The opinions and attitudes of prospective primary school teachers on the use AI applications in education
|
|
Teaching (Today for) Tomorrow: Bridging the Gap between the Classroom and Reality 3rd International Scientific and Art Conference |
|
|
| Section - Education for digital transformation | Paper number: 1 |
Category: Original scientific paper |
Abstract |
|
|
|
Key words: |
|
|
Introduction
InArtificial today'sIntelligence digital(AI) society,has recently experienced unprecedented growth in whichmany we have grown upindustries, and areeducation activeis members,among the most promising fields for its implementation. Although the potential of AI has been recognized for quite a long time, it is clearonly in the last couple of years that research has shifted more and more towards the integrationeffects of informationAI technology withinin the educational systemfield. isThis becomingincrease increasinglyin significantinterest has occurred alongside the developments in AI technologies like machine learning and natural language processing, which revolutionize education (Bognar,UNESCO, 2016)2023; World Economic Forum, 2023). In education, the contextcapability of computerAI science,to teachersimprove faceteaching and learning cannot be overemphasized since service delivery has been personalized and made more accessible. In this context, AI refers to systems that can acquire and process large amounts of information, identify learning patterns, and provide individual learning processes. Learners can also progress through their learning curves and at their learning rate, effectively producing results (Alneyadi et al., 2023; HolonIQ, 2023). In addition, ITS, chatbot, and auto-grading solutions can decrease the dualteachers’ challengeworkload and make them focus more on the invention and improvement of teachingthe foundationalcreation knowledgeof meaningful interactions with students and enablingon the development of thinking skills (World Economic Forum, 2023). AI technologies can be used in learning as recommendations in potential development areas and optimization of tasks based on a student’s capability. This leads to a situation where even the poor performers and the bright students toreceive overcomethe keyassistance barriersthey require to understanding.be productive. Personalized learning platforms have been shown to engage students more effectively, boosting motivation and improving academic outcomes and retention rates (Gningue et al., 2022; Ibrahim et al., 2022). AI can notice that a student requires assistance by analyzing the student’s performance profile. Instead of providing general assistance, it assists where the student is most likely to require help (Alarabi & Wardat, 2021). AI systems can analyze student performance data to identify areas that may require special attention. These obstaclessystems includecan thresholdhelp concepts—detect corestruggling ideaslearners thatearly representand aprovide transformativesupport pointsbefore inissues learning.get Theseout conceptsof arehand. notFor onlyinstance, fundamental,using butbig alsodata servecan astrack athe gatewaystudent’s activity, attendance, and performance to deeperidentify understandingthose thatat often requires a significant shift in perspective to master.
Meyerrisk and Land (2003) have identifiedsuggest the thresholdappropriate conceptintervention or change in the fieldlearning approach (Niall McNulty, 2023). Moreover, other AI systems, such as Mindspark, assess the students’ responses and recommend the right exercises and materials to fill the students’ knowledge gaps and provide each learner with the right level of challenge and support (Calibraint, 2023). AI systems in education ascan aalso setcontrol the speed of ideaslearning, that,thus onceallowing understood,fast becomelearners transformativeto butprogress arefaster initiallywhile challengingslow andlearners unfamiliar.can Regardlessprogress ofat whethertheir weown adoptpace. aThis constructivistlearning approach ormakes anotherit learningpossible theory, threshold concepts represent points at which students are likely to encounter learning difficulties. To further define threshold concepts, Meyer and Land state that they are integrative, as they show previously unknown ways of linking ideas; irreversible, as the new way of thinking becomes part offor the learner onceto theylearn haveat trulyhis understoodor it;her own pace, depending on his or her ability, making the learning process very effective. For instance, a study reveals that AI can use the student’s performance data to offer feedback and boundaryrecommend markers,learning asresources theysuitable definefor the boundaries of partstudent (orBenzakour all)et ofal., a set of ideas. An entire subject area may have its boundary marked by a single threshold concept, mastery of which indicates competence in that area (Meyer & Land, 2003)2022).
Dr.They Tuckeralso highlightsassist educators by pointing out students who may require extra attention and assistance. In this case, AI makes it easier to provide targeted support, guaranteeing that every learner gets the help required to succeed (Kurilovas, 2018). In addition, adaptive learning platforms have been integrated into different educational contexts. They have helped students learn at their own pace and enhanced both the interest and the drop-out rates (Chiu et al., 2023).
Harry & Sayudin (2023) highlight that there is a potential for artificial intelligence to change the educational scenario; however, there are specific barriers that require attention for this to happen. It is beneficial as it can personalize lessons for students, reduce administrative tasks for staff, and provide real-time learning updates during class. Still, they have to calculate the benefits and risks, protect information gleaned from students, and control prejudices from AI systems to make integrative authorities feasible. When this comes together with the limitations, AI could make you more efficient and deliver educational help to all-inclusive students. Integrating AI in education is rewarding with several characteristicsbenefits but presents core concerns that have to be strategically resolved. The challenges of thresholddata conceptsprivacy, bias in algorithms, and emphasizesthe fivefeasibility mainof characteristics:rolling transformative,out irreversible,AI integrative,systems troublesome,have been factors of great concern. Training affected the AI systems, proving adverse as they replicated the biases in the training data and boundedcreated (SJSUgaps Schoolin areas such as provision for healthcare services and allocation of Information,resources, 2013)including special education for various groups of students (Sutaria 2022). With this, privacy concerns come into play when AI categories together and utilize students' information without consent, raising concerns about surveillance and data usage governance in educational settings (GUJJULA et al., 2023; McNulty, 2023).
ResearchOther important factors are ethical considerations such as fairness and accessibility of the content. AI should be developed and deployed in a way that does not reinforce prejudices and provides equal opportunities for all learners. This includes explaining how AI algorithms work and use data (UNESCO, 2023). Constant supervision and assessment are needed to ensure the process is fair and does not perpetuate societal educational inequalities. For example, Baker & Hawn (2022) note that caution is needed when implementing AI in education since it can lead to algorithmic bias that is unfavorable to some students. They recommend the use of diverse data in order to reduce bias in the AI models. Likewise, Chaudhuri & Mohanty (2023) also talk about the need for bias-aware algorithms to capture and address biases that might be inherent in historical data patterns. In this way, institutions can ensure that AI systems are trained on thresholda conceptsdiverse set of data to reflect the diverse student body. Malik (2024) highlighted the importance of continuously using diverse and representative data and monitoring the AI systems’ performance to create efficient and ethical AI systems.
Furthermore, the UK’s Information Commissioner’s Office (ICO) also notes that fairness in computerAI scienceis hasnot highlightedonly certaina conceptstechnical concept that refers to algorithmic fairness. It needs a more extensive and context-based approach that takes into account the social and legal consequences of data processing, so that AI does not enhance the existing unfair distribution of resources (ICO, 2023). By implementing these principles, AI systems can be more explainable, non-discriminatory, and inclusive to all students with equal opportunities to access educational materials. There is a great deal of interest and controversy regarding the use of AI in education, as transformativethe presence of AI in the educational process increases. Since AI technologies like intelligent learning platforms, automated grading tools, and challengingintelligent fortutoring studentssystems asare theybeing oftenincorporated requireinto significantclasses, cognitiveteachers' change.perceptions Amongof thesesuch concepts,technologies object-orientedgive programminginsight standsinto outthe duefuture toadoption itsof complexityAI andtechnologies potentialin toeducation. enableThese deeperperspectives understandinginfluence the integration and application inof differentAI areastools and raise the question of computer science (Boustedt et al., 2007). However, the conceptsrole taughtof technology in primarylearning school have not been the focus of such research, leaving a gap in the understanding of threshold concepts that younger students should overcome.contexts.
ThisThe aim of this paper examinesis to identify the thresholdperception conceptsof students from the Faculty of Teacher Education at the University of Zagreb towards the application of Artificial Intelligence (AI) in computer science education, with a focus on primary school.learning. The research aims to identify conceptsthe thatpositive and negative attitudes students have towards AI, which subject they think should be supported by AI, and which AI applications are particularlymost challengingbeneficial for studentsteachers and students. The study, conducted through an online questionnaire, aims to distinguishidentify those that can be considered threshold concepts based on their characteristics. Using the nominal group technique,prospective primary school teachersteachers’ wereattitudes involvedtoward inthe a structured research processbenefits and thresholdpossible concepts were proposed on the basisdrawbacks of shared insights. By addressing this topic, this research aims to improve teaching practices and support learning and teaching, as by understanding these concepts, teachers can better address the obstacles students face and guide them towards understanding the key concepts.
Characteristics of threshold concepts in education
In this section, the following characteristics of threshold concepts are described according to Meyer & Land (2003): transformativeness, irreversibility, integrativeness, troublesomness, and boundedness.
Transformative is described and associated with events that leave a lasting impression and are unforgettable, such as passing a driving test. The transformation of attitudes, values or understanding often represents a decisive point in our lives. This change not only shapes our identity, but also has a profound impact on our daily lives. Through this development, the new understanding gradually integrates into our biography and becomes an inseparable part of who we are. This process does not happen immediately, but unfolds gradually and permeates all aspects of our lives. It is not just a matter of adopting a new attitude or a new value. Rather, this transformation becomes part of our inner being and shapes the way we perceive the world around us. Through this integration, the new understanding becomes a fundamental element of our identity and influences our thoughts, feelings and actions. This change does not occur in isolation, but has a profound impact on our relationships, our work and our life choices. The importance of such transformations lies in their ability to promote growth and development as individuals. They encourage us to look at things from different angles, which gives us a broader view of life. Furthermore, these changes often coincide with personal growth and make us stronger and more resilient to life's challenges.
Irreversibility,AI in the context of knowledge,teaching standspractices, foras awell profoundas leveltheir preferences regarding the use of learning,AI where what we have mastered becomes an integral part of our intellectual repertoire. It is like riding a bike or swimming, where a learnt skill becomes inherenttools and indelible.technologies Throughin the process of irreversibility,learning knowledgeand becomesteaching. imprintedThis study provides a systematic discussion of the possibilities and concerns of AI in ourthe memorylearning environment, and its purpose is to inform the future directions for teacher preparation and policies in integrating AI into primary education responsibly.
Teachers' attitudes toward the use of AI in education
AI's integration into education has brought together a range of feelings and opinions from educators, which span from optimism to caution. Teachers broadly believe in the benefits of AI, primarily in augmenting personalized learning while, at the same time, lightening administrative work. For example, AI can provide services that relieve teachers of the burden by automatically providing feedback and monitoring students' progress and automating tasks such as grading, thereby giving the teacher ample time to teach and offer more personalized services. This, according to Kim (2023) and Wang et al. (2023), can go a long way in minimizing teacher burdens. Moreover, a lot of educators believe that resistsAI forgetting,has the potential to create customized learning experiences for students, allow for differences in learning speed, and even promote emotional well-being through the use of AI-powered chatbots (Kim, 2023; Pejnović, 2024). In addition, AI enhances the possibility of real-time monitoring and analysis of student performance, thus enabling the teacher to intervene on time when some students struggle (Kostrić, 2021). For example, intelligent tutoring systems make individual feedback available through student performance analysis and help students in challengingtheir situations.weak Thisareas phenomenonwithout waiting for conventional testing periods. Most educators believe that with differentiated instruction, AI will significantly help close achievement gaps and ensure that all students reach their full potential. Some subjects, like mathematics, science, and language learning, were found more effective with ITSs, which give domain-specific feedback in detail (Kurni et al., 2023; AIWS, 2023). As AI continues to evolve, the integration of natural language processing and emotional recognition in ITSs promises to further enhance personalized learning by responding not only to cognitive but also affective states of learners (Fernández-Herrero, 2023).
However, there are quite a few fears among teachers concerning the potential problems of artificial intelligence. One of the most severe concerns are the data privacy issue and the ethical implications of using AI in schools. Many teachers fear that the extensive data collection for AI-enabled personalization can be comparedproblematic tofor ridingstudent aprivacy, bike.especially Asif soonsensitive as we master a technique, it becomes part of our muscle memory. No matter how long we have not ridden a bike, when we do it again, the process naturally emerges from our subconscious. Similarly, irreversibility in learning means that once acquired, knowledge becomes a permanent skill that is activated regardless of a prolonged period of disuse.
Integrative learning means that what was previously hidden or not fully understood is made accessible in its context. This quality of learning has the power to link separate concepts together so that they are brought together into a holistic understanding. Ideas that were once separate are now connected, creating a broader understanding that enriches individual perspectives. This integration process can be likened to putting together pieces of a jigsaw puzzle. Each individual piece represents a particular concept, and through integration, these individual pieces become an integral part of a larger, complete landing. Furthermore, integrativenessinformation is not justhandled aboutproperly putting(Felix, different2020; conceptsBeović, together,2023). itFurthermore, goesconcerns ahave stepbeen furtherraised byregarding creatingthe anpotential expanded understanding that enriches our perceptionloss of teacher autonomy and the world around us. This dimensionhumanizing of integrativenesseducational significantlyprocesses influencesdue to artificial intelligence. According to educators, AI should not replace the vital human aspects of teaching, such as emotional support and the development of individualcritical understanding.thinking Ideasskills, which are crucial for student development (Beović, 2023; Guilherme, 2017; Wang et al., 2023). Teachers seem indifferent regarding their preparedness for the implementation of AI in schools. While some teachers believe that they can use AI with ease, others observe that this cannot be achieved without deeper training that will equip them with the relevant skills in applying AI (Kim 2023). Conversely, other studies have previouslyestablished isolatedthat nowpreparation becometo partuse AI differs depending on prior use and resource experience (Wang et al. 2023). Teachers advocate for specialized training initiatives that encompass both the technological components of aartificial widerintelligence networkand the associated ethical implications, thereby guaranteeing their preparedness to utilize AI responsibly within the educational setting (Felix, 2020; Alharbi, 2024).
One major factor affecting teachers' attitudes towards AI is their readiness to integrate AI into their teaching practices. However, surveys and studies show that most teachers feel unprepared and ill-equipped for the integration of connections,AI leadingtools—often due to a richerlack of training or resources required for use in classrooms (Beović, 2023; Wang et al., 2023). Teachers who feel unconfident about using AI systems might not use these resources as intensely, widening the gap between those who benefit from AI-supported learning and those who do not (Kim, 2023). Professional development, therefore, becomes necessary to empower the teachers with knowledge of AI technologies and imbue them with the confidence to apply these tools in their pedagogical approaches. Educators are asking for more training that would highlight ethical AI use, data privacy, and the potential effect AI can have on the relationships within the classroom (Kostrić, 2021; Felix, 2020). These initiatives would assist educators in managing the intricacies of artificial intelligence, ensuring that its deployment is consistent with educational objectives and principles (Wang et al., 2023).
Professional development of teachers in an effective way to overcome this gap in AI-supported learning needs to be geared toward both the technical knowledge of AI and deeper experienceunderstandings of knowledge.
Troublesomeness,issues thrown up by AI-data privacy concerns, with its potential consequences for classroom relationships. This would imply that such professional development should align with emphases within the DigCompEdu framework on empowering teachers across multiple dimensions:. In developing their own competencies with digital resource development, pedagogical methodologies, and evaluative techniques, educators might become more confident in learningexercising responsible use of AI tools. Furthermore, such professional development would allow instructors to better nurture students' digital literacy in ways that help them grapple with the ethical and practical problems linked to AI. This comprehensive strategy guarantees the incorporation of artificial intelligence in a manner that improves educational results while preserving the human-centric elements inherent in pedagogy (Kostrić, 2021; Kim, 2023; Wang et al., 2023). This balanced approach is based on the understanding that AI is not a substitution for teachers but rather their augmentation that helps with grading and lesson planning so that the teacher can bespend associatedmore andtime describedinteracting with certain concepts that may seem counter-intuitive or unpleasant. However, it is essential to face these challenges in order to understand them. Often these concepts are associated with situations that cause discomfort or are counter-intuitive, and this discomfort may stem from misconceptions. Especially when solving problems in physics, beginners are often confronted with various misconceptions and contradictions. However, when they dedicate themselves to solving these challenges, they not only overcome their preconceptions but also reach a new level of understanding. Wrestling with counterintuitive ideas becomes a path to deeper understanding, and this process allows individuals to reach new heights in their learning eliminating underestimation in the process.
Boundedness implies the presence of definitive boundaries. These boundaries serve as transitions between different conceptual areas, defining boundaries to other thresholds and introducing us to new areas of understanding. In a particular subject area, specialized terminology takes on a new meaning defined precisely by these boundaries. Boundedness implies not only the presence of endpoints, but also the possibility of exploring and expanding these boundaries to deepen our understanding and discover new meanings that emerge within these defined boundariesstudents (MayerAlharbi, and2024; Land,AIWS, 2003), (SJSU School of Information, 2013)2023).
ThresholdApplications conceptsof AI in computer science education
The research on threshold concepts in computer science conducted by BoustedtNg et al. (2007)2023) focusedperformed ona identifyingsystematic termsreview thatof could49 correspondstudies toover thresholda conceptsperiod of two decades, with the result being a much-needed insight into the growth of AI education. The review highlights three key research questions: types of learners involved in AI education, tools used, and validatingteaching themapproaches withapplied. students, followed by checking whetherIn the criteriaearly fordays, thresholdAI conceptseducation are met. At the Conference on Innovations in Computer Science Education in June 2005, 33 computer science experts from nine countries were surveyed to select terms that met the criteria for threshold concepts. In November 2005, a similar study was conducted at a conferencefocused on computer science ineducation Finland,at the resultshigher ofeducation whichsector focusedlevel. onBy the2021, „hardhowever, toAI learn”literacy aspectshad ofgained threshold concepts (McCartney and Sanders, 2005). Subsequent studies at different universitiesmomentum in severalK-12 countries showed that students identified „control structures”, „sequential thinking”, „parameters”, „objects” and „memory models” as threshold concepts. Object-oriented technologies and pointers were selected for in-depth analysis as they fully met all the criteria for threshold concepts (Boustedt et al., 2007).
Object-oriented programming (OOP) is based on the idea that a program consists of objects that represent interconnected parts of a solution, in contrastdue to the classicdevelopment proceduraland model, which views a program as a sequenceemergence of instructions.age-appropriate OOPteaching enablestools. moreThe efficientpedagogical code organization and simplifies the maintenance and scaling of large programs. It is used in languages such as Java and Python, while the procedural model is suitable for languages such as C and Pascal (Jovanović, 2012). Although OOP offers numerous advantages, students often report difficulties in learning it, especially with basic concepts such as classes and objects. Research among first-year university students shows that many have experienced OOP as a challenge that requires rethinking. Nevertheless, most students emphasize that mastering OOP has helped them to understand more complex programming concepts and enabled them to transfer the skills learned to other areas, such as software engineering, demonstrating the transformative nature of OOP. Such experiences show that although the learning process is long and complex, it provides long-term benefits and enables students to apply their knowledge in different contexts (Boustedt et al., 2007).
Pointers have beenstrategies identified as a threshold concept in computer science because their understanding is often challenging for students, especially when they are used as parameters in programs. Students reported difficulty connecting abstract theory to the practical application of pointers. One student described having difficulty understanding pointers until he realized that they simply represent a specific memory location, which made the concept clearer. Having mastered pointers, students began to apply this knowledge in broader contexts such as hardware and operating systems by using pointers in practical work, for example in assembly language. Understanding pointers enabled them to apply objects and references more successfully in programming, leading to greater confidence in tackling more complex computing problems (Boustedt et al., 2007).
Recent research by Kallia and Sentance (2020) provides additional insights in the contextarticle are dominated by collaborative project-based learning and the use of functions.game Theirelements. studySuch emphasisesmethodologies encourage problem-solving, creativity, and engagement in students. Teaching tools leveraged for scaffolding AI concepts in students ranged from robotics to software platforms and intelligent agents. Nevertheless, challenges were noted, including the transformativelack of suitable resources for young learners and integrative nature of concepts such as „parameters”, „parameter passing”, „return values”. These concepts were found to be challenging for secondary school students and their mastery facilitates deeper understanding and integration of knowledge across programming topics. The authors identified „procedural decomposition” as a potential threshold skill requiring extensive practice to bridge theoretical understanding and application.
The study conducted by McSkimming, Mackay & Decker (2023) identified two threshold concepts for intermediate non-computer science students: „algorithmic runtime”majors and „memorythe management”,complexity emphasizingof some AI concepts. The review highlights the challengesgrowing importance of AI literacy and transformativeshows understandingthe associatedneed withfor theseeducators concepts.to Theadapt researchtheir byteaching Govenderpractices &to Olugbaraintegrate (2022)interdisciplinary reflectedand oninteractive thresholdtools that would make AI concepts for developing programming skills in first-year information technology students, providing insights into supporting teaching strategies and identified procedures and functions and programming constructs (such as selection, iteration, and variable manipulation) as critical threshold concepts for teaching computer programmingunderstandable to first-year ITall students.
These studies highlight the ongoing interest in understanding and addressing threshold concepts to enhance computer science education.
MethodologyTools designed to support teachers
Intelligent Tutoring Systems are among the most prominent AI tools used in education. These systems simulate one-on-one human tutoring by providing personalized learning experiences to students based on their individual needs. They use algorithms to track student progress, analyze areas of difficulty, and provide tailored feedback. For instance, platforms like ASSISTments in the United States help students, particularly in mathematics, by offering tailored instruction based on their performance in practice exercises (Holmes & Tuomi, 2022). ITSs have shown particular efficacy in subjects like mathematics, science, and language learning. In mathematics, ITS platforms can guide students through problem-solving processes by offering hints, identifying misconceptions, and providing real-time feedback (Luckin et al., 2016). For example, in physics and chemistry, AI-driven platforms like Knewton and ALEKS adapt content delivery based on students’ responses, helping them grasp complex concepts through gradual mastery of smaller components (Su & Yang, 2023). In language learning, AI-powered tutors like Duolingo offer personalized exercises based on a student's proficiency level, accelerating language acquisition through repetitive and targeted practice.
Systems like Carnegie Learning utilize AI algorithms to evaluate a student's comprehension and modify lesson plans, ensuring that each concept is mastered before moving on to more advanced topics (Cope et al., 2020). Such systems are particularly effective in subjects like mathematics and science, where understanding builds on previous knowledge. A significant benefit of adaptive learning platforms is their capacity to customize the learning experience for students with different skill levels. For example, students who find fractions or algebra difficult may receive extra practice exercises, while more advanced students face more challenging problems (Perrotta & Selwyn, 2019). This method guarantees that all students, no matter their learning speed, can achieve a solid understanding of the subject.
Furthermore, the use of AI platforms enables students to visualize the results of experiments that are hard to conduct in real life, such as ecosystems and chemical reactions. This also helps students to grasp concepts that are abstract in nature and also enables learning through experimentation (Holmes & Tuomi, 2022).
Another learning application of AI is in the area of tests and evaluations , where students can be given results and feedback immediately. Software such as Gradescope and AutoGradr employ machine learning to self-grade students’ assignments and tests, thus freeing teachers’ time while giving students timely and uniform feedback. Assignments in subjects such as computer science and mathematics entail code or problem-solving, and AI tools can check the correctness and effectiveness of the student’s responses (Cope et al., 2020). Feedback systems that are facilitated by AI also facilitate self learning since students can go through their mistakes and enhance their comprehension before the final submission of their work. In writing and composition courses, technologies such as Grammarly give immediate feedback on grammar, syntax, and style, which can be used to improve the students’ writing skills through repeated practice (Su & Yang, 2023).
ClassDojo is an example of an AI-based application that helps teachers to maintain discipline and control the level of students’ participation. ClassDojo enables teachers to monitor student behavior in the classroom in real time, communicate with parents, and differentiate lessons according to students’ needs. By so doing, this tool helps instill a more organized and responsive classroom management system and thus, allows the teachers to spend more time teaching rather than managing the class (Carnegie Learning, 2024).
Otter.ai and ModMath are great examples of AI technologies that are rather helpful for students with disabilities. Otter. ai allows students with hearing impairments to follow lectures and discussions by providing real-time speech transcription. ModMath, on the other hand, is intended for students with dyslexia or motor skill challenges, enabling them to solve mathematical problems on a computer without writing by hand. These tools enhance more participation and access to classroom content and materials so that all students can participate irrespective of their physical or learning disabilities (Khan Academy, 2023).
Tools designed to support students
In addition to being helpful for teachers, AI solutions offer several tools that are very helpful for students, including individual learning, improving cooperation, and creativity. One of the most engaging ways AI supports student learning is through educational games, such as Minecraft: Education Edition and DragonBox, two of the best applications for kids and parents. In Minecraft, AI builds dynamic educational experiences and makes programming, mathematics, science, and history come alive for students. This way, the game is designed to fit the student’s answers, and thus, every learner is engaged at his or her level of understanding, promoting deep learning while exploring and solving problems (Alawajee & Delafield-Butt, 2021). Minecraft in education benefits learning and social engagement. International Journal of Game-Based Learning (IJGBL), 11 (4), 19-56.). Likewise, DragonBox employs the use of artificial intelligence to tailor math problems to the learner’s level and thus is especially beneficial in teaching the basics of math (Holmes & Tuomi, 2022).
Smart applications like Socratic by Google, and Photomath change the way students solve their assignments. Socratic uses artificial intelligence to provide students with answers to questions they pose on diverse topics and in the process provides individualized explanations that are suited to the student’s learning ability and speed (D'Mello & Graesser 2023). It is most beneficial for homework and study purposes and offers extensive answers to the questions in academic areas such as mathematics, science, and humanities. Likewise, Photomath is an AI-based app to solve mathematical problems by capturing the problems and providing solutions along with the explanations to help students to learn the concepts behind the solutions (Zain et al., 2023). These tools are very useful in assisting students in dealing with complex topics since they are given real-time feedback and instructions.
The mainuse of AI has also improved language learning through the development of applications such as Duolingo and Babbel. Duolingo adapts the lessons depending on the learner’s performance, which means that the tasks given to the students are always challenging, but not beyond the learners’ capabilities. The use of an AI-driven system makes it easier to track the mistakes that the student makes and adjust the subsequent lessons to correct the mistakes. Likewise, Babbel incorporates AI to assist the student in monitoring their progress in learning the vocabulary and grammar in the course (Kessler et al., 2023). These language learning tools help in making it easier for the students, especially those in primary and middle schools, to learn languages by making it a unique experience.
Another aspect where the AI tools such as Miro and Padlet have brought a change in education is collaboration. Miro is an AI-based collaboration tool that allows students to form groups, discuss concepts, and participate in visual planning. This tool promotes effective collaboration in real-time as students can create a board that they can all work on, add notes, and build projects together (Karsen et al., 2022). Padlet, on the other hand, supports collaborative learning by allowing students to post and display their ideas, projects and notes in an attractive digital canvas. It also has AI components that assist in managing and categorizing the content to enhance the students’ teamwork. Such platforms encourage engagement and collaboration, both important competencies in today’s work settings.
In science education, tools such as Labster and Phet Interactive Simulations offer students virtual labs and interactive experiments, allowing them to delve into complex scientific concepts without needing physical lab equipment. Labster features AI-driven simulations that enable students to perform biology, chemistry, and physics experiments in a virtual environment, allowing them to test hypotheses and investigate scientific principles safely (Tsirulnikov et al., 2023). Similarly, Phet Interactive Simulations offers AI-powered simulations across various scientific disciplines, giving students hands-on experience with experiments that may be otherwise out of reach due to limited resources. These tools are especially valuable for students who lack regular access to fully equipped labs, helping them engage more deeply with scientific inquiry.
To encourage creativity, platforms like Canva and Animoto enable students to showcase their ideas through visual design and video production. Canva is a design tool powered by AI that assists students in crafting visually striking projects, including posters, presentations, and infographics, by offering customizable templates and design suggestions generated by AI (Belda-Medina & Goddard, 2024). Animoto is a video creation tool that leverages AI to simplify the video-making process. It provides ready-made templates and enhanced effects that make it easy for even younger students to create videos.
AI also assists students in staying organized through tools such as Todoist and MyStudyLife. Todoist is a task management app that utilizes AI to help students monitor their assignments, exams, and activities by sending reminders and allowing for customizable to-do lists. Similarly, MyStudyLife enables students to oversee their school schedules, deadlines, and exam dates, aiding them in managing their time more effectively. These tools are particularly beneficial for students who find time management challenging, providing AI-driven solutions to help them keep up with their academic obligations (Bouchrika, 2024).
Mental health plays a vital role in the well-being of students, and AI tools such as Woebot and Moodpath offer valuable emotional support for those facing anxiety, stress, or other emotional difficulties. Woebot is an AI-powered chatbot designed to assist students in managing their mental health by providing real-time support and coping strategies (Fitzpatrick et al., 2017). Moodpath monitors students' emotional states and delivers insights and resources to help them better comprehend and handle their feelings. These AI tools introduce a new dimension of emotional support, especially for students who may lack immediate access to mental health professionals.
AI applications in education provide numerous advantages for students, including personalized learning experiences, improved problem-solving tools, collaborative platforms, and emotional support. It is believed that with the improvement of artificial intelligence technology, its applications in education will also become more and more powerful, creating opportunities for students to study in progressively interactive, personalized, and supportive environments.
Methodology
The aim of this paper is to proposeexplore thresholdthe conceptsopinions and attitudes of prospective teachers from the Faculty of Teacher Education at the University of Zagreb regarding the use of Artificial Intelligence (AI) in education. The study aims to find out the students’ perceptions of AI, positive and negative attitudes towards AI, the courses they think will benefit most from incorporating AI, and the specific AI applications that are most useful for computerteachers scienceand taughtstudents. This quantitative study is based on an online questionnaire and aims to identify the perceptions of prospective primary school teachers on the opportunities and limitations of AI in primaryteaching school.practices, as well as their preferences on the use of AI tools as learners and teachers. The basicpaper methodologyformulates ofthree the research was the nominal group technique (NGT), which is suitable for gaining a deeper insight into the participants' perception and understanding of threshold concepts. The nominal group technique (NGT) was found to be a very effective method for promoting critical thinking through discussions that involve a small number of participants, provide clear and focused instructions and allow for constructive feedback. NGT fulfils all these criteria while ensuring the full participation of all group members, which is especially useful in an educational context.
NGT is widely used in various disciplines, e.g. medicine, information technology, politics, management and education, where it serves as a method for evaluated discussions (Macphail, 2001). In education, NGT is used in the design and evaluation of curricula and as a pedagogical method that encourages active participation (Chapple & Murphy, 1996). Studies have shown that NGT increases participant’s productivity and problem-solving skills through structured discussions (Madar, 1982).
hypotheses:
ProcedureHypothesis 1: There are no statistically significant differences in opinions and attitudes toward using AI in teaching based on the students' year of study.
TheHypothesis application2: processThere are no statistically significant differences in opinions and attitudes toward using AI in teaching based on the students' study module.
Hypothesis 3: There are no statistically significant differences between the average rating of NGTpositive consistedand the average rating of sixnegative steps, beginning with an oral presentation that covered the definitions and examples provided. A presentation was prepared, which included precise definitionsaspects of threshold concepts, along with examples of threshold concepts with explanations according to their characteristics. For demonstrating threshold concepts, „looping” and subprograms were selected as examples that fully align with the threshold concept. Simultaneously, as an example of a demanding concept, recursion was analyzed and identified as a concept that requires additional understanding but is not a necessary prerequisite for further learningAI in programming.teaching.
The secondsurvey step,included 56 students from the University of Zagreb, Faculty of Teacher Education, including all five years of academic studies and different modules: Croatian Language, Art Culture, Informatics, Educational Sciences, and English and German Language programs. The questionnaire was voluntary and anonymous and consisted of three sections. The first part was used to collect data on the study year and the module or program students were studying.
silentFollowing ideathe generation,previous requiredsection, the participants were asked to individuallyanswer reflectsome onquestions regarding their opinions of artificial intelligence in the educational sector. Participants were given statements and writewere downthen asked to indicate their thoughtslevel regardingof thresholdagreement concepts.or Theydisagreement completedwith the presented statement. The instrument utilized a questionnaireLikert-type thatscale requiredbetween them1 and 5, with each number corresponding to try to identify concepts and answer whether these concepts are fundamental, demanding, and whether they meet the characteristicsfollowing levels of thresholdconsensus:
1 - completely disagree (lowest level of agreement), 2 - mostly disagree (partial disagreement), 3 - neutral (neither agree nor disagree), 4 - mostly agree (partial agreement) 5 - completely agree (highest level of agreement).
The third step, idea discussion, took place after individual reflections. Participants were divided into smaller groups with 4 to 7 participants to further deepen their understanding and discuss the proposed threshold concepts. Through constructive discussion, participants contributed to a deeper understandingpart of the conceptsquestionnaire consisted of questions about the possible application of AI technologies in learning. Students were asked which subjects they think would be most helped by AI, which AI tools are helpful for teachers, and theirwhich applicationAI tools are helpful for students in computer scienceprimary education. In this group decision-making phase, each group selected one or more concepts thatFurther, they believedwere metasked allhow often they would incorporate AI tools in their teaching in the characteristicsfuture. of threshold concepts.
To obtain clear results and rankAt the proposed threshold concepts, voting was conducted via the online platform Padlet. After the voting and reviewingend of the results,questionnaire, conclusionsthey responded to an open-ended question: „What changes do you think AI will bring to primary education in the next 10 years?“ The answers were drawn,coded and potentialgiven thresholdcategories, conceptswhich werewill identified.be Lastly,discussed in the reportnext with a summarysection of the procedure, decisions,paper, and finalthe resultsstatistically wassignificant written.
were
analyzedParticipants
Thethe participantshelp of the researchappropriate werestatistical primary school teachers of computer science (N=53) who participated in the activities organized by the Professional Council of Computer Science Teachers of Primorje-Gorski Kotar County at Vežica Primary School in Rijeka in February 2024.
In addition to the questions on threshold concepts, participants were asked to provide demographic data, including their educational profile, professional experience and the subjects they teach. Table 1 shows the age and gender of the participants. The participants in the survey have a wide age range, from 24 to 63 years, with an average age of 39 years. This age diversity suggests that participants are at different stages in their careers, which could influence their responses given their varied teaching experience.tests.
TableResults 1
Age of the participants
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
discussion
The majoritystudy included 56 students from the Faculty of participantsTeacher wereEducation, women,University aof smaller number were men (Chart 1). One respondent did not specify their gender.Zagreb.
In Chart 1, students' responses regarding their year of study are displayed. The distribution is as follows: 12% of students are in their first year of study, 29% in second year, 36% in third year, 9% in fourth year and 14% in fifth year.
Chart 1
Distribution of participantsstudents by genderyear of study
When it comes to their study module or study program, Chart 2 shows the distribution of students in each module: 21% Croatian Language module, 12% Art Culture module, 32% Informatics module, 11% Educational Sciences module, 20% English Language program, 4% German Language program.
AllChart participants2
Distriburion of studenty by study module/program
In the questionnaire, after students provided their year of study and study module/program, they were asked to respond to whether, during their studies so far, they had engaged with topics related to artificial intelligence (AI) or educational technology in any course. 64.3% of the students asnwered yes and 35.7% answered no.
The following question was: „How would you rate your familiarity with artificial intelligence technologies?“ Students answered on a scale from 1 to 5, where 1 indicated „not familiar at all“ and 5 indicated „very familiar.“ The answers are displayed in Chart 3, where it can be observed that most students answered 3 (somewhat familiar) 39.3% and 4 (familiar) 35.7%. Only one student answered nor familiar at all and only five students rated their familiarty as very familiar.
Chart 3
Familiarity with atificial intelligence techologies
To determine the method for analyzing the collected research results, the Kolmogorov-Smirnov test was conducted. The test indicated that the distributions of the measured dependent variables significantly deviate from the normal Gaussian distribution. Given this result and the sample size, non-parametric statistical indicators and results from non-parametric statistical tests will be taken into consideration.
The data shown in Table 1 provides an analysis of student perceptions regarding the positive aspects of AI in education. The results show that students have a somewhat favorable view of AI's role in enhancing educational experiences. For instance, AI's potential to improve access to education for students with special needs received one of the highest average scores (M = 3.53), indicating that students see value in AI's ability to create more inclusive learning environments. Similarly, the potential of AI to facilitate lesson planning and organization for teachers was rated relatively highly (M = 3.54), suggesting that students recognize the practical benefits AI can offer to teachers.
However, the perceived ability of AI to allow teachers to focus more on individualized support received a lower average score (M = 2.89), highlighting some skepticism about AI's capacity to enable more personalized teacher-student interactions. Despite this, the overall positive mean score of 3.20 for AI-related aspects suggests that students generally view AI as a useful tool in enhancing both learning and teaching processes in primary education.
Table 1
Students’ perceptions regarding the positive aspects of AI
|
Positive aspects |
Mean |
Median |
Mode |
Std. Deviation |
|
|
AI can be useful in primary education. |
3.13 |
3 |
3 |
1.04 |
|
|
AI can help personalize educational experiences for students. |
3.16 |
3 |
4 |
1.09 |
|
|
AI can improve learning efficiency by providing students with immediate feedback. |
3.21 |
3 |
3 |
1.17 |
|
|
AI technologies can increase student engagement through interactive learning methods. |
3.05 |
3 |
4 |
1,22 |
|
|
AI can improve access to education for students with special needs. |
3.53 |
4 |
4 |
1.20 |
|
|
AI can allow teachers to focus more on providing individualized support to students. |
2.89 |
3 |
3 |
1.12 |
|
|
AI can facilitate lesson planning and organization for teachers. |
3.54 |
4 |
4 |
1.15 |
|
|
Average rating of positive AI aspects. |
3.20 |
3.42 |
3.57a |
.94 |
The data shown in Table 2 suggests that students perceive several significant negative aspects of AI in education, with the highest concern being that AI may reduce students' critical thinking by providing ready-made solutions (M = 4.59). This indicates a strong belief that relying on AI could prevent independent problem-solving and creativity in the learning process. The concern that AI technologies could lead to an over-reliance on technology instead of creative learning methods (M = 4.41) reflects a strong belief among respondents that there is a potential downside to integrating AI into education. This high mean score indicates that many students fear that AI might shift the focus from fostering creativity to overusing technology. Additionally, many students expressed concerns about the ethical implications of AI, such as privacy violations and algorithmic bias (M = 4.14), highlighting the importance of addressing these issues in educational AI deployment. Students are also concerned that AI may not yet be advanced enough to fully understand individual student needs (M = 4.18), reflecting skepticism about AI's ability to provide the same level of personalized support that teachers can. Interestingly, the complexity of AI technology and its potential to complicate teachers' work received a lower score (M = 3.04), suggesting that while students see AI as a challenge, they are more worried about its effects on teaching quality and creativity than its difficulty. Overall, the average rating of negative attitudes towards AI in education (M = 3.97), reflects a relatively high level of educationconcern andamong workstudents. asThis computersuggests sciencethat, teachersoverall, atstudents primaryare schoolsmore (Tableapprehensive 2).than Theoptimistic averageabout lengththe potential drawbacks of time spentAI in theeducational teachingsettings. profession is around 10 years. A total of 33 participants have between 0 and 9 years of professional experience, 7 participants have between 10 and 19 years of professional experience, 11 participants have between 20 and 29 years of professional experience, and 2 participants have more than 30 years of professional experience.
Table 2
Participants‘Students’ workperceptions experienceregarding the negative aspects of AI
|
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
| |
|
|
|
|
| |
|
2 |
|
|||
|
The use of AI in education may diminish the importance of the teacher's role. |
|
|
5 |
1.43 |
|
|
AI technologies could lead to an over-reliance on technology instead of creative learning methods. |
4.41 |
5 |
5 |
.93 |
|
|
AI is not yet advanced enough to fully understand students' individual needs. |
4.18 |
4 |
5 |
.93 |
|
|
AI may reduce students' critical thinking by providing ready-made solutions. |
4.59 |
5 |
5 |
.82 |
|
|
I am concerned about the ethical implications of using AI in education, such as student privacy violations, algorithmic bias, and reducing the human element in teaching. |
4.14 |
5 |
5 |
1.15 |
|
|
Average rating of negative aspects of AI. |
3.96 |
4.08 |
4.67 |
.77 |
This strong concern highlights the necessity for careful and responsible implementation of AI technologies, ensuring that they complement, rather than replace, traditional pedagogical methods. It also suggests a demand for more transparency and discussion around how AI will impact teaching and learning, student engagement, and creativity.
The Kruskal-Wallis test was conducted to determine whether there are differences in opinions on the use of AI in teaching based on the year of study. The test showed a statistically significant difference regarding the year of study for two statements (Table 3). However, when comparing the arithmetic mean ranks for each year, it is not possible to conclude that students change their opinion consistently based on their year of study (for example, second year students are less likely to believe that AI can help students with special needs, but fifth year students believe this even less than second year students). While there is a statistically significant difference between certain groups of respondents depending on their year of study, the results are irregular across categories. Therefore, Hypothesis 1 is partially confirmed.
Table 3
Differences in attitudes toward the use of AI in teaching based on the year of study
|
Differences in attitudes toward the use of AI in teaching based on the year of study. |
Chi-Square |
df |
Asymp. Sig. |
|
Perceived familiarity with AI technologies. |
5.629 |
4 |
.229 |
|
AI can be useful in primary education. |
3.753 |
4 |
.440 |
|
AI can help personalize educational experiences for students. |
3.806 |
4 |
.433 |
|
AI can improve learning efficiency by providing students with immediate feedback. |
1.843 |
4 |
.765 |
|
AI technologies can increase student engagement through interactive learning methods. |
5.712 |
4 |
.222 |
|
AI can improve access to education for students with special needs. |
11.444 |
4 |
.022 |
|
AI can allow teachers to focus more on providing individualized support to students. |
3.446 |
4 |
.486 |
|
AI can facilitate teachers' planning and organization of lessons. |
9.723 |
4 |
.045 |
|
AI technologies in education may be too complex to use and could complicate teachers' everyday work. |
2.031 |
4 |
.730 |
|
The use of AI in education may diminish the importance of the teacher's role. |
8.100 |
4 |
.088 |
|
AI technologies could lead to an over-reliance on technology instead of creative learning methods. |
6.157 |
4 |
.188 |
|
AI is not yet advanced enough to fully understand students' individual needs. |
8.637 |
4 |
.071 |
|
AI may reduce students' critical thinking by providing ready-made solutions. |
3.872 |
4 |
.424 |
|
I am concerned about the ethical implications of using AI in education, such as student privacy violations, algorithmic bias, and reducing the human element in teaching. |
8.690 |
4 |
.069 |
|
Average rating of positive AI aspects |
5.332 |
4 |
.255 |
|
Average rating of negative aspects of AI |
8.776 |
4 |
.067 |
ParticipantsTo varydetermine whether there are differences in opinions on the use of AI in teaching based on the study module, the Kruskal-Wallis test was conducted for each statement and for the average rating of positive aspects. The test identified statistically significant differences based on the module for three statements shown in the combinationTable 4, and for the average rating of subjects they teach, including computer science, mathematics, English, physics and technical education what can be seen in Table 3.
Table 3
Number of participants by job position
|
|
|
|
|
|
|
|
|
|
Results
The results collected through the survey provide insight into opinions of computer science teachers regarding the threshold concepts.
Within the stage of silent idea generation, the participants highlighted several key fundamental concepts in computer science education, including „Branching”, „Functions”, „HTML”, „File Storage”, „Addressing” and „Logical Conditions”, which were recognized as essential for understanding computer science. „Branching Algorithm” was marked as a fundamental concept by 15 participants, while „Functions” and „Logical Conditions” received 13 votes. „HTML” and „File Storage” were also identified as fundamental by 11 participants, and „Addressing” was also ranked highly.
Challenging concepts that represent learning obstacles include „Branching”, „Database Relations” and „File Storage”, which require additional attention and resources for adequate understanding.
The interconnection of concepts was also important, and the participants recognized that „Branching”, „Functions”, „Logical Conditions”, „HTML” and „File Storage” are connected, meaning that mastering these concepts facilitates understanding of otherpositive aspects of computer science.
The transformative nature of concepts, which indicates their potential to lead to deeper understanding, was also highlighted for „Branching”, „Functions”, „HTML” and „Logical Conditions”, while the irreversibility of these concepts was considered high, meaning that students rarely forget these terms once they have mastered them.
Concepts such as „Database Relations” and „Branching“ were marked as bounded, meaning that their understanding may vary among students, requiring continuous repetition.
Some concepts, such as „Attributes”, „Database”, „File”, „Elements in Canva”, „Internet”, „Output Devices”, „Copying” and „References in Word”, were not recognized as fundamental or challenging by most participants, indicating their lower importance in basic computer science education.
During the idea discussion phase, the participants concluded that some concepts, such as „Branching” and „Logical Conditions”, could be combined because they are interconnected, and that „Data Organization” and „File Storage” could also be merged, as data organization is key to understanding the process of file storage. After the discussion phase, we move on to the voting phase.
In order to obtain clear results and ranking of the proposed threshold concepts, a voting was conducted through the online platform Padlet. In the following table, the proposed threshold concepts and the voting results for each proposed concept can be seen. Because some teachers were undecided on certain concepts, the total number of votes differs per concept and is less than 53.AI.
Table 4
ProposedDifferences thresholdin conceptsattitudes toward the use of AI in teaching based on the study module
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
Educational Sciences |
|
||
|
Informatics |
|
||
|
AI can improve learning efficiency by providing students with immediate feedback. |
English Language |
25.41 |
|
|
German Language |
38.25 |
||
|
Croatian Language |
24.79 |
||
|
Art Culture |
14.50 |
||
|
Educational Sciences |
39.17 |
||
|
Informatics |
33.67 |
||
|
AI can improve access to education for students with special needs. |
English Language |
22.05 |
|
|
German Language |
33.25 |
||
|
Croatian Language |
22.08 |
||
|
Art Culture |
14.00 |
||
|
Educational Sciences |
36.08 |
||
|
Informatics |
38.32 |
||
|
Average rating of positive AI aspects. |
English Language |
24.55 |
|
|
German Language |
28.00 |
||
|
Croatian Language |
22.00 |
||
|
Art Culture |
15.93 |
||
|
Educational Sciences |
36.58 |
||
|
Informatics |
36.41 |
When observing the arithmetic means of ranks, they indicate that students from the Informatics module generally express more positive opinions on three statements compared to other modules. For the average rating of positive aspects, a post-hoc analysis was conducted to determine between which modules the differences were statistically significant. Post-hoc tests revealed that students from the Informatics module express significantly more positive attitudes toward the use of AI in teaching compared to students from the English Language, Croatian Langugage, and Art Culture modules. However, no significant difference was found between students from the Informatics module and those from the German Language or educational sciences modules. Therefore, it cannot be concluded that Informatics module students have more positive attitudes toward AI compared to other modules and Hypothesis 2 is partially confirmed.
The Wilcoxon Signed Ranks test was conducted to determine whether there is a statistically significant difference between the average ratings of positive and negative aspects of AI in teaching. The result indicated that there is a statistically significant difference (Table 5).
Table 5
Differences in attitudes toward the use of AI in teaching based on the study module
|
Wilcoxon Signed Ranks Test |
Average rating of negative aspects of AI Average rating of positive aspects of AI |
|
Z |
-3.846b |
|
Asymp. Sig. (2-tailed) |
.000 |
By analyzingcomparing the votingmedians, results,it can be concluded that there is statistically significantly greater agreement with the negative aspects than with the positive aspects of using AI in teaching (Chart 4) so here it can be concluded that the Hypothesis 3 is refuted.
Chart 4
Average rating of positive and negative aspects of using AI in teaching
In the next section of the questionnaire, students were presented with a list of applications that could be useful for teachers, and they were asked to select which ones they considered the most useful using a multiple-choice format. They were also able to write applications that they consider useful. The applications were as follows:
Intelligent Tutoring Systems: Carnegie Learning's MATHia – This system uses artificial intelligence to adapt math lessons to students based on their responses, providing personalized instructions and explanations.
Automated Grading and Feedback Systems: Grammarly – An AI tool that automatically reviews and evaluates written text's grammatical and stylistic aspects, offering feedback to help students improve their writing assignments. Turnitin – A platform that uses AI to analyze student papers, detect plagiarism, and provide feedback on the originality of the content.
Adaptive Learning Platforms: Khan Academy – An online educational platform that uses AI to personalize lessons and track student progress, adjusting the material according to their needs. DreamBox – An adaptive math learning platform that uses AI to tailor instructional content in real time based on student responses and progress.
AI for Data Analysis and Student Assessment: Civitas Learning – A platform that uses AI to analyze large amounts of student data to predict their success and provide interventions to improve outcomes. Edmentum – A system that leverages data analytics and AI to assess student progress and generate reports for teachers to understand student needs better.
AI Tools for Classroom Organization and Management: ClassDojo – A platform incorporating AI to help teachers manage student behavior and engagement. Teachers can track student progress, communicate with parents, and customize student activities.
AI Tools for Supporting Students with Special Needs: Otter.ai – An AI-powered transcription tool that assists students with hearing or writing difficulties by providing real-time transcription of lectures. ModMath – An AI application designed for students with dyslexia or motor disabilities, allowing them to solve math problems on a digital platform without the need for handwriting.
As shown in Chart 5, most students (N=30) see the potential of AI tools for supporting students with special needs which is closely followed by Automated grading and feedback systems (N = 26). Two students answered none.
Chart 5
Usefulness of AI application designed to support teachers
When it comes to applications designed to support students, the list was as follows.
AI-driven educational games: like Minecraft: Education Edition, use AI to create dynamic educational scenarios where students can learn programming, math, science, and other subjects through interactive gameplay. DragonBox is an educational game that utilizes AI to tailor math challenges to the student’s knowledge level, especially helping with learning basic mathematical concepts.
AI tools for learning and problem-solving: Socratic by Google, an AI app that answers students' questions and provides problem solutions in various subjects with tailored explanations. Photomath is an AI tool that allows students to scan math problems and receive step-by-step solutions and explanations to better understand the material.
Language learning: Duolingo is an AI tool that personalizes lessons according to the student’s progress, adjusting the difficulty of thresholdtasks for more effective language learning. Babbel is another AI language-learning app that adapts lessons to the student’s needs and tracks progress in vocabulary and grammar.
AI tools for collaboration and teamwork: Miro provides an AI-powered platform that enables students to work in teams, brainstorm together, and visually plan projects. Padlet is an online tool using AI to facilitate collaboration, allowing students to share ideas, work, and projects in one place.
AI tools for scientific experiments and simulations: Labster, offer virtual labs and scientific experiment simulations, enabling students to explore biology, chemistry, physics, and other scientific fields. Phet Interactive Simulations uses AI-driven simulations that allow students to explore various scientific concepts wasand proposedexperiments basedin onan inputinteractive fromenvironment.
Creativity highlightingand thosedesign: Canva is an AI design tool that clearlyhelps receivedstudents highercreate visually appealing projects, such as posters, presentations, and infographics, with customizable templates and elements. Animoto is an AI-powered video creation tool that allows students to make video content using templates and AI-driven effects easily.
Creative writing development: Storybird is an AI platform that guides students through creating and writing their own stories, providing ideas and visual support.
Research and information gathering: Kiddle is an AI-powered search engine tailored for children, ensuring safe and relevant results for younger students when searching for topics for their projects. Elicit is an AI tool that assists students in researching various topics by suggesting relevant articles and sources and providing summaries.
AI tools for virtual and augmented reality: Google Expeditions, an AI-powered AR and VR platform that allows students to take virtual trips and explore historical, scientific, and cultural locations.
Time management and organization: Todoist is an AI tool that helps students organize their tasks and activities by providing reminders and customizable to-do lists. MyStudyLife is an AI app that helps students track school assignments, deadlines, and exam dates, assisting them in better managing their time.
Mental health and emotional support: Woebot is an AI chatbot that provides emotional support asto thresholdstudents, conceptshelping them manage anxiety, stress, and emotional challenges. Moodpath is an AI app that monitors students' emotional states and provides tools to better understand and manage their emotions.
As shown in computerChart science6, education.most students (N=29) see the potential of AI tools used for language learning and scientific experiments and simulations (N=29). This is closely followed by AI tools for Creativity and design (N=28). On the low end of the score, students don’t really see the benefits of using AI tools for research and information (N=12) and creative writing (N=6), also, one student answered „None.“
Logical conditions”
Chart an6
Usefulness of votes,applications designed to support students
Regarding the subjects they think should be supported by AI (Chart 7) the subject Informatics was mentioned most often (N=40).
Chart 7
Usefulness of applications in various subjects
Here is a categorization of the provided text into a few key categories:
1. Positive Impacts of AI on Education:
Ease of Use for Teachers: AI will likely facilitate the preparation of materials and assignments, reducing the workload for teachers. Teachers will be able to track and evaluate student progress more easily and make their classes more engaging.
Enhanced Learning Tools: More tools will make learning more interesting both in class and independently. AI will provide personalized learning, adapting to each student's pace, interests, and needs, helping students with 42different votesabilities to progress better.
„YES”Administrative Automation: AI will automate administrative tasks like grading and onlyprogress onetracking, votereducing the burden on teachers.
„NO”.Interactive „CellLearning addressingfor (Abstract Concepts: AI will help explain complex topics visually, especially in spreadsheet)”subjects alsolike receivednatural asciences.
Increased numberEfficiency: ofAI votes,can withmake 40teaching votesmore „YES”efficient and onlylessen twothe votesburden „NO”.on „Flowchart”teachers, receivedmaking 35classes votesmore „YES”effective and only 6 votes „NO.” „Data organization(storage)” received 39 votes „YES” and 6 votes „NO”. „Variables”, although not receiving unanimous support, can be considered a threshold concept given the predominant support.fun.
Discussion2. Negative Impacts of AI on Education:
Over-reliance on AI: Both teachers and students may become too dependent on AI, leading to a reduction in independence and creativity. Over-reliance on AI might diminish critical thinking skills and logical reasoning among students.
Decreased Teacher Involvement: The resultsrole of teachers might be diminished, with AI tools handling tasks that were previously teacher-led, potentially leading to a decrease in the importance of the surveyteacher-student provide several important insights into primary school IT teachers’ views on threshold concepts. Concepts such as „logical conditions”, „cell addressing”, „variables”, and „data organization” received strong support, indicating their importance for computer science teaching. The high consensus on „logical conditions” (42 out of 43 participants who voted for this concept) may reflect its fundamental role in programming, particularly in helping students understand decision-making in algorithms. This strong support suggests that teachers view it as a prerequisite for grasping more advanced programming concepts.relationship.
In contrast, the slightly lower support for the concept of „flowchart” (35 votes) may indicate variability in how teachers integrate visual algorithm representation into their teaching practices. While flowcharts are valuable for structuringLaziness and planningReduced code,Critical someThinking: teachersStudents maymight rely on alternative methods, such as pseudocode,AI to introducecomplete theseassignments, skills.leading to a lack of effort and reduced engagement in learning. AI might encourage a „shortcut“ mentality, where students seek easy solutions rather than learning and problem-solving.
Reduced Socialization and Communication: The strong support for „cell addressing” (40 votes) highlights its relevance not only for spreadsheet data management but also for broader applications like data analysis. This result underscores the increasing emphasis on data literacy in primary education, aligning with the growing importanceuse of dataAI sciencemay skills. However, it also raises questions about whether teachers feel adequately equippedlead to teachless thesesocial interaction and communication among students, affecting their social skills effectively.
These findings suggest that while core programming concepts such as „logical conditions” and „variables” are well-established in the curriculum, there is roomability to explore how concepts like „data organization” and „flowcharts” can be better integrated. Future research might investigate how these concepts are taught in practice and the challenges teachers face in developing student’s understanding of them.focus.
Conclusions3. Mixed Impacts and Concerns:
TheBalance topicbetween ofBenefits thresholdand conceptsRisks: representsWhile notAI onlycan aprovide challenge,valuable but also an opportunitytools for innovationlearning, especially in computerareas where schools lack resources (e.g., science teaching.labs), Clearlythere definedis thresholdconcern conceptsabout providemaintaining aemotional foundationand human interaction in education, which is crucial for high-qualitychild education and promote the development of the skills required in the digital age.development.
Methodologies like the nominal group technique allow structured dialogue among teachers, creating a spaceNeed for exchangingGradual experiencesImplementation: Technology should be introduced gradually to avoid negative impacts on critical thinking and findingensure commonthat solutionsAI toenhances teachingrather challenges.than Basedreplaces onhuman the research conducted using the nominal group technique with primary school teachers the following threshold conceptselements in computer science education have been proposed: logical conditions, cell addressing (in spreadsheet), flowcharts, data organization, and variables. These concepts have been highlighted as transformative points that enable deeper understanding of the subject matter and progress for students.
In the future work, these proposals will be validated through further research involving subject-matter experts, who will contribute to developing a solid rationale confirming that these concepts have the defining characteristics of threshold concepts. The integration of educational games has considerable potential to support students in mastering complex topics while increasing their engagement and motivation in the learning process. Following this validation, game-based learning activities will therefore be designed to facilitate student’s acquisition of these concepts.education.
Acknowledgment4. Skepticism and Caution:
Skepticism toward AI Integration: Many fear that the integration of AI will lead to a loss of key educational values, such as creativity, problem-solving, and personal development. Some believe AI will lead to laziness and a lack of effort in both teaching and learning processes.
Potential for Negative Long-term Effects: Some express concern that the long-term impact of AI will be a decrease in socialization and hands-on learning, with children becoming overly reliant on technology for solutions.
Conclusion
In conclusion, this study examined the attitudes of prospective primary school teachers towards the use of AI in education, focusing on three hypotheses. The researchfirst hypothesis proposed that there are no statistically significant differences in opinions and attitudes based on the students' year of study. The results partially supported this hypothesis, as some differences were found across study years, particularly regarding the perception of AI’s role in supporting students with special needs and facilitating lesson planning. Therefore, it was concluded that the year of study has beenonly fundeda byminor influence on shaping attitudes and opinions, as the Erasmus+observed Programmedifferences were not consistent across all evaluated categories.
The second hypothesis suggested that there are no statistically significant differences in opinions based on the students' study module. This hypothesis was also partially confirmed, as students from the Informatics module exhibited significantly more positive attitudes towards AI compared to those from the language and arts modules. Nevertheless, no significant differences were found between Informatics and some other modules, such as Educational Sciences, highlighting a subtle connection between the study module and attitudes toward AI.
The third hypothesis, which stated that there are no statistically significant differences between the average ratings of thepositive Europeanand Union,negative KA220-SCHaspects -of Cooperation partnershipsAI in schoolteaching, education,was underrefuted. The results indicated a statistically significant difference, with students rating the projectnegative “aspects of AI—such as concerns about critical thinking, creativity, and ethical implications—more strongly than the positive aspects, such as AI’s potential to improve learning efficiency and support personalized education.
Overall, what these findings show is the complexity of integrating AI into education. Prospective teachers not only see the potential benefits of using AI with students but also are very cautious about the possible drawbacks. They worry about the potential impact of creative AI on the kind of work that students do and on the kinds of decisions that teachers make about the kind of work that students do. And they especially worry about the kind of impact that creative AI could have on student autonomy.
Literature
Harry, A., & Sayudin, S. (2023). Role of AI in Education. Interdiciplinary Journal and Hummanity (INJURITY), 2(3), 260-268.
Alneyadi, Saif, Wardat, Yousef, Alshannag, Qasim, & Abu-Al-Aish, Ahmad. (2023). The effect of using smart e-learning app on the academic achievement of eighth-grade students. EURASIA Journal of Mathematics, Science&Math educationaland gamesTechnology Education, 19(4), em 2248.
HolonIQ. (2023). Artificial intelligence in education: 2023 survey insights. HolonIQ. https://www.holoniq.com
Sutaria, N. (2022). Bias and ethical concerns in machine learning. ISACA J., 4, 1-4.
McNulty, N. (2023). AI data privacy in schools: Navigating the challenges and solutions. Retrieved from preschoolhttps://www.niallmcnulty.com
GUJJULA, university”R., & Sanghera, K. (023-2023). 1-HR01-KA220-SCH-000165485)Ethical Considerations and Data Privacy in AI Education. Journal of Student-Scientists' Research, 5.
UNESCO. (2023). The challenges and opportunities of artificial intelligence in education. UNESCO. https://www.unesco.org
Literature
R.
S.,Bognar,& B.Hawn, A. (2016)2022). TeorijskaAlgorithmic polazištabias e-učenja.in Croatianeducation. International Journal of Education:Artificial HrvatskiIntelligence časopisin za odgoj i obrazovanje, 18(1)Education, 225-256.32, Available online: https://hrcak.srce.hr/file/229424, accessed 28.12.2023.1052–1092.
Boustedt,Chaudhuri, J.S., Eckerdal,& A.Mohanty, I. (2023). The importance of bias mitigation in AI: Strategies for fair, ethical AI systems. UXmatters.
ICO. (2023). What about fairness, bias and discrimination? ICO.
Ibrahim, Hamza Khalifa, Al-Awkally, Noor Alhooda Milood, Samad, Abdul, Zaib, Waqar, & Hamza, Muhammad. (2022). Covid-19 Pandemic and Its Impact on Psychological Distress, Malignancy and Chronic Diseases: A Scoping Review. Eduvest-Journal Of Universal Studies, McCartney,2(5), 1017–1021.
Gningue, S. M., Peach, R., Moström,Jarrah, J. E., Ratcliffe,A. M., Sanders,& Wardat, Y. (2022). The Relationship between Teacher Leadership and School Climate: Findings from a Teacher-Leadership Project. Educ. Sci. 2022, 12, 749.
Alarabi, K., & Zander,Wardat, C.Y. (2007)2021). ThresholdUAE-based conceptsteachers’ inhindsight computerjudgments science:on dophysics theyeducation existduring the COVID-19 pandemic. Psychology and areEducation they useful? ACM Sigcse Bulletin, 39(1)Journal, 504-508.58(3), 2497–2511.
Breidenbach, D., Dubinsky, E., Hawks, J., & Nichols, D.Calibraint. (1992)2023). Development of the process conception of function. Educational studiesAI in mathematics,Education: 23(3),A 247-285.Comprehensive Guide to Benefits and Applications.
CELatElon,Niall CenterMcNulty. (2023). AI predictive analytics for Engagededucation: Learning.Transforming (2013,learning September 18)outcomes. DecodingRetrieved thefrom disciplines and threshold concepts. Accessed December 10, 2023, Available online: https://youtu.be/Wqe_kKFoOq4.www.niallmcnulty.com
Falkner,Calibraint. N.(2023). AI in education: A comprehensive guide to benefits and applications. Retrieved from https://www.calibraint.com
Benzakour, A., Altsitzioglou, P., Lemée, J. G.M., Vivian,Ahmad, R.A., J.Mavrogenis, A. F., & Falkner,Benzakour, K.T. (2022). Artificial intelligence in spine surgery. International Orthopaedics.
Kurilovas, E. (2013)2018). ComputerAdvanced Sciencemachine learning approaches to personalize learning: Learning analytics and decision making. Behavior and Information Technology.
Chiu, T. K. F., Xia, Q., Zhou, X., Chai, C. S., & Cheng, M. (2023). Systematic literature review on opportunities, challenges, and future research recommendations of artificial intelligence in education. Computers and Education: Artificial Intelligence.
Chaudhuri, S., & Mohanty, I. (2023). The Firstimportance Thresholdof Concept.bias 2013mitigation in AI: Strategies for fair, ethical AI systems. UXmatters.
Malik, A. (2024). Addressing bias in AI: Strategies for bias mitigation with OpenAI. Signity Solutions.
AIWS. (2023). Revolutionizing learning: The advantages of AI intelligent tutoring systems. AI World School. https://aiworldschool.com
Fernández-Herrero, J. (2024). Evaluating recent advances in affective intelligent tutoring systems: A scoping review of educational impacts and future prospects. Education Sciences, 14(8), 839. https://doi.org/10.3390/educsci14080839
Kurni, M., Mohammed, M. S., & Srinivasa, K. G. (2023). Intelligent tutoring systems. In A beginner's guide to introduce artificial intelligence in teaching and learning (pp. 29-44). Cham: Springer International Publishing.
Alharbi, W. (2024). Mind the gap, please! Addressing the mismatch between teacher awareness and student AI adoption in higher education. International Journal of Computer-Assisted Language Learning and Teaching in Computing and Engineering. Macau: Macao, pp. 39-46. Available online: https://doi.org/10.1109/LaTiCE.2013.32.(Mind-the-Gap-Please!_-A…).
GruićBeović, I.E. (2005)2023). EnteringTheoretical theApproaches storyto labyrinth.Teacher AnTraining Investigationin Primary Education. University of Zadar.
Felix, C. V. (2020). The Role of the EffectsTeacher ofand Placing the Participants in Different Actantial Positions within the Dramatic World Created by a TheatreAI in Education Programme. [DoctoralInternational disertationPerspectives oron Master’sthe thesis]. Birmingham: FacultyRole of Technology in Humanizing Higher Education (Chapter_3_The_Role_of_t…).
Guilherme, A. (2017). AI and education: The importance of teacher and student relations. AI & Society.
Kim, J. (2023). Leading teachers' perspective on teacher-AI collaboration in education. Education and Information Technologies (Kim2023.Leadingteachers…).
Pejnović, L. (2024). The Use of AI in the Education System and Its Potential Consequences. University of Central England in Birmingham.Zagreb.
Hoić-BožićWang, X., N.,et & Holenko Dlab, M.al. (2021)2023). UvodPreparing ufor e-učenje:AI-enhanced obrazovnieducation: izazoviConceptualizing digitalnogand doba.empirically Rijeka:examining Sveučilišteteachers’ uAI Rijeci,readiness. OdjelComputers zain informatiku.Human Behavior.
JovanovićCope, B., N. (2012). Objektno orijentisano programiranje. Kruševac: SIGRAF, Visoka poslovna škola.
Kallia,Kalantzis, M., & Sentance, S. (2021). Threshold concepts, conceptions and skills: Teachers' experiences with students' engagement in functions. Journal of Computer Assisted Learning, 37(2), 411-428.
Klafki, W., Schulz, W., Cube, F., Moller, C., Winkel, R., & Blankertz, H. (1992). Didaktičke teorije. Zagreb: Educa.
Kyriacou, C. (1995). Nastavna umijeća: Metodički pristup za uspješno podučavanje i učenje. Zagreb: Educa.
Macphail, A. (2001). Nominal group technique: A useful method for working with young people. British Educational Research Journal, 27(2), 161-170. Accessed May 1, 2024, Available online: https://doi.org/10.1080/01411920120037117.
Madar,Searsmith, D. (1982)2020). UsingArtificial nominalintelligence groupfor techniqueeducation: toKnowledge fosterand productiveits behaviorassessment in groupAI-enabled discussions.learning Teachingecologies. PoliticalEducational Science,Philosophy 9(4),and 185-189.Theory. Accessed May 1, 2024, Available online: https://doi.org/10.1080/00922013.1982.994291800131857.2020.1728732​
Holmes, W., & Tuomi, I. (2022). State of the art and practice in AI in education. European Journal of Education, 57(4), 542-570 (Euro J of Education - 2…).
McCartney,Luckin, R., Holmes, W., Griffiths, M., & Sanders,Forcier, L. B. (2016). Intelligence Unleashed: An Argument for AI in Education. Pearson.
Perrotta, C., & Selwyn, N. (2019). Deep learning goes to school: toward a relational understanding of AI in education. Learning, Media and Technology.
Su, J., & Yang, W. (2023). Unlocking the power of ChatGPT: A framework for applying Generative AI in education. ECNU Review of Education.
Carnegie Learning. (2024). MATHia® by Carnegie Learning Wins “Best Artificial Intelligence Solution” in The EdTech Awards 2020. Business Wire.
Khan Academy. (2023). How We Built AI Tutoring Tools - Khan Academy Blog.
Edmentum. (2020). AI-Driven Data Analytics for Personalized Learning.
Tsirulnikov, D., Suart, C., Abdullah, R., Vulcu, F., & Mullarkey, C. E. (2023). Game on: immersive virtual laboratory simulation improves student learning outcomes & motivation. FEBS Open bio, 13(3), 396-407.
Fitzpatrick, K. K., Darcy, A., & Vierhile, M. (2005)2017). WhatDelivering arecognitive thebehavioral thresholdtherapy conceptsinto computeryoung science?adults U:with Proceedingssymptoms of thedepression Koliand Callinganxiety 2005using a fully automated conversational agent (Woebot): A randomized controlled trial. JMIR Mental Health, 4(2), e19.
Karsen, M., Pangestu, H., & Kristin, D. M. (2022, August). Acceptance of Miro and Padlet as Collaboration Tools on Hybrid Flipped Learning & Case-Based Learning in Education 4.0 (a case study approach). In 2022 International Conference on ComputerInformation ScienceManagement Education,and T. Salakoski & T. MäntyläTechnology (ur.),ICIMTech) str. 185.
McSkimming, B. M., Mackay, S., & Decker, A. (2023, October). Identification of Threshold Concepts for Intermediate Computer Science Students. In 2023 IEEE Frontiers in Education Conference (FIE) (pp. 1-5)577-582). IEEE.
Meyer,Holmes, J. H. F.W., & Land,Tuomi, R.I. (2003)2022). ThresholdState Conceptsof the art and Troublesomepractice Knowledgein LinkagesAI toin Wayseducation. European Journal of Thinking and Practising. U: Improving Student Learning Ten Years On. Rust, C. (ur.)Education. Oxford.
MinistarstvoD'Mello, znanostiS. iK., obrazovanja& [MZO]Graesser, A. (2023). Intelligent tutoring systems: How computers achieve learning gains that rival human tutors. In Handbook of educational psychology (2018)pp. 603-629). Kurikulum nastavnog predmeta Informatika za osnovne škole i gimnazije. Accessed April 12, 2024, Available online: https://skolazazivot.hr/wp-content/uploads/2020/06/INF_kurikulum.pdf.Routledge.
MinistarstvoZain, znanostiI. iN., obrazovanja [MZO]. (2019). Kurikulum međupredmetne teme Učiti kako učiti za osnovne i srednje škole. Accessed April 12, 2024, Available online: https://tinyurl.com/45rv8bu4.
Olugbara, Thamotharan P. Govender1 Oludayo O. “Reflection on Threshold Concepts for Developing Computer Programming Skills of First-year Information Technology Students.”
Resnick,Setambah, M., Maloney, J., Monroy-Hernández, A., Rusk,Othman, N., Eastmond, E., Brennan, K., ... & Kafai, Y. (2009). Scratch: programming for all. Communications of the ACM, 52(11), 60-67.
SJSU School of Information. (2023). The Expert Searcher and Threshold Concepts. Accessed December 10, 2023, Available online: https://youtu.be/4I1Ue0vpcMw.
Vinner,M. S., & Dreyfus,Hanapi, M. H. (2023). Use of Photomath Applications in Helping Improving Students’ Mathematical (Algebra) Achievement. European Journal of Education and Pedagogy, 4(2), 85-87.
Alawajee, O., & Delafield-Butt, J. (2021). Minecraft in education benefits learning and social engagement. International Journal of Game-Based Learning (IJGBL), 11(4), 19-56.
Su, J., & Yang, W. (2023). Unlocking the power of ChatGPT: A framework for applying generative AI in education. ECNU Review of Education, 6(3), 355-366.
Kostrić, M. (2021). AI in early childhood education: Opportunities and challenges. University of Zagreb.
Modrušan, A. (2023). The impact of artificial intelligence on the teaching profession: Opportunities and challenges. University of Zagreb.
Kessler, M., Loewen, S., & Gönülal, T. (1989)2023). ImagesMobile-assisted language learning with Babbel and definitionsDuolingo: comparing L2 learning gains and user experience. Computer Assisted Language Learning, 1-25.
Belda-Medina, J., & Goddard, M. B. (2024). AI-Driven Digital Storytelling: A Strategy for theCreating conceptEnglish as a Foreign Language (EFL) Materials. International Journal of function.Linguistics JournalStudies, 4(1), 40-49.
Bouchrika, I. Best School Organization Apps for researchTime Management, Note-Taking & Mind Mapping in mathematics education, 20(4), 356-366.2024.