Generative <scp>AI</scp> in Mathematics, Science, and <scp>STEM</scp> Education: Research, Applications, and Emerging Themes

Abstract

The advent of generative artificial intelligence (GenAI) has ushered in a transformative era for mathematics, science, and STEM education. GenAI is a highly sophisticated technology capable of mastering human language and leveraging human knowledge to identify patterns beyond human perception (Bozkurt 2023a). Its applications in natural language processing include chatbots, virtual assistants, language translation, and text generation, with growing research interest in its intuitive, human-like interaction capabilities (Bozkurt 2023b, Bozkurt and Sharma 2024; Lim 2024; Sharma and Bozkurt 2024). The emergence of tools like ChatGPT and other large language models (LLM) has introduced both opportunities and challenges (Redmond-Sanogo et al. 2024). From enhancing instructional strategies to streamlining assessment, GenAI is reshaping the educational landscape. However GenAI is not a replacement for teachers but a powerful tool to complement their work (NCTM, 2024). Although GenAI has the potential to enhance efficiency and creativity, its limitations and ethical considerations necessitate thoughtful integration (Khennouche et al. 2024; Liu et al. 2024). The editorial, Navigating the Artificial Intelligence Frontier (Redmond-Sanogo et al. 2024), illuminated the role of GenAI in fostering efficiency and critical thinking while cautioning against over-reliance on tools. It shared the need for understanding the complex interplay between technology, pedagogy, and ethical considerations in mathematics, science, and STEM education. This special issue explores the implications of GenAI in mathematics and STEM education, offering insights into its integration in classrooms, teacher preparation programs, and mathematics teacher education programs. This introductory article provides an overview of the special issue, situates it within the current literature, and highlights the emerging themes, trends, and future directions in GenAI research and practice. Since the publication of Navigating the Artificial Intelligence Frontier (Redmond-Sanogo et al. 2024), the academic discourse around GenAI integration has expanded significantly. In fact, at the recent 2024 School Science and Mathematics Association (SSMA) Convention in Knoxville, Tennessee, AI was the main topic in seven presentations and mentioned in several others. Recent research on GenAI in mathematics, science, and STEM education generally focuses on three main areas: perceptions of GenAI, its applications in teaching and learning, and strategies for developing AI literacy and skills among students. One of the central themes in the recent literature is educators' perceptions of GenAI, as these perceptions play a critical role in determining AI's adoption and integration in educational contexts (Ingersoll and Strong 2011; Venhatesh et al. 2003; Yue et al. 2024). Yue et al. (2024) highlighted the importance of teacher perceptions, noting that teachers with positive perceptions “often craft more robust and stimulating learning experiences that nurture higher-order thinking and problem solving abilities, which are essential for AI literacy” (p. 19509). Their study of 1664 K-12 teachers from Mainland China revealed that those with higher technological pedagogical content knowledge (TPACK), confidence, and positive perceptions were more likely to adopt GenAI in their classrooms. Similarly, Galindo-Dominguez et al. (2024) surveyed educators across primary and secondary schools in Spain, finding that while many recognized GenAI's potential to foster personalized learning and improve efficiency, teachers were concerned about its accuracy and ethical implications. Expanding on this, Collie and Martin (2024) investigated the role of contextual, occupational, and background factors on 389 Australian teachers' motivation and engagement with GenAI. Their findings indicated that professional growth striving and autonomy-supportive leadership were significant predictors of teachers valuing and integrating GenAI, with secondary teachers showing greater motivation and engagement than their elementary counterparts. Lim (2024), through a metaphor analysis of preservice early childhood educators, identified both positive and negative conceptualizations of AI in education. Positive metaphors, such as AI as “seed” and “key,” represented innovation and potential, while negative metaphors such as “maze,” reflected the complexity and challenges of AI integration. Chiu et al. (2024) emphasized the need for increased AI literacy among educators, showing that positive perceptions were often correlated with familiarity and confidence in using AI tools. Opesemowo (2025) highlighted a dual perspective, expressing optimism about the potential of AI to transform education while also acknowledging ethical and equity concerns. Collectively, these studies reveal the dual nature of educators' perceptions of GenAI, reflecting both optimism and apprehension. They underscore the critical importance of targeted professional development and revisions to teacher education programs to build confidence, address misconceptions, and promote equitable and effective GenAI integration. While educators' perceptions of GenAI shape its adoption and integration, its practical applications in teaching and learning reveal the transformative potential of this technology across various educational disciplines. The application of Gen AI in teaching and learning has shown promising results across mathematics, science, and STEM education. Huang and Qiao (2024) demonstrated that incorporating AI into a Science, Technology, Liberal Arts, and Mathematics (STEAM) curriculum improved high school students' computational thinking and interdisciplinary problem-solving skills. Similarly, Marouf et al. (2024) emphasized the benefits of intelligent tutoring systems in supporting personalized learning pathways. Bewersdorff et al. (2025) examined the potential of multimodal large language models (MLLMs) to transform science education. Their research highlighted the role of MLLMs in fostering engagement, personalized learning, and supporting multimodal content creation. Bozkurt and Sharma (2024) also hypothesized that ChatBots have the potential to “act as conversational agents that bridge the gap between traditional educational methods and the dynamic, social, and individualized needs of students” (p. vii). They conducted a systematic review of AI applications in education, identifying best practices for incorporating GenAI into instructional strategies. They highlighted the potential of AI to support differentiated instruction, automate routine tasks, and facilitate student engagement through interactive simulations and personalized feedback. (Gadanidis and Johnathan Tan 2024) explored the integration of AI concepts into Canadian mathematics curricula, showcasing how abstract mathematical principles underpin AI technologies like neural networks. Their work suggests that connecting AI with mathematics can demystify AI systems and foster deeper engagement. Despite these successes, Gadanidis and Johnathan Tan (2024) critiqued the current lack of alignment between AI tools and core disciplinary knowledge, particularly in mathematics education. Assessment practices are pivotal when integrating GenAI into education. Mao et al. (2023) reviewed the opportunities and challenges of using GenAI for assessments. They found that effective assessment frameworks should ensure that AI-enhanced tools align with learning objectives and are tailored to meet diverse learner needs. For instance, integrating adaptive learning systems allows for formative assessments that adjust to a student's current understanding, providing immediate feedback and customized learning pathways. (Bozkurt and Sharma 2024) emphasized that assessments using GenAI should not solely focus on outcomes but also consider the learning process, leveraging AI's ability to provide insight into student engagement and cognitive development. Furthermore, their analysis demonstrates the importance of using GenAI to streamline grading practices while ensuring fairness and reducing educator workload. Recommendations also highlight the need for educators to critically evaluate AI-driven assessment tools, ensuring transparency in scoring and feedback mechanisms. Gadanidis and Johnathan Tan (2024) suggest that integrating AI with project-based assessments can foster creativity and collaboration, bridging the gap between theoretical knowledge and practical application. However, challenges persist, including potential biases in AI algorithms and the risk of over-reliance on automated systems. Addressing these issues requires teachers and preservice teachers to build their capacity to design and interpret AI-augmented assessments. These findings suggest that while GenAI can enhance teaching and learning, careful design and alignment with educational goals are critical. Leveraging GenAI thoughtfully ensures that technology complements rather than replaces current teaching methods. Next, we will discuss the effective implementation, which requires a parallel focus on developing AI literacy and skills among both students and educators. AI literacy is a multidimensional concept encompassing the ability to understand AI principles, apply and use AI technologies, evaluate and create AI solutions, and consider ethical implications (Ng et al. 2021). It also involves fostering knowledge, skills, and values or attitudes as core educational outcomes, as highlighted by UNESCO (2022). Developing AI literacy and skills among students and teachers is essential for preparing them to navigate a technology-driven world (Touretzky et al. 2019). Specifically, providing early education on the foundational principles, ethical considerations, and societal effects of AI technology can promote AI's conscious and safe use while fostering the development of informed and responsible AI users (Ng et al. 2021). Gadanidis and Johnathan Tan (2024) conducted a study in Canada to explore how the mathematical principles used to develop AI systems to learn, make predictions, and generate outputs (neural networks) are taught in schools. They found that AI concepts were taught in the last 2 years of school and wanted to determine how AI literacy and a foundational understanding of AI concepts (linear algebra, big data management, statistics and probability, and differential and multivariate calculus) could be taught in earlier grades. They provided examples of ways they developed the foundational understanding of concepts such as limit, infinity, and matrices as early as second grade. They called for resources that integrate AI concepts into K-12 curricula while maintaining a strong connection to disciplinary knowledge. Huang and Qiao (2024) also recommended expanding research to other grade levels to refine pedagogical models and ensure seamless integration of AI across diverse educational contexts. Yang et al. (2024) explored the development of AI literacy in Kindergarten children in Hong Kong through technology-enhanced embodied learning. They discovered that when guided by the teacher, children could develop AI literacy through age-appropriate interactions. In addition to developing students' AI literacy and knowledge, Galindo-Dominguez et al. (2024) proposed tailoring AI training programs to the specific needs of educators at various educational levels to build their confidence and competence in using AI tools effectively. They highlighted that empowering teachers with practical training and resources can bridge the gap between theoretical knowledge and the practical application of AI in diverse educational settings. These strategies are essential to ensure that both students and teachers are equipped to navigate and leverage AI in educational settings. By equipping educators and students with the necessary skills to effectively navigate and leverage GenAI in educational settings, these strategies lay the foundation for exploring the broader implications and innovative contributions of GenAI, as highlighted in this special issue. This special issue presents a curated collection of articles addressing the multifaceted impact of GenAI in mathematics and STEM education. These collectively advance our understanding of GenAI's role in education. They address topics such as teacher perceptions and instructional innovations. They also examine GenAI from multiple perspectives—from mathematics teacher educators, to inservice teachers, to preservice teachers, and as tools for curriculum development. The four articles in this special issue each bring a unique perspective and insight into GenAI's role within the mathematics and STEM education community. The first article, “Exploring AI Integration in Math Education: Preservice Teachers' Experiences and Reflections on Problem-Posing Activities with ChatGPT” by Kim, Park, and Joung, investigated how prospective mathematics teachers interacted with ChatGPT, a GenAI chatbot, when exploring problems in mathematics education. Specifically, this study examined how they recognized and corrected errors in problems posed by ChatGPT. While prospective mathematics teachers effectively identified and verified correct problems, they had difficulties identifying and correcting these errors. This study suggests that using AI-based activities in teacher preparation programs provides opportunities for future educators to adapt to the AI-driven educational landscape. While AI tools have the potential to enhance learning, they can't replace the human teacher. This article also explored the ethical issues associated with AI, such as privacy and educational surveillance issues. While this study had a small sample size and only focused on fractions, it does provide insight and suggestions for future research areas. Secondly, in “Empowering Mathematics Teacher Educators: Exploring AI-Driven Mathematical Tasks,” Aqazade, Mauntel, and Atabas explores how mathematics teacher educators design mathematical tasks to support future mathematics teachers in using AI tools effectively. It investigates how the mathematics teacher educators' knowledge domains of content, pedagogy, technology, contextual, and pedagogical content knowledge inform their engineering of generating tasks through ChatGPT. The authors find that MTEs primarily use pedagogical and pedagogical content knowledge to craft prompts and that their prompt engineering techniques involve revising, generating, or editing tasks. Critical friend dialogues helped the authors identify key lessons learned. Lessons include different conceptualizations of a mathematical task, the importance of articulating learning goals, and the need to navigate and revise these goals during ChatGPT interactions. Findings from this study illuminate the potential of GenAI tools in mathematics teacher education when critically evaluated to ensure high-quality experiences. Using a TPACK- integrated model of mathematics teacher educator knowledge, future opportunities include exploring the role of AI tools in mathematics teaching and learning. In “STEM Teachers' Perceptions, Familiarity, and Support Needs for Integrating Generative Artificial Intelligence in K-12 Education,” Cheah and Kim used self-reporting to explore the perceptions, familiarity, and support needs of K-12 STEM teachers in using GenAI. This study found diverse perceptions with an almost equal split between positive and negative views regarding the impact of Generative AI on education. A correlation was found between teacher perceptions of Generative AI and their familiarity with its integration into educational settings. Professional development that caters to the diverse levels of GenAI familiarity and perceptions among teachers is needed, according to the findings of this study. Differentiating to address the diverse perspectives and needs of educators effectively is key in this area. The fourth article, “A Mile High and an Inch Deep: Exploring ChatGPT as a Mathematics Curriculum Development Tool” by Sawyer and Aga, explored ChatGPT as a tool for developing elementary mathematics curriculum. One hundred ninety-one ChatGPT- generated tasks based on Common Core standards were analyzed for levels of cognitive demand, using Stein and Smith's Task Analysis Guide Framework (1998). While most tasks were high cognitive demand, findings showed they were often repetitive and lacked mathematical explanations. Inaccuracies and potential biases highlight the need for critical curation when using AI for curriculum development. Therefore, mathematics teacher educators need to assist in developing prompt engineering techniques and critical curation strategies with preservice and inservice teachers to prepare them to effectively and critically use AI tools. While helpful tools, GenAI chatbots cannot replace the human teachers' expertise and judgment in meeting learners' specific needs. In the fifth article, Exploring the Interactions among Instructors, Learners and AI in Facilitating Mathematics Learning for Prospective Elementary Teachers, Yilmaz, Galanti, Naresh, and Kanbir explored the interactions between AI, prospective elementary mathematics teachers (PEMTs), and instructors in a mathematics content course. The study used the AI tool Khanmigo to facilitate small-group problem-solving and examined PEMTs' perceptions of their learning experiences. The findings show that AI-PEMT interactions created robust opportunities for PEMTs to construct mathematical knowledge. In contrast, instructor-PEMT interactions provided a support system for navigating the challenges of using AI. The study also highlights the need for instructors to be intentional in planning and facilitating AI explorations to foster student engagement beyond simply getting answers to questions. The implications of the study suggest that AI can be a valuable tool in mathematics education, but its effectiveness depends on instructors' careful integration and support by instructors. The use of AI in teaching and learning is expanding rapidly. With this expansion comes a need for targeted professional development such as described by Cheah and Kim, training to expand knowledge and understanding, and communication with all education stakeholders to better understand the opportunities and challenges presented by integrating generative AI in STEM teaching and learning. Yilmaz also described a need for careful, intentional interactions with GenAI to support learning in their work, which focused on using Khanmigo with preservice teachers. We can apply what we already know about technology to understand and grow. For example, in their article, Park and Joung consider teachers' generative AI use through the lens of TPACK, a construct we have used for two decades to consider domains of teacher knowledge related to technology integration. Through this work, we also acknowledge the importance of using generative AI tools judiciously, recognizing ways students and teachers might misuse the tool and circumvent opportunities for growth rather than using the tool to provide efficiency and increased access. Challenges include issues with the accuracy of the information generated, especially when used to generate curricular materials as described by Sawyer and Aga. The challenges reported from the study regarding outputs that were sometimes repetitive, erroneous, or at incorrect grade levels are good warnings to heed when using technology to create questions, tasks, rubrics, and other lesson components. Prompts should be engineered thoughtfully, such as in the cyclic process described by Aqazade, Atabas, and Mauntel of revising, editing, and generating. This dynamic use of GenAI as a back-and-forth tool rather than a one-off is a key characteristic of GenAI. This lesson of thinking of GenAI use as an ongoing is that can be integrated in professional development opportunities for teacher use and when students use generative AI for specific tasks. While this special issue valuable it also highlights for The interaction of GenAI in science, and STEM education is a complex with that have to be this issue focuses on teacher use for tasks related to we that schools are GenAI tools for tasks such as For student use in is a need to focus on using GenAI tools to support the students' to use AI tools as an to determine the or is also a need to explore the use of GenAI tools to support students in specific such as Learners and research in GenAI is expanding to be a for around and of AI which can educators and teacher educators in on teacher and student GenAI This is the most as current GenAI integration each school or based on experiences but often or a broader of implications. research and should explore how GenAI can be across a broader of instructional and assessment including and These tools the potential for more and learning particularly for students with diverse such as Learners or students with By leveraging GenAI to design multimodal educational educators can better address learning needs. is a need to the integration of GenAI in science, and mathematics should focus on how students can use GenAI to enhance their learning rather than simply This GenAI's role in fostering critical and skills through interactive adaptive learning and project-based tasks. promising is using GenAI to support feedback in both formative and assessments. studies could evaluate how GenAI systems can adjust to learners' offering personalized suggestions and to meet specific learning Expanding on this, might explore how GenAI can enhance teacher feedback by routine assessments while ensuring that feedback and the ethical of GenAI use Developing for GenAI integration across educational contexts is to ensure and This exploring the potential biases in content, ensuring privacy in data and addressing around educational a shared can educators, and in informed about GenAI professional development and teacher education programs should to GenAI tools and strategies. This prompt engineering critical curation of content, and ongoing with GenAI systems as tools for learning. and also examine how to these programs effectively across diverse educational settings, ensuring and By these future the can its understanding of GenAI's transformative potential while addressing the and challenges associated with its integration into mathematics, science, and teacher education. This was and with the of the human The human authors critically and the content to academic The authors also and potential biases in the The of the is the of the human

Ähnliche Arbeiten