Is Artificial Intelligence an educational resource in Physical Education? A systematic review

Josep Bofill

Gil Pla-Campas

Enric M. Sebastiani

Abstract

Artificial Intelligence (AI) is breaking into Physical Education (PE) at an accelerated pace. Although the studies carried out are very recent, we present in this article a first exploration of the initial impact that AI is having on PE. Through a systematic review according to PRISMA standards, we examined the scientific literature on recently published studies between 2019 and 2024 that analysed how AI can contribute to improving learning in PE. For this purpose, a search was carried out in the specialised databases ERIC, ProQuest, Scopus and Web of Science (WoS), in which a total of 241 articles were found. After applying the established inclusion and exclusion criteria, a total of 10 studies were included and analysed according to three categories: scientific evidence on the use of AI in PE, areas of implementation of AI in PE, and educational use of AI in PE. The results showed a lack of research on the application of AI in PE, especially at primary and secondary eduation stages in Europe, suggesting that its integration is still embryonic. They also highlighted the potential of AI, such as video and voice analytics, Intelligent Computer Assisted Instruction (ICAI) and the Internet of Things (IoT) to personalise learning in PE, improve student satisfaction, physical performance and teaching effectiveness. However, it emphasises the need for further studies to explore the real impact of AI on the learning and development of PE competences.

Introduction

In the contemporary era, characterised by unprecedented technological advances, education faces transformational challenges and opportunities. Artificial Intelligence (AI) has begun to significantly influence various sectors, including education. As AI advances, its integration in the field of education and thus in Physical Education (PE) emerges as a promising field of study, offering potential to improve the way we teach and assess the area’s competencies. In this context, this article presents a systematic review of the existing literature on the application of AI in PE, which is interested in exploring how AI is being used to improve learning and assessment in this field.

In this sense, the incursion of digital technology in the field of PE in recent decades has been increasing by implementing digital technologies in PE through mobile applications (Gil-Espinosa et al., 2020; Lavay et al., 2015; Pulido González et al., 2016); accelerometers, GPS trackers and wearable technology to record physical activity (Marttinen et al., 2019); and the use of video for movement analysis (Koekoek et al., 2018). In addition, we can find the implementation of active video games that promote physical activity (Birinci et al., 2021; de Lima et al., 2020; Salgado & Scaglia, 2020). Nonetheless, despite these advances, the specific field of AI in PE still seems to remain relatively unexplored. This gap in research highlights the potential of AI to personalise learning, collect data, provide real-time feedback and offer a variety of learning tools to foster students’ interest and maintain their motivation to learn (Lee & Lee, 2021). Therefore, it is an opportune moment to find out the current trends in the use of AI in this area.

Similarly, the adoption of AI in education has recently gained momentum, with tools such as ChatGPT and DALL-E generating both fascination and concern among the education community (Delgado et al., 2024). As a result, educational institutions are adapting to the emerging capabilities of generative AI. This development has triggered debates on several critical issues such as preparedness, ethics, trust, impact and added value of AI in education, as well as the need for regulation, governance, research and training to manage its rapid evolution (Grassini, 2023). Nevertheless, AI not only encompasses the field of generative AI, but also opens the door to the fields of Machine Learning, Deep Learning and Natural Language Processing (NLP) (World Commission on the Ethics of Scientific and Technological Knowledge, COMEST, 2019, p3).

In order to harmonise all this technological profusion, regulatory measures are being implemented such as: the UNESCO guide for policy makers on AI in education (UNESCO, 2021) and the European Commission’s proposal to create a regulatory framework for AI (UNESCO, 2009). However, it remains to be seen whether these regulations taken by the actors in the education system have a real regulatory effect (Bond et al., 2024). Nonetheless, there are also important ethical considerations that need to be addressed when introducing AI in PE. Aspects such as data privacy and biases in AI algorithms are crucial issues that require careful attention to ensure responsible and beneficial implementation of these technologies (Moncada, 2024).

Using a systematic review methodology, this work analyses recent studies exploring the integration of AI in PE. To this end, practical applications of AI are examined, which includes personalised AI-based training systems, the use of motion analysis to improve sport technique, interactive learning platforms and the automatic assessment of physical performance. In addition, the potential benefits of these technologies, such as improved accuracy of assessments, increased student motivation and personalisation of learning, are discussed.

In this emerging and constantly changing context, where the scope of AI in PE is unknown, a systematic review is proposed with the following objectives: (1) To examine the existing scientific evidence on uses of AI that contribute to improving student learning in PE; (2) To understand the areas of implementation of AI in PE; and (3) To analyse emerging trends in the use of AI in PE.

In pursuit of these objectives, the study aims to provide a current view, in this decade, of how AI is being applied and can be developed in PE to improve learning processes in the subject and to clearly guide future lines of research and pedagogical practice in this emerging field.

Methodology

This study follows a systematic review design as per the guidelines and standards established in PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) (Moher et al., 2009), approved by the Blanquerna-URL Research Ethics Committee (ID 2223011D).

In order to locate and identify relevant studies, a bibliographic search was carried out in the different databases in the field of Health Sciences and Sport Science, specifically in ProQuest, Scopus and Web of Science (WoS). The search for publications was conducted on articles published between January 2019 and February 2024, using the following descriptors selected by the authors: “physical education” and “artificial intelligence”. These search descriptors must appear in the Title, Abstract or Keyword fields. 

Selection procedure

The selection of data for this article was made by the principal investigator, according to inclusion and exclusion criteria which are summarised in Table 1. 

Table 1

Inclusion and exclusion criteria.

See Table

The selection process continued with the identification, screening and eligibility phases outlined in Figure 1. These phases attempt to ensure the proper selection of items by applying the above criteria (Table 1) with these successive procedures: (1) removal of duplicate articles; (2) exclusion of irrelevant descriptive studies; and (3) extraction of relevant data from the final filtered articles. The online study selection software Rayya (Ouzzani et al., 2016) was used and with the consensus of the three investigators, 10 studies that met the inclusion criteria were selected for review.

Figure 1

See Full Size

Selection analysis

Once the studies selected and included in the review had been organised, the following data were extracted, organised by journal of publication, country of origin of the article, methodology, educational stage and results obtained, which are presented in Table 2. 

Table 2

Summary of selected articles.

See Table

Results

The content analysis of the selected articles was carried out using the three following categories: scientific evidence of the use of AI in PE, areas of implementation of AI in PE and educational use of AI in PE. 

Scientific evidence on the use of AI in PE

The papers of this systematic review have been published in different journals in other fields of Health and Sport Science; most of them (6) of technological and mathematical subjects such as Applied Mathematics and nonlinear sciences, Electronics, Computer-Aided Design and Applications, Scientific Programming and Electronics; and only two (2) of Health and Sport Science: Frontiers in public health; and two (2) of general scope: Heliyon and Sustainability. With regard to the stage of education, nine of the ten articles in the systematic review were carried out at university or college level. All studies used a quantitative methodological approach, mainly through surveys, evaluations and tests. It is important to note that only the study by Ba & Liu (2022) employed advanced statistical analysis, while the other articles were limited to descriptive statistical analysis (frequency tables, percentages and simple graphs). This methodological limitation must be considered when interpreting the results of our review.

Areas of AI implementation in PE

Regarding the AI used in PE, this varies from study to study. In this systematic review, studies using video and speech recognition, Intelligent Computer Aided Instruction (ICAI) and Internet of Things (IoT) were grouped together.

Sang and Chen (2022) and Yang et al. (2020) introduced speech recognition using intelligent robots to assist PE teachers. In this way, using the voice recognition system, the robot can answer the students’ questions and gather their feedback. With this human-computer interaction, students’ individual development and autonomous learning ability were enhanced (Sang & Chen, 2022). On the other hand, two studies focused on video recognition-based AI using the Kinect algorithm (Zhang et al., 2022) and Kinect v2 (He et al., 2024) to analyse motion. In the same direction, Liu (2022) used motion recognition by recording fitness movements to give feedback to students immediately.

Three articles focused on ICAI. Wu et al. (2022) conducted a survey on the application of wireless sensors and ICAI technology in PE, targeting students and teachers. Hu (2020) used intelligent computer-assisted badminton teaching where teachers create effective teaching programmes according to teaching objectives and carry out targeted teaching, which can effectively improve badminton lessons. Guo (2022) used the ICAI system to be able to select questions for students to answer, to be able to monitor the PE lessons and to be able to evaluate students’ behaviour in the different tasks set.

Finally, two articles focused on the use of IoT. Ba & Liu (2022) used AI to assess students’ performance and predict their results in PE tests. Meanwhile, Yu and Yang (2023) combined AI and IoT to study the application mode of practical and innovative teaching in university PE through data analysis with the application of algorithms.

Educational use of AI in PE

As can be seen in Table 2, it was possible to find similarities between the different articles in relation to the educational use of AI in PE. First of all, some papers put emphasis on assessing students’ satisfaction with the integration of AI technologies in their learning process. Yang et al. (2020) investigated student interest and attitude towards learning in PE lessons by means of a questionnaire, comparing the results with a control group and obtaining higher satisfaction in students who used AI. Along the same lines, He et al. (2024) analysed student satisfaction through a questionnaire including satisfaction with the experience, interest in the sessions, attractiveness of the interactive teaching system and promotion of learning and observed a better satisfaction on the group that used video analysis with AI. Zhang et al. (2022) and Hu (2020) observed that the use of virtual simulation technology improves student interest and motivation in PE lessons. Wu et al. (2022) conducted a survey on the opinion of PE teachers and students, and found that 40% of students were very satisfied with the use of the smart PC in PE lessons and 67% were satisfied with the use of the smart CC-AS in PE lessons. However, this was not an experimental intervention, but rather a survey of teachers’ and students’ opinions.

Secondly, other studies focused on evaluating the results obtained in sport tests or events with the use of AI in PE. He et al., (2024) analysed the improvement of 400-m sprint performance observing better results in the post-test having received systematic training with AI support. On a similar note, Yu & Yang (2023) concluded that the implementation of a new model of PE by introducing AI can improve students’ physical test scores. Focusing on sports, Hu (2020) used a badminton test to compare the results of the control group and the experimental group, with better results in the experimental group. Finally, Ba & Liu (2022) used IoT in their study, focusing on the intelligent algorithm based on the feedforward neural network (FNN) and thus can effectively predict students’ score in the national college PE exam. 

Thirdly, some studies focused on improving the dynamics of PE classes, such as efficiency, communication and personalisation. In relation to the efficiency of PE classes, Yu and Yang (2023) found that a PE model incorporating IoT and AI improves teaching efficiency compared to the traditional PE model, although this improvement is observed after two weeks. On the same note, Guo (2022) concluded that the integration of AI in PE management can improve the efficiency of student learning. On the other hand, Liu (2022) and Sang and Chen (2022), in their studies on an intelligent teaching system for basic movements in PE and the use of a voice recognition assistant respectively, concluded that AI can improve communication in PE by providing more feedback and personalisation of students’ learning and thus promoting their autonomy. 

Discussion

Scientific evidence on the use of AI in PE

Regarding the characteristics of the articles included in the review, it should be noted that all the publications in the systematic review are from China. The concentration of research on the application of AI in PE in China, as evidenced in our systematic review, can be attributed in part to the country’s unique educational structure, which actively integrates PE at higher education levels, including schools and universities, and to differences in its pedagogical models.

The authors suggest that PE at university educational stages provides fertile ground for innovation and research at the intersection of technology and PE. Nine of the ten articles in the systematic review have been conducted at these stages, with a result similar to other systematic reviews (Zhou et al., 2023). Engagement with PE at these educational stages creates significant opportunities for the development and application of AI solutions aimed at improving the quality of teaching and learning in this field.

In line with this, the review also highlighted the almost non-existent research on AI in PE at primary and secondary level. This absence is particularly striking given the relevance of PE in the educational curriculum and its potential to benefit from AI applications, such as personalising learning, analysing physical performance and promoting healthy lifestyles (Lee & Lee, 2021). Nevertheless, the selected university studies and the results we have observed from them may have transferability to secondary or primary school.

Another critical finding of this systematic review is the variability in the quality of the included studies. It is important to note that many of the reviewed studies present descriptive statistical analyses and questionable research quality. For example, the work of Sang and Chen (2022) relied mainly on authors’ opinions and surveys, without advanced statistical analysis. Moreover, in some cases, interventions are not well detailed (Liu, 2022; Zhang, 2021), which makes replicability and accurate evaluation of results difficult. The possible introduction of bias due to reliance on self-reported data and the lack of rigour in statistical analyses also merits attention. The lack of consideration of confounding variables in many studies may affect the reliability of the findings, introducing additional distortions in the results. Despite these limitations, the integration of AI in PE shows promising potential.

Areas of AI implementation in PE

Concerning the areas of implementation of AI in PE, the results of this systematic review underline its diversity and potential by demonstrating different areas of implementation of AI in PE to enhance learning in this field. The use of voice and video recognition, ICAI and IoT illustrates an innovative landscape where technology not only facilitates interaction between students and teachers, but also promotes more autonomous and personalised learning. For example, the use of intelligent robots that respond to questions through voice recognition represents a significant advance in human-computer interaction, offering a richer learning experience tailored to the individual needs of students.(Sang & Chen, 2022). In addition, the implementation of technologies such as Kinect for motion analysis and

he combination of AI with IoT for the study of practical applications in PE show how the integration of these tools can offer a more accurate and detailed approach to physical performance and sport activity. (He et al., 2024; Yu & Yang, 2023). The ability to provide immediate and personalised feedback to students, based on detailed analysis of their movements, highlights the potential of these technologies to transform the teaching of PE, allowing for more objective and individually tailored assessment. 

Educational use of AI in PE

The results on the satisfaction and increased interest of students in AI-assisted PE classes are indicative of how emerging technologies can revitalise traditional teaching and learning methods. However, it is important to keep in mind that the real goal of incorporating AI in PE goes beyond mere student satisfaction and should focus on objectively improving subject-specific learning.

On the other hand, findings related to academic performance highlight that the potential of AI can lead to significant improvements in both students’ physical performance and their academic achievement related to PE. In line with this, AI technologies, such as motion analysis and IoT-based systems, can provide detailed assessments and real-time feedback. (Ba & Liu, 2022; He et al., 2024; Liu, 2022; Sang & Chen, 2022; Yang et al., 2020; Zhang et al., 2022). This focus on personalisation, precision in teaching and evaluation of PE can not only increase the effectiveness of training sessions, but also motivate students by providing them with a clearer understanding of their own progress and areas for improvement.

Finally, studies focusing on classroom effectiveness, communication and personalisation underline the importance of integrating AI into the dynamics of the EF classroom to improve teaching efficiency and foster greater student interaction and engagement. The ability of AI to provide instant and personalised feedback is a significant added value, promoting learner autonomy and greater understanding of PE concepts (Lee & Lee, 2021). 

However, the promising educational future through the integration of AI in PE is not without ethical and practical challenges. Specifically, it is crucial to consider the privacy of student data, avoid biases in AI algorithms and understand the potential impact on the teacher-student relationship. Addressing these challenges effectively is essential to ensure that AI actually benefits the educational process and does not introduce new inequalities or ethical problems. Continued exploration of how AI can influence PE teaching is not only necessary, but essential to ensure that AI is used in a way that maximises the benefit to students and contributes to the improvement of PE learning.  

Conclusions and future lines of research

Through the results obtained and despite the growing interest in the integration of advanced technologies in education, this review has revealed a scarcity of research, specifically addressing the use of AI in PE, particularly in the primary and secondary stages of education in the European context. In the same vein, it was observed that the studies found in the first search of the review focused exclusively on investigating the different AI tools in PE, but none of them explored whether the use of AI influences the teaching of PE. This was an inclusion criterion and, for this reason, studies were not included in the review.

This reality not only highlights the need for more research in this area, but also suggests that the adoption of AI in PE teaching is at a very embryonic stage and probably without significant educational experiences to investigate. This leaves ample room to explore how AI can enrich and transform pedagogical practices in this field. Along these lines, Celik et al. (2022) concluded that AI offers teachers different opportunities to improve the planning, implementation and evaluation of their teaching. 

Regarding the areas of AI implementation in PE, video and speech recognition and analytics, ICAI and IoT are possible areas of AI that can contribute to the improvement of PE by providing personalisation of learning and enriching the learning experience with real-time data and feedback. It remains to be seen whether the education system as a whole finds these resources necessary to achieve its objectives or whether the system already has the resources.

Concerning the educational use of AI in PE, the results show that AI can improve the satisfaction, outcomes and effectiveness of PE sessions. While these aspects are important to validate the acceptability and feasibility of AI in the classroom, there is a clear lack of studies that delve deeper into the direct impact of AI on students’ learning and improvement of specific PE competences. This gap in research suggests that, while AI advances may be positively received by the educational community, there is still much to be explored in terms of the actual usefulness of AI in PE learning.

Against this background, the present review invites future research to venture into the exploration of AI in PE, especially in primary and secondary education. It is imperative that future studies focus not only on technical and satisfaction aspects, but also on assessing how AI can transform PE learning. Future research should adopt multidisciplinary methodologies to address these questions, working closely with educators, technologists and students to design and evaluate pedagogically sound classroom implementations of AI tailored to the specific needs of the PE domain. This step should be taken into account insofar as there has recently been a strong debate on the appropriateness of the use of digital devices in the classroom in schools (Moncada, 2024; UNESCO, 2021). Future studies could benefit from a more diversified approach with a specific focus on primary and secondary education. In addition, it is essential that the interventions are specified in detail to facilitate their replicability, and that the process of data collection is clearly described. It is also important that studies include both qualitative and quantitative data, as this can contribute to a more complete, comprehensive and holistic view of the study problem. (Castañer et al., 2013).

Another limitation of the present work is the incipient stage of academic development of this field of study, which implies a relatively sparse in advance literature base, limiting the ability to carry out a comprehensive analysis with a broad empirical basis. Another limitation of the study lies in its focus on the use of AI in PE and does not consider other emerging digital technologies such as Virtual Reality or Augmented Reality whose use in PE can help to improve this subject (Zhou et al., 2023).

In summary, this systematic review underlines the lack of scientific literature with much more verifiable evidence on the use of AI in PE and represents a significant opportunity to enrich the field of PE by incorporating AI. Future studies can address the identified gaps and explore new research directions in order to define where the use of AI can improve the quality of FE.