Abstract

The role of Artificial Intelligence (AI) in science education gained momentum as educators pursued innovative and effective strategies to enhance students' learning outcomes. The review article analyzed various research reports in physics, chemistry, and biology (PCB) education that discussed the development of AI systems to achieve learning objectives and the utilization of AI for evaluating students' performances. The article classified the role of AI in PCB education separately, offering a comprehensive overview of its applications in these specific fields. The advantages of incorporating AI techniques into PCB education were also examined, with some reports indicating that AI-powered learning systems could enhance motivation, conceptual understanding, problem-solving skills, engagement, critical thinking, and reduce cognitive load among students. The review concluded by discussing the overall benefits and challenges of using AI in science education. Some challenges included ethical concerns, teacher training, and the need for ongoing research to ensure that AI technologies were effectively integrated into PCB education. However, the potential benefits of AI, such as tailored learning experiences and improved students' academic achievements, suggested that AI had a crucial role to play in shaping the future of science education and empowering the next generation of learners.

Introduction

In recent years, the significance of cultivating a broader range of learning skills and competencies beyond traditional subject knowledge in the field of science education was increasingly recognized. The reinforcement of this emphasis was seen through various educational frameworks, such as the National Curriculum in England, the National Education Policy (NEP 2020), and the Next Generation Science Standards (NGSS). Skills like metacognition, critical thinking, creative thinking, and collaboration have been recognized as essential for success in today's rapidly evolving world [1]. In the past, several pedagogies were developed and implemented to address the challenges encountered in teaching and learning science and to enhance students' conceptual understanding [2] [3]. These pedagogies were evaluated for their effectiveness and were based on different theoretical frameworks, such as inquiry-based learning, constructivism, cognitive load theory, conceptual change theory, social learning theory, Kolb's learning theory, self-regulated learning theory, situated learning theory, multiple intelligences theory [4] [5]. In the late 1980s, the emergence of computer and digital technologies drew the attention of educators to technology-based science learning [6]. However, the proliferation of mobile devices, internet availability, and educational software in the 2000s increased the accessibility and diversity of technology-based science learning [7] [8]. Interactive educational technologies, including online courses and tutorials, virtual and remote laboratories, simulations, games, multimedia presentations, and mobile apps, became popular among students and teachers, particularly during and after the COVID-19 pandemic [9] [10]. Electronic devices replaced traditional study tools like copybooks, and teachers adopted various learning platforms such as Google Classroom, Edmodo, Power School, and Moodle. Massive Open Online Courses (MOOCs) such as Coursera have become increasingly popular, indicating that society values electronic and distance learning approaches. On the other hand, educational institutions and researchers have developed increasingly interactive and intelligent technologies such as Virtual Reality (VR), Augmented Reality (AR), and Artificial Intelligence (AI) to enhance students' academic performance [11]. AI, in particular, was a rapidly advancing computer science field that involved the creation of intelligent machines and software capable of performing complex tasks that required human-like intelligence. By utilizing advanced algorithms and machine learning (ML) techniques, AI analyzes data to recognize patterns and make decisions [12]. AI-based educational software could customize the learning experience to suit each student's unique learning style and prior knowledge, encouraging active participation in the science learning process. The potential of AI in connecting learning to real-world applications was significant, as it simulated scenarios and contexts, resulting in increased motivation and understanding among students [13]. Additionally, AI offered multiple representations of abstract concepts, including interactive visualizations and virtual reality simulations. Furthermore, AI assisted in formative assessment by analyzing student data, identifying areas of difficulty, and providing targeted feedback. The development of AI systems had the potential to revolutionize science education by providing new opportunities and innovative approaches for students to learn and engage with scientific concepts. However, achieving an inclusive scientific understanding through AI systems requires interdisciplinary collaboration among computer scientists, natural scientists, and philosophers of science. The use of advanced computational methods, especially in AI, has the potential to transform scientific innovation, and these novel techniques facilitated the acquisition of new scientific knowledge [14]. The integration of AI in science education could transform traditional pedagogical approaches, leading to deeper conceptual understanding among students.

Background of the Study

John McCarthy, a well-known computer scientist and mathematician, first introduced the term "Artificial Intelligence" during the 1956 Dartmouth Conference, which was widely considered the birthplace of AI research. The conference brought together a group of scholars who discussed the possibility of creating machines that could imitate human intelligence and solve complex problems. This event marked the beginning of AI as a multidisciplinary field that encompasses computer science, mathematics, psychology, neuroscience, and other related disciplines. Figure 1 presents a timeline highlighting the significant advancements in AI-based technologies that occurred in the field of education, more than sixty years after the Dartmouth Conference. The evolution of AI has led to remarkable advancements in various techniques, including speech recognition [15], natural language processing [16], intelligent robotics [17], image processing [18], energy systems [19], autonomous vehicles [20], Fintech[21], healthcare [22], and more. Educational researchers utilized various AI technologies, such as intelligent teaching agents [23] [24], ML [25], learning management systems [26] [27], Chatbot language systems[28], and personalized learning environments [29], to augment students' learning outcomes and optimize the impact of their research. Studies reported that intelligent teachable systems showed significant improvement in learning among elementary students in different grade levels[30] [31] [23]. Moreover, the research also suggested that the teaching agents could equip students to understand new scientific concepts from regular lessons, even in the absence of AI software [30]. In a study, [27] it was demonstrated that AI-supported systems that took into account both cognitive and affective aspects of students' engagement boosted their task engagement and performance. The AI ethics curriculum allowed instructors the flexibility to opt for individual learning, collaborative learning, or blended learning approaches to attain their students' learning objectives.

The integration of AI technologies into science education encompassed the incorporation of diverse pedagogical methods to enhance students' learning outcomes, problem-solving abilities, and higher-order skills. Researchers [32] combined RoboGen software with inquiry-based learning and found it to be an effective educational tool for teaching robotics and AI to students, even those without prior experience in Evolutionary Robotics. Researchers further observed that the RoboGen platform is adopted by various universities globally, and ongoing enhancements to the software are expected to accelerate its uptake. Another research group [33] employed a blended learning approach that combined AI with the flipped classroom strategy in a mental health education course. The study's findings suggested that this blended learning approach stimulated students' interest in learning mental health education more than traditional teaching methods. In addition, researchers [34] [35] reported that the integration of AI with collaborative and project-based learning enhanced students' academic achievements and creative thinking skills. To improve students' scientific learning skills and abilities, prioritizing conceptual understanding was critical. Science educators explored several technologies to enhance students' understanding of key scientific concepts. The use of screencasts and simulations was investigated to improve the conceptual understanding of energy concepts in chemical bonding [36]. This study employed eye-tracking techniques to assess the cognitive load differences between the use of screencasts and simulations. An integration of AI into a game virtual laboratory and evaluated its efficacy in teaching linear momentum to undergraduate Physics students [37]. The effectiveness of the game virtual laboratory was compared to that of virtual laboratories. The study's findings revealed that the use of the game virtual laboratory resulted in a better conceptual understanding of linear momentum compared to virtual laboratories. In a recent study, [38] leveraged Amazon's Alexa skills to enhance students' learning outcomes. Researchers developed ForAlexa, an online tool that enables the creation of question-and-answer and random-quote forms, to develop apps for teaching evolutionary biology in their classes, as well as for alternative learning methods for students with special needs. Prioritizing comprehension over rote memorization was commonly associated with developing a thorough conceptual understanding of science subjects. This type of learning requires the ability to apply prior knowledge to new situations and overcome misconceptions and naive conceptions of scientific concepts [39]. With the rapid advancement of AI and its potential benefits, we argue that utilizing AI technologies is an effective means of enhancing conceptual understanding in science education for learners from diverse educational backgrounds.

Outcome

This review article aimed to examine the impact of AI on physics, chemistry, and biology (PCB) education and provided a comprehensive overview of the topic. It investigated how AI could aid in identifying and addressing student challenges, enhance critical thinking, learning skills, and create innovative educational experiences in PCB education. As we venture into the domains of physics, chemistry, and biology education, it becomes increasingly apparent that artificial intelligence holds the promise to transform conventional pedagogical approaches. By examining its capacity to foster critical thinking, address student hurdles, and pave the way for innovative educational experiences, we endeavor to shed light on the transformative possibilities AI brings to the forefront of pedagogy in these scientific disciplines. Through this comprehensive overview, we hope to inspire educators, researchers, and policymakers to embrace AI as a powerful ally in the quest for enhancing the quality and accessibility of PCB education.

The study investigates the following research questions (RQs):

RQ1. How are AI strategies and technologies employed in PCB education to empower students and what are the associated outcomes?

RQ2. What challenges are associated with integrating AI into PCB education, and what suggestions can be proposed to overcome them?

Evidence

The influence of AI on science education became a subject of growing interest in the field of educational research. This study aims to comprehensively examine the impact of AI on PCB teaching and learning by conducting a literature survey. The databases and the resources selected are based on the PRISMA guidelines [40] and the flow chart is given in Figure 2. A variety of sources, such as journal articles, conference proceedings, and online sources, were selected and analyzed to provide an overview of the current state of knowledge on the topic. To find relevant articles, several databases, including the Education Resources Information Center (ERIC), Web of Science, and Google Scholar, were searched using different keywords such as 'artificial intelligence in education', 'artificial intelligence in science education', 'AI in physics, chemistry, biology education', 'benefits of AI in science learning', 'AI in medical education', and 'AI in miscellaneous science courses'. In addition, backward literature surveys were conducted to identify more appropriate articles for review. The research reports selected for discussion were limited to peer-reviewed journals and filtered based on subject, technology-based pedagogy, research methodologies, and evaluation approaches implemented by the researchers.

Practice

The integration of AI technology into science education not only offered personalized learning opportunities but also enhanced cognitive abilities, resulting in a more engaging and enjoyable educational experience that fostered self-motivation. AI-powered simulations, virtual laboratories, and interactive tools created learning environments that allowed students to gain hands-on experience with complex scientific problems, investigate data, and draw conclusions. These student-centered teaching techniques were personalized and engaging, influencing students' cognitive skills and enabling them to think critically and understand complex concepts in virtual science learning environments. One of the research reports highlights the potential of AI chatbot programs to serve as valuable educational tools, enhancing nursing college students' motivation in their studies and supporting self-directed learning when compared to traditional learning [41]. The AI system's intelligent tools assisted students and teachers with initial scientific practices such as collecting suitable data, interpreting/evaluating data, and predicting the outcome based on experience. This approach helped learners to focus their time solely on developing core abilities such as critical and creative thinking, hypothesis formulation, methodology development, and theory generation. The combination of AI technology and learner abilities enhanced individual conceptual understanding in science education and enabled learners to connect and apply knowledge in real-world scenarios. Furthermore, these interactive simulations provided teachers with timely feedback, allowing them to assess students' progress and evaluate the effectiveness of their pedagogies.

RQ1: Impact of AI in PCB Education: Empowering Students and Outcomes

(i) AI Technologies and Strategies in Physics Education

Intelligent Tutoring Systems (ITS) were computer programs that acted as tutors and were recognized to provide several advantages in physics education. The use of ITS offered flexibility in learning, as students could learn at their own pace, or repeat lessons at any time, and were cost-effective in terms of maintenance. Computer tutors also engaged more students than human tutors, as they became more affordable, possessed infinite patience, and were considered less judgmental. Additionally, computer coaches offered reproducible instructions that continuously improved with feedback from the user community. There were various online computer systems available, specifically designed to evaluate students' answers to physics questions. Some examples of these web-based homework (WBH) systems include ALEKS, WebAssign, Sapling Learning, WileyPLUS, Expert TA, Mastering Physics, LON-CAPA, and OpenStax Tutor. These systems were user-friendly for educators and offered vast databases of problems that were sourced from and aligned with well-known physics textbooks. Andes [42] was a classical physics ITS that underwent extensive development and was widely recognized in the field of physics education. The system has been developed by researchers at the University of Pittsburgh's Learning Research and Development Center and the United States Naval Academy since 1996. Andes operated as a coached learning environment for classical physics, using an AI system to determine the user's mental state and provide appropriate guidance and feedback. Andes aimed to be minimally intrusive, meaning it did not offer guidance tailored to improve overall decision-making abilities crucial to effective problem-solving. From the student's viewpoint, it appeared that Andes emphasized a formulaic problem-solving approach more.

Physics educators have utilized various technologies to monitor learning processes and outcomes. In 1999, [43] Personal Assistants for Learnings (PALs) were used, in the form of computers, to teach science majors at Carnegie Mellon University about the applications of Newton's laws in their introductory physics course. The authors found that students found the PALs to be useful in enhancing their problem-solving skills and understanding of physics concepts and principles, and they recommended this approach to improve learning outcomes in physics education. Similarly, in 1997, Conati and colleagues presented a detailed account of the construction and structure of a Bayesian network [44]. The network was automatically created using the output from a physics problem solver that generated abstract solution plans and sequences of actions deemed acceptable for solving the problem. They also provided an initial assessment of the Assessor's update algorithms' performance on a representative network. In another investigation, [42] the fundamental principles were delineated that guided the development of Andes. These principles included inducing students to create new knowledge by providing hints that promoted independent problem-solving, creating a paper-like interface to facilitate the transfer of learning, delivering prompt feedback after each action to maximize learning opportunities and minimize time wastage, and permitting students to exercise flexibility in the sequence of actions and skip steps as needed. In 2004, researchers [45] developed an AutoTutor to teach Newtonian qualitative physics and computer literacy. Their approach was influenced by explanation-based constructivist learning theories, ITS which adapts to student knowledge, and research on tutorial discourse dialogue patterns. The study found that technology use in AutoTutor motivated students to engage in physics learning. Researchers [46] used Bayesian networks to construct probabilistic models of students for the Andes, an educational system for Newtonian physics that emphasizes encouraging student initiative and autonomy during the learning process. Based on the evaluations of the system's coaches, it was observed that students exhibited better learning outcomes and improved problem-solving abilities compared to traditional instruction.

In 2007, Krusberg conducted a comparative study of three primary technologies used in physics education, namely Physlet Physics, the Andes (ITS), and Microcomputer-Based Laboratory (MBL) [47]. The study aimed to evaluate the potential of these tools to promote conceptual change, enhance problem-solving skills, and achieve the objectives of the conventional physics laboratory. In 2013, [48] the Virtual Physics System (ViPS) was introduced and assessed its effectiveness in comparison to the traditional method of constructing and experimenting with actual pulleys. The results showed that ViPS was effective in aiding students' learning and correcting their misconceptions. Furthermore, virtual experimentation within the ViPS was found to be more efficient than physical experimentation with pulleys. The pedagogical and technical methods that were described and used to implement reciprocal teaching, a technique in which both students and a computer coach teach each other how to solve physics problems [49]. Their study focused on the topics of kinematics, dynamics, energy, momentum, rotations, and oscillations, and found a significant improvement in students' problem-solving abilities.

Researchers [50] utilized Moodoo, a system designed for modeling spatial teaching dynamics, to engage first-year undergraduate students in a physics unit. The study produced a set of conceptual mappings based on x-y positional data of teachers, which were used to create higher-level spatial constructs informed by Spatial Pedagogy concepts. The resulting metrics associated with these constructs enable novel investigations into classroom activity and potentially advance our understanding of physical behaviors and teacher-student proximity across different learning environments. The ML techniques were employed to grade open-ended physics questions, using algorithms such as Support Vector Machines (SVM), Gini, k-Nearest Neighbors (KNN), Breiman's Bagging, and Freund and Schapire's AdaBoost.M1 [51]. Their research findings showed that Adaboost.M1 had the most accurate predictive models and outperformed all other algorithms. Another research team [52] employed random forest classification to develop and evaluate prediction models aimed at identifying students at risk of struggling in introductory physics courses. In another study, the physics course outcomes were predicted by both the random forest and logistic regression models [53]. Additionally, some research focused on the development of AI systems and robots [54] to enhance problem-solving skills. Instructors demonstrated the use of a structured problem-solving approach that students were expected to follow, and the decision-making steps involved in this process were made clear and visible. Most of the research in physics education focused on exploring the concept of "reciprocal teaching" that involves using computers, where the computer coached students, and students, in turn, coached the computer in problem-solving.

(ii) AI Technologies and Strategies in Chemistry Education

Compared to AI technologies, AR and VR technologies were more prevalent in chemistry education, as supported by several studies. These technologies are employed for various concepts such as chemical bonding, laboratory experiments, gas laws, and molecular structure, as supported by several studies [55] [56] [57] [58]. In a few studies, eye-tracking technology was used to decrease students' cognitive load and augment their conceptual understanding [59]. In chemistry education research, AI was used not only for teaching but also for other activities such as student behavior detection [60], automated assessment [61], and at-risk student prediction [62]. In the future, AI technology and computer algorithms are expected to be used for all types of science education, including chemistry at all levels. Chemistry education researchers trained their students to develop intelligent systems and computer algorithms to achieve their learning objectives. For example, Oh and Kang [63] introduced a software manipulation and AI speech recognition training program to students, which included an AI carbon dioxide fountain experiment. This experiment facilitated students' conceptual understanding of scientific concepts and processes in a more logical and organized manner. It also showcased an instance of AI technology integration across various domains. An AI-driven system was primarily employed in laboratory courses to elucidate or predict chemical reactions. Birk [64] created PIRExS, an expert system designed to simulate human thinking and educate students on chemical reactivity. This tool served as a supplementary learning environment, with the instructor responsible for creating scenarios that allowed the program to function as an effective teaching aid. Birk discussed how the development of computer software taught about chemical reactivity and provided an approach to teach various topics in general chemistry. Healy and Blade [65] observed a high level of student engagement in designing organic syntheses and retrosyntheses based on their learning outcomes. During this course, the researchers employed IBM RXN for Chemistry, a web-based graphical user interface (GUI) that offered access to an AI-powered tool for learning activities. In another study, Schrier and others [66] demonstrated a simple form of AI in planning chemical experiments. It was created to minimize the total number of experiments required using various comparison sorting algorithms in computer science, such as insertion sort, binary insertion sort, merge sort, and merge insertion sort. This research found a connection between a chemistry experiment and computer science sorting algorithms.

Some studies were carried out to encourage student’s independent learning through technologies that helped them to recognize and learn from their logical errors. Soong and team [67] reported that ACD/Structure Elucidator, an AI software package, was utilized to aid in teaching upper-level organic chemistry classes about NMR spectroscopic principles and their application in determining the structure of complex organic molecules. This technology report provided students and instructors with a valuable tool to understand how complex organic structures were identified by examining correlations in multi-dimensional NMR spectra. In another research, [68] a study was conducted that introduced a classroom exercise for first-year science and engineering students. The exercise required the students to create a correlation to predict the boiling point of organic compounds using a data set of over 6,000 compounds. An artificial neural network (ANN) was employed to process the data and generate an engineering-quality correlation. The feedback from the students was exceptional, with the recognition that this was a rapidly growing area with many opportunities to incorporate these tools into design and engineering. Likewise, a workshop was [69] designed on ML, incorporating chemical examples and consisting of five Python notebooks and an undergraduate chemistry curriculum. The workshop covered topics including data modeling for classification and regression, as well as data visualization for a wine data set. The response from students was reported positive, with their interest in programming and ML. The research findings indicate a growing integration of AI technology in chemistry education, which includes the use of emerging technologies such as AI-powered teaching agents, eye-tracking techniques, learning analytics, etc., for pedagogical research. Research reports demonstrated the efficacy of AI tools and software in facilitating students' conceptual understanding of scientific concepts and processes, promoting engagement, and fostering independent learning. Specifically, AI systems offered an immersive digital experience, particularly for laboratory courses, and helped students acquire contextual knowledge during the learning process. The integration of AI technology in chemistry education complemented traditional teaching methods and prepared students for modern laboratory technology. Moreover, digital technologies, including AI, proved the potential to enhance the teaching of chemistry subjects that required the extensive use of structural drawings for explanations. With further advancements in AI technology, there was great promise for improving chemistry education and providing students with valuable tools for their conceptual understanding and critical thinking.

(iii) AI Technologies and Strategies in Biology Education

In the past few years, there has been significant progress and transformative changes in intelligent systems, which led to the creation, retention, and examination of data not just within technology, but also in various other fields. In the last five decades, both AI technology and biology have flourished. The biosciences and biotech sectors made remarkable advancements in drug discovery with the aid of sequencing and other high-throughput methods. Additionally, AI-based tools and applications were used in biological education to teach and assess learning outcomes in various learning environments. The evaluation of biology education research yielded data on the frequency of biological and everyday conceptions. Furthermore, it was discovered that the conceptions held by students were dependent on the contextual framework employed [70] [71]. The evaluation tools examined biological and everyday conceptions, which encompass the conceptions described in biology education [72]. Three evaluation tools, namely the Conceptual Inventory of Natural Selection (CINS) [70], Assessing Contextual Reasoning about Natural Selection (ACORNS) [73], and EvoGrader [74], were used to assess students' biological and everyday conceptions. The CINS was designed to facilitate constructivist and socio-constructivist learning, employing open-question formats to elicit information about everyday conceptions when the test was developed. The ACORNS tool responded to the criticism of forced choice by incorporating open response formats. EvoGrader provided visual representations of the percentages of biological and everyday conceptions, as well as their co-occurrences and a condensed categorization of students' responses.

A research group [75] employed the Siette system to evaluate learning outcomes in a botany laboratory involving dried plants in a study. This system aided in the understanding of the role of assessments, and various assessment elements and strategies were explored and integrated within the Siette environment. Another research group, [73] utilized a corpus of 565 undergraduates' biology evolutionary explanations to evaluate the effectiveness of an automated assessment program, Summarization Integrated Development Environment (SIDE). The results revealed that SIDE performed better than human expert scoring when scoring models were constructed and tested at the individual item level, but its performance deteriorated when suites of items or complete instruments were employed to build and test scoring models. Researchers [76] conducted a quasi-experiment to compare the impact of AI-enabled E-books and regular E-books on students' biology learning and found no significant difference in students' learning outcomes between the two types of books. Another study, [77] demonstrated the utility of combining tactile tools and AI to facilitate the learning of biology among the blind and visually impaired. Another widely used technique in biology education is ML. ML is a subfield of AI that involves the development of computer programs that are capable of enhancing their performance using experience, typically through the process of training. Bertolini and colleagues [78] utilized five ML methods to assess the predictive efficacy of predictive modeling in undergraduate students' outcomes in biology learning. The findings revealed that individual ML methods, particularly logistic regression, exhibited poor prediction performance, whereas ensemble ML methods, notably the generalized linear model with elastic net (GLMNET), achieved high accuracy. Deep learning, a subset of AI, is another popular technique among biology educators. Deep learning is one of the most significant techniques in contemporary AI research and is effectively used in various tasks, including image recognition, natural language processing, speech recognition, and others. It is a form of ML algorithm inspired by the structure and functions of the human brain and employed to construct models that could learn and make predictions or decisions based on large amounts of data [79]. The utilization of novel deep learning techniques that incorporated neural networks broadened the scope of ML into the realm of biology education.

Another important area in AI-based learning or teaching is medical education. Early efforts were focused on using computer-based technology for three-dimensional interactive anatomy teaching platforms. These platforms allowed students to navigate through anatomical structures and view them from different angles and layers. Examples of such products included the Primal Pictures website and Netters Interactive 3D anatomy. ANN and Convoluted Neural Networks (CNN) are two types of AI deep machine tools modeled on human brain neural networks. Hence, AI technology offers several benefits for medical education, such as serving as a friendly tutor, enabling self-paced learning, reducing the need for manpower, enabling remote operation, providing cost-effective solutions, and more. The interest in using AI in medical education has grown considerably in recent years, as evidenced by the increasing number of publications and citations. AI was used in all aspects of medical education, including curriculum development, analysis, learning, and assessment [80]. For instance, interactive teaching programs like Anatomy Chatbots [81], formative assessment tools, or clinical application quizzes were adapted and programmed into the deep learning AI framework to assess students' deep and logical learning and knowledge application in a clinical setting [82]. The ability to store and process massive quantities of raw, unstructured data by AI algorithms allowed the development of intelligent computing systems with sophisticated decision-making capabilities. In biology education, a variety of AI applications were employed, including the precise identification of the 3D geometry of biological molecules, such as proteins. However, the utilization of these methods in analyzing biological data necessitates heightened caution and rigorous scrutiny to elevate the standards of biological education and facilitate the students to enhance their conceptual understanding of biology.

Innovation

The use of AI in science education become increasingly popular in recent years due to the numerous benefits it offers. According to the literature survey, one of the most significant advantages of using AI in science education was the flexibility provided in learning. Although AI systems and intelligent interactive agents were effective in communicating with students, it is important to note that student-teacher interactions also played a crucial role in clearing doubts and preventing misconceptions during classroom sessions [83]. AI systems help teachers monitor students' performance to provide suitable assistance for their science learning. AI served as a virtual assistant for teachers to assess students' daily learning activities and homework, especially in larger class sizes where time constraints hinder human evaluation. Furthermore, AI systems aided teachers in designing lesson plans and implementing pedagogies for their courses [84]. Precise feedback analysis from students helped educators and education sectors refine and enhance the quality of education [85]. A pictorial representation of AI techniques and their applications in PCB education is given in Figure 3. One of the key benefits of AI-based learning is its convenience, as students have the flexibility to prioritize their learning according to their schedules. ITS allowed students to learn at their own pace, or repeat lessons whenever they needed to, and take as much time as they needed to understand concepts fully. This flexibility proved especially useful for students who struggled with certain topics or who had different learning styles. Another benefit of using AI in science education was the cost-effectiveness of ITS in terms of maintenance. Computer tutors catered to a larger number of students than human tutors, as they were more economical and could be programmed to have unlimited patience. Additionally, computer tutors were considered to be less judgmental, which helped students reduce anxiety and increase engagement in the learning process. In addition to flexibility and cost-effectiveness, AI-supported technologies also offer other advantages in science education. For example, online computer systems, such as the WBH systems, were designed specifically to evaluate students' responses. WBH systems offered vast databases of problems that were sourced from and aligned with well-known science textbooks, making it easier for educators to create and grade assignments [86]. Another example of an ITS is Andes, a classical physics system that uses AI to determine the learner's mental state and provide appropriate guidance and feedback [87].

AI-supported technologies were found to be useful in enhancing students' problem-solving skills, critical thinking skills, and conceptual understanding in PCB education. PALs and AI-supported reciprocal teaching were two such technologies that had been demonstrated to be effective in this regard. Bayesian networks and ViPS were also found to be useful in aiding students' learning and correcting their misconceptions. ML was another significant subfield within the field of AI. Its algorithms were capable of analyzing massive amounts of data and detecting patterns that may not have been easily identifiable by humans. Consequently, machines were empowered to make more precise predictions and decisions. ML also automated repetitive tasks and processes, saving time and reducing the risk of errors. Moreover, it could analyze data in real-time, providing insights and information that could be used to make informed decisions. Various pedagogies, such as active learning, inquiry-based learning, project-based learning, and problem-based learning, were utilized in science education to implement ML and enhance student academic achievements [88]. ANN was another example of AI-supported technologies used in science education. These networks were used to process data and generate engineering-quality correlations to learn laboratory experiments. Software tools, such as LUCID (Learning and Understanding through Computer-based Interactive Discovery), provided a web-based GUI that encouraged student participation in the learning process. AI technologies were also used for other activities such as student surveillance, algorithmic evaluation, and predictive analytics for student success. Schools started using AI-powered software to monitor student attendance, behavior, and engagement levels. AI technologies enable students to develop their intelligent systems and computer algorithms to achieve their learning objectives, allowing for independent learning and taking charge of their education. AI-supported technologies offer evaluation tools to assess students' understanding of concepts and prevent misconceptions. For example, three evaluation tools, namely CINS, ACORNS, and EvoGrader, were used to assess students' biological and everyday conceptions. Peer assessment was also incorporated into some AI-powered courses to enhance students' understanding of concepts. As AI technologies continued to evolve and become more accessible, they were likely to become increasingly important in science education, providing new ways to engage students and enhance learning outcomes. The majority of AI-based science education courses have been deployed at the undergraduate level, with a few illustrative instances presented in Table 1.

Table 1 A Few Cases of AI-Based Science Education Courses at the Undergraduate Level

Adaptive educational systems utilize a range of AI techniques, such as Fuzzy Logic, Decision Trees, Neural Networks, Bayesian Networks, Genetic Algorithms, and Hidden Markov Models [89]. However, a standardized approach for identifying the ideal learning theory and AI methodology to implement in a given science learning context was not yet established. In comparison to chemistry and biology, there were fewer research reports available on the use of AI in supporting physics education. While AI was effective in tasks such as image and speech recognition, which applied to biology and chemistry, it might have been less effective in physics education as it involved abstract concepts and mathematical reasoning. Furthermore, physics education possibly had distinct research priorities compared to biology and chemistry education. The initial step in achieving effective AI-powered physics education was the creation of a practical software framework that provided accessible guidance to students. This framework aimed to improve metacognitive abilities related to decision-making in problem-solving. The ultimate objective was to develop an ITS that supplemented conventional classroom instruction, helping students gain competence in solving physics problems through iterative guided practice, with an emphasis on the decision-making process. Chemistry education has both advantages and disadvantages. AI-powered systems excelled at explaining atomic and molecular level concepts to students at the school level, where it may have been difficult for teachers to stimulate imagination and experience at this level. Additionally, AI-based systems could predict chemical reactions and retrosyntheses, which reduced research costs and time. AI learning also supported higher education in chemistry, including research literature, data collection, instrumentation principles, hardware, data interpretation, spectral analysis, etc. However, when compared to physics and biology, chemistry learning using AI lacked the same level of conceptual understanding and practical experience. For example, students studying chemical reactions or laboratory experiments in a traditional setting had to physically observe the appearance of chemicals, detect their smell, and comprehend safety precautions before carrying out reactions. However, AI-based learning systems might not have provided the same level of attention to safety precautions and did not offer the sensory experience of chemicals and solvents. This could have resulted in discomfort when handling chemicals in an actual laboratory. The field of biology education research was a promising area for the application of AI-based teaching and learning techniques, which had the potential to revolutionize the methods of teaching and learning biology. According to a literature review, biology educators have utilized various AI-powered approaches such as chatbots, neural networks, deep learning, and ML to improve students' learning outcomes. These algorithms were designed to monitor the learning process and assess learning outcomes. In higher education especially in the healthcare field, the use of AI-powered systems was more prevalent. For instance, robotics technology was being used to enhance surgical precision and accuracy, and image processing techniques were used to process medical images to determine the type of disease and test results. Natural language processing was utilized to convert unstructured medical chart data into interpretable formats, and voice recognition captured important patient information in electronic medical records [90]. AI-powered systems like ANNs, decision tree algorithms, and instance-based algorithms were widely used for drug discovery and pharmaceutical product management [91]. AI had the potential to transform the teaching and learning of science by providing educators with interactive and engaging learning experiences for students, as well as access to powerful tools that could aid in the understanding of complex concepts in PCB subjects.

RQ2: Challenges and Suggestions for Integrating AI in PCB Education

Creating or implementing AI-powered PCB learning environments that met specific standards was a challenging task for science educators. These standards encompass: (i) ensuring barrier-free accessibility of technology for all students, including those with special needs or disabilities; (ii) designing pedagogy that places equal importance on AI-based learning, teacher-student interactions, and students' feedback on their learning; (iii) incorporating technology that enhances students' thinking abilities and skills, such as problem-solving, critical thinking, and creativity; (iv) considering the ethics and privacy concerns of teachers, students, and parents; and (v) ensuring that techniques and technology are flexible enough to be refined based on the needs of students and teachers. The following are some of the typical obstacles (Figure 4) discovered from the literature survey in the incorporation of AI technologies into PCB education.

(i) Funding Support

The cost of designing an AI-based PCB learning program at the high school level or college level depends on various factors [92]. The main factors were the type of AI technologies and resources used the complexity and scope of the program, the size of the student population, teacher training, and the level of customization required. It is advisable to consult with experts and vendors to obtain accurate and realistic cost estimates for a specific AI-based PCB learning program. Implementing AI-based pedagogy in economically weaker or rural academic institutions was challenging due to logistics, such as access to technology, internet connectivity, and the lack of trained teachers. These factors limited the ability of these institutions to fully utilize the potential of AI in enhancing the educational experience for students. However, science educators could utilize or seek a support system to address these challenges. The partnerships with technology companies, non-profit organizations, or government agencies help to provide access to technology and internet connectivity.

(ii) Ethical Concerns

AI-based PCB education raised concerns about data privacy, bias, and the role of human teachers [93]. The important ethical concern was the potential for AI to collect and store sensitive students’ and teachers’ data without consent or knowledge. These included information on the learning progress, preferences, and behavior patterns. It was important for educators and developers to implement strong data privacy policies to ensure that the student’s information should be used only for educational purposes and not shared with third parties without explicit consent. Another concern was the potential for bias in AI algorithms [94]. The AI-based educational tools might reflect the biases in their recommendations and decisions. These results perpetuate existing inequalities in education, such as gender or racial biases, and lead to further discrimination. Additionally, the utilization of AI in science education questioned the role of human teachers. Though the ITS provides personalized learning experiences, and real-time feedback, and is less judgmental, the studies showed that it was impossible to replace human intelligence. The teacher-student interactions were mandatory to enhance the emotional and social interactions that were critical to effective learning.

(iii) Curriculum, Pedagogy, Teacher Development and Assessment

Curriculum development would play a crucial role in the process of incorporating AI-based education which could help in students’ conceptual understanding of PCB subjects and familiarity with the latest technologies. The design of curriculum development based on AI should be suitable for students across various educational levels. The lower-grade students should engage with the basics of AI techniques such as machine learning and data analysis [95], whereas advanced neural networks and deep learning could be used for the higher-grade or college-level students [96]. By tailoring the curriculum, science educators could ensure relevant and suitable education to the needs of different age groups and educational levels. The curriculum also provides students with opportunities to work with AI technologies in a practical setting and teachers to reinforce the concepts that are taught in the classroom. These activities enhance students’ deeper understanding of AI and its applications in science learning and also provide them with valuable skills for their future careers. In designing pedagogy for AI-based physics, chemistry, and biology education, it is essential to prioritize personalization [97]. AI's adaptive capabilities allow content and pacing to align with students' unique needs, fostering a more tailored educational experience. Moreover, active learning strategies that encourage problem-solving, group activities, and hands-on experiments should be integrated to maximize student engagement [98]. AI can support these activities by providing resources, real-time feedback, and data analysis, enhancing the learning process. Continuous assessment, facilitated by AI, enables timely and informative feedback, guiding students on their learning journey [99]. Instructors, equipped with AI-driven analytics, can make data-driven decisions about their teaching methods, addressing areas where additional support may be required. Furthermore, embracing a variety of multimodal content types, such as videos, simulations, interactive models, and virtual labs, accommodates diverse learning styles [100]. Collaboration and teamwork skills, promoted by AI tools, are crucial in the scientific disciplines [101]. Ethical considerations regarding AI use, along with professional development for educators and ongoing research and evaluation, are also key components of a well-rounded AI-based pedagogical approach [102]. Lastly, ensuring accessibility for all students, regardless of their background or abilities, is a fundamental principle in the design of AI-enhanced PCB education.

Well-trained teachers who understood both the technology and the pedagogy were required for the effective integration of AI technologies into the classroom. Teacher education research highlights the importance of providing technology training for teachers, and some studies specifically emphasize the significance of Technological Pedagogical Content Knowledge (TPACK) [103] [104] for educators. Teacher training in AI could equip PCB educators with the skills and knowledge necessary to integrate AI into their teaching practices, design effective learning experiences, and maximize the benefits of AI in learning PCB concepts [105] [106]. Moreover, the training could help teachers build a culture of innovation and experimentation in science education. Also, the replacement of human teachers with AI was impossible. Instead, AI tools could be utilized to support teaching and enhance student learning outcomes. Schools and educational institutions must prioritize the training and professional development of teachers in AI to equip them with the necessary tools to provide the best possible learning experience for their students. Designing assessment techniques for AI-based physics, chemistry, and biology education necessitates a multifaceted approach to harnessing the potential of artificial intelligence. These assessments must adapt to the individualized learning journeys of students, adjusting question difficulty and feedback in real time [107]. Formative assessments, such as quizzes and interactive exercises, offer insights into students' ongoing progress, while summative assessments, including comprehensive exams, measure overall comprehension [108]. AI analytics play a pivotal role, providing instructors with data-driven insights into student performance. Automated grading, facilitated by AI, ensures objective and efficient evaluation, saving educators valuable time. Adaptive testing strategies challenge students appropriately by tailoring question difficulty to their skill level. Additionally, assessments enriched with multimedia elements, like videos and simulations, mimic the complexity of real-world scientific practices. Ethical considerations, encompassing privacy, fairness, and transparency, should guide assessment design. Ultimately, these assessments should be accessible to all, accommodating diverse learning needs and ensuring that no student is left behind [109]. Continuous evaluation and refinement are key, allowing AI-based assessment techniques to evolve and better serve the educational goals of physics, chemistry, and biology students.

Conclusions

In recent years, AI has emerged as a game-changer in various fields, including science education, leading to a growing interest among researchers in exploring its impact on PCB education. AI was employed in science education to enhance students' comprehension of concepts through customized learning experiences, instantaneous feedback, and targeted interventions. AI was also capable of analyzing students' learning patterns, detecting gaps in their understanding, and adapting to their learning requirements. In this review, a qualitative research study was conducted using a literature review as a research design and method. The study examined scientific articles from journals, reputed publications, and international conference reports to evaluate the impact of AI on different aspects of PCB education such as implementation, and learning outcomes. The research revealed that the advancement of computers and related technologies had led to the development of new pedagogies centered around AI, which brought numerous benefits to students and educators. AI had significant prospects in the field of PCB education, primarily by enhancing access to learning opportunities, expanding tailored learning experiences, and refining methods and techniques for more effective learning outcomes. Moreover, AI facilitated the adaptation and personalization of educational resources to cater to the unique needs and abilities of each learner, resulting in enhanced learning experiences. Despite having had concerns about AI replacing human teachers, the study found that AI-powered applications were unlikely to replace teachers but rather complemented and empowered them, freeing up time for more teaching and research activities. For instance, even though the advantage of the ChatGPT portal over other online resources, there was still a chance of getting erroneous responses to scientific queries. However, this presents an opportunity for students to explore science in a unique way that incorporates both human and AI expertise in their learning experience. The adoption of those systems and educational environments led to enhanced efficiency and effectiveness of teachers, ultimately resulting in improved instructional quality. Studies demonstrated that AI had a noteworthy influence on various aspects of science education, including management, teaching, and learning. To implement AI-based science education, having a comprehensive understanding of AI ethics and engaging in extensive, longitudinal research through cross-disciplinary collaborations were crucial. According to the study, as research on AI-based science education continued to grow, educators were expected to receive more practical guidelines and examples that would introduce new approaches to teaching and learning. Innovative AI applications for teaching and learning emerged, and AI gained greater significance in higher education. Therefore, educators and researchers had to acknowledge and adapt to these technological advancements to reap the full benefits of AI in science education. However, there was still much research needed to fully understand the benefits and limitations of these approaches and to develop effective AI-based tools and interventions for PCB education.