of students in an online environment with 175 first-year undergraduate students of the Faculty of Science, University of Split, in the course Programming 1 in the academic year 2021/2022. The course covered the basic principles of programming and the theoretical and practical aspects of application development using the Python programming language. The teaching content was delivered through the online platforms Moodle and CloudMap&Flash with four areas and nine teaching units. The teaching process took place according to the hybrid model and was carried out from 4-Oct, 2021, to 20-Feb, 2022 in four research phases. The research aims to determine how students perceive and engage in online course activities on compulsory and optional e-learning platforms. We developed an original course activities model with criteria to measure the behavior of students in learning activities and knowledge assessments.

ABSTRACT. Integrating math, computer science and technology into the classroom can improve students' understanding and appreciation of these subjects, and using games to illustrate algorithms is particularly effective in making abstract concepts more concrete and accessible to learners.

The idea behind [ai] explore exhibition is to show the application of math and computer science in solving real-world engineering problems. By presenting visualizations created with the HEDAC algorithm, students can see first-hand how mathematical concepts are used in practical contexts.

The development of two games, [ai] explore game using GeoBoard and a game using micro:bits, to explain the algorithm behind the exhibits is a creative way to reinforce learning. Games have the potential to make learning more interactive and fun, thus increasing student engagement and understanding. In this paper, we describe how new math and engineering results can be explained and turned into a playable game that can be used in the classroom. Finally, we present the results of a short survey of students about their attitudes towards the games presented and the results of the activities carried out.

ABSTRACT. This systematic review investigates educational robotics' impact on enhancing learning experiences. It focuses on student engagement, motivation, and learning outcomes across various educational levels. Through the analysis of studies from elementary to university-level courses, a significant positive impact of robotics on learning performance and the development of 21st-century skills, such as computational thinking and problem-solving, is identified. Robotics is highlighted as a powerful tool for creating interactive and immersive learning environments. These environments support academic achievement, critical thinking, and collaboration. However, challenges like accessibility, teacher training, and curriculum integration are also discussed. The review suggests that educational robotics has transformative potential for pedagogical practices, deserving further exploration and integration into educational curricula.

ABSTRACT. Artificial intelligence (AI) is a powerful technology that can transform traditional business models and legal processes, offering new opportunities and challenges for both companies and legal profession- als. This article explores the fundamental nature of AI, using influential studies and research to provide insight into its complex characteristics and consequences. Using case studies and practical examples, we illus- trate the pragmatic use of AI in different domains, including supply chain management, marketing, and strategic decision making. Moreover, the article examines potential opportunities and possibilities in AI technol- ogy, considering emerging trends and their potential impact on business models and legal practice. The article serves as a valuable resource for understanding how these technologies are revolutionizing the economic and legal dynamics within enterprises.

ABSTRACT. Decision tree, as a decision-making method in conditions of risk and uncertainty, represents a practical graphical-visualization tool for decision-making. Additionally, modern businesses, especially banking and other financial institutions, handle large volumes of data, making an appropriate tool more essential than ever. Machine learning and associ- ated predictive models, such as decision trees and random forest, provide a solid foundation for creating systems to make quality and timely de- cisions. This paper explains the background of the decision tree method and related models that are based on it. In the end, our own implemen- tation of a system for detecting negative customer comments based on a synthetic dataset using predictive models is presented.

ABSTRACT. The article provides a comparison of Convolutional Neural Network (CNN) and Reinforcement Learning (RL) applied to the field of autonomous driving within the CARLA (CAr Learning to Act) simulator for training and evaluation. The analysis of results revealed CNNs better overall performance, as it demonstrated a more refined driving experience, shorter training durations, and a more straightforward learning curve and optimization process. However, it required data labelling. In contrast, RL relayed on an exhaustive (unsupervised) exploration of different models, ultimately selecting the model at timestep 600,000, which had the highest mean reward. Nevertheless, RL’s approach revealed its susceptibility to excessive oscillations and inconsistencies, necessitating additional optimization and tuning of hyperparameters and reward functions. This conclusion is further substantiated by a range of used performance metrics (objective and subjective), designed to assess the performance of each approach.

ABSTRACT. Medical images are often exceedingly large in width and height, limiting the maximum batch size when training convolutional neural networks and requiring models with a large number of parameters. Typically, images are uniformly downsampled, leading to losing fine-detailed information. Instead of uniformly downsampling images, we introduce a two-stage end-to-end segmentation network utilizing image crops to reduce network input size. Initially, a uniformly downscaled image is first segmented with a rough segmentation module, and the rough segmentation is used as a saliency map to crop the original high-resolution image to a region of interest. This crop is then re-segmented with a fine segmentation module. Our method's effectiveness is demonstrated in segmenting lesion boundaries in clinical images across two datasets. We establish that this technique maintains comparable segmentation quality to a baseline model while reducing the network input size. Furthermore, our approach enhances the robustness of segmentation outcomes with smaller input sizes, outperforming uniformly downscaled images and baseline models. This improvement is consistent in both in-sample and out-of-sample evaluations.

ABSTRACT. The domains of inspection, emergency response, and surveillance have been revolutionized in recent years by the increase in capabilities of autonomous mobile robots (AMRs), such as unmanned aerial vehicles (UAVs), unmanned ground vehicles (UGVs) or even quadrupedal robots. While a large body of work covers topics addressing static scenarios, working with dynamic points of interest remains relatively problematic. The nature of the problem brings with it the real time adaptability challenges, efficient decision making, and uncertainty. Available literature mostly concentrates on research-oriented specialized tools running controlled experiments. However, there is a lack of comprehensive frameworks supporting coverage of scenarios dealing with non stationary objects.

This paper introduces a multi-object planning framework for autonomous mobile robots operating with dynamic points of interest. The framework integrates open source software and low level frameworks to simulate multiple agents moving on unknown trajectories. Advanced planning algorithms can be deployed and tested in simulation, as well as the real world AMRs thanks to the software-in-the-loop (SITL) approach. The architecture uses ROS for high level communication, Ardupilot for interfacing with the vehicle hardware, and Gazebo for the physics simulation. Core modules allow configuring various dynamic agents and implementing various planning algorithms. As a proof of the system capabilities, use cases of ship inspection and agriculture monitoring using a UAV are presented. The resulting framework can serve as a basis for research, education, and deployment purposes on the topic of advanced planning algorithms for AMRs.

ABSTRACT. Adaptation to the level of knowledge of each student remains one of the key challenges of e-learning and education in general. E-learning systems provide opportunity for systematic data collection about learning activities offering valuable insights into the students’ knowledge. In order to achieve the personalised learning, this study introduces a Knowledge Design and Delivery Model (KDDM) for intelligent tutoring systems. This model uses a hybrid approach that combines traditional overlay student models with fuzzy logic and multi-criteria decision-making methods. Unlike popular machine learning approaches, these methods do not require existing datasets and they allow direct teacher involvement in knowledge delivery. The KDDM associates student stereotypes with Bloom’s revised taxonomy levels, providing a reference point for the cybernetic model. KDDM has been successfully implemented and examined in a two-year experiment which confirmed its effectiveness on 370 participants from two universities in two countries.

ABSTRACT. Artificial intelligence (AI) is a powerful tool that has been widely used in various industries, including software development. In this study, we explore the perceived impact of AI tools on the quality of software development code. The study aims to provide a comprehensive understanding of the current state and potential future trends of arti- ficial intelligence in software engineering. Through a survey conducted in various tech companies, the findings of this study aimed to provide insight into the effectiveness of AI assistance in software development, particularly focusing on code quality. The overall results show that there is high satisfaction among developers using AI tools, with more than three-quarters of them stating that the adoption of these tools positively impacted their overall satisfaction and productivity in the software development sector.

ABSTRACT. With integration of collaborative robotics, robotic vision, and machine learn-ing, the high-end frontier reached in assistive robotics is likely to help devel-op promising solutions to the challenges associated with the needs of aging populations and persons with disabilities. This paper examines recent developments, challenges, and opportunities that relate to the application of these technologies for human assistance. The paper will begin with an explanation of the basic concepts and the rationales driving assistive robotics, and then move on to the technological ground with basic collaborative robots, robotic vision, and machine learning, stressing how they can make human assis-tance better. Case studies and real-world applications will help to present the potential of integrated assistive robotics not only in healthcare and rehabilitation, but also in domains such as elder care and living on one's own. This paper also highlights some of the critical challenges inclusive of robustness, reliability, human-robot interaction, safety, and ethical considerations along with some of the emerging trends and future research directions related to it. The paper sets up a visionary outlook, where integrated assistive robotics—realized fully through human-centered design, interdisciplinary collaboration, and innovation in technology development—could revolutionize care and support toward fostering the inclusivity, dignity, and independence of those with diverse needs.