Revealing the true potential and prospects of augmented reality in education

Abstract

Augmented Reality (AR) technology is one of the latest developments and is receiving ever-increasing attention. Many researches are conducted on an international scale in order to study the effectiveness of its use in education. The purpose of this work was to record the characteristics of AR applications, in order to determine the extent to which they can be used effectively for educational purposes and reveal valuable insights. A Systematic Bibliographic Review was carried out on 73 articles. The structure of the paper followed the PRISMA review protocol. Eight questions were formulated and examined in order to gather information about the characteristics of the applications. From 2016 to 2020 the publications studying AR applications were doubled. The majority of them targeted university students, while a very limited number included special education. Physics class and foreign language learning were the ones most often chosen as the field to develop an app. Most of the applications (68.49%) were designed using marker detection technology for the Android operating system (45.21%) and were created with Unity (47.95%) and Vuforia (42.47%) tools. The majority of researches evaluated the effectiveness of the application in a subjective way, using custom-made not valid and reliable tools making the results not comparable. The limited number of participants and the short duration of pilot testing inhibit the generalization of their results. Technical problems and limitations of the equipment used are mentioned as the most frequent obstacles. Not all key-actors were involved in the design and development process of the applications. This suggests that further research is needed to fully understand the potential of AR applications in education and to develop effective evaluation methods. Key aspects for future research studies are proposed.

Introduction

The current epoch is marked by swift advances in Information Technology (IT) and its pervasive applications across all industries. The most prominent technological terms are Augmented Reality (AR), Virtual Reality (VR), and Mixed Reality (MR), which have gained popularity for professional training and specialization. AR has been defined variously by researchers in the fields of computer science and educational technology. Generally, AR is defined as the viewing of the real physical environment, either directly or indirectly, which has been enriched through the addition of computer-generated virtual information (Carmigniani & Furht, 2011). Azuma (1997) described AR as a technology that combines the real with the virtual world, specifically by adding virtual-digital elements to the existing real data. This interactive and three-dimensional information supplements and shapes the user's environment. Azuma (1997) proposed that AR systems should exhibit three characteristics: (i) the ability to merge virtual and real objects in a real environment, (ii) support real-time interaction, and (iii) incorporate 3D virtual objects. Milgram and Kishino (1994), to avoid confusion among the terms AR, VR, and MR, presented the reality-virtuality continuum (see Fig. 1).

Fig. 1

Reality—Virtuality Continuum [Adapted from Milgram and Kishino's (1994)]

Figure 1 illustrates that Mixed Reality (MR) lies between the real and virtual environments and includes Augmented Reality (AR) as well as Augmented Virtuality (AV). AR refers to any situation where the real environment is supplemented with computer-generated graphics and digital objects. In contrast, AV, which is closer to the virtual world, augments the virtual environment with real elements (Milgram & Kishino, 1994). Unlike VR, AR aims to mitigate the risk of social isolation and lack of social skills among users (Kiryakova et al., 2018).

AR is recognized as a novel form of interactive interface that replaces the conventional screens of devices such as laptops, smartphones, and tablets with a more natural interface, enabling interaction with a virtual reality that feels completely natural (Azuma, 1997). AR can be classified into four main categories based on its means and objectives:

• Marker-based AR: Marker tracking technology uses optical markers (flat structures with long edges and sharp corners, also known as triggers or tags), captures the video input from the camera, and adds 3D effects to the scene. This type of augmented reality is mainly used to collect more information about the object and is widely used in department stores and industries (Schall et al., 2009).

• Markerless or location-based AR: This technology gets its name because of the readily available features on smartphones that provide location detection, positioning, speed, acceleration and orientation. In this type of AR the device's camera and sensors use GPS, accelerometer, compass, or other location-based information to recognize the user's location and augment the environment with virtual information (Kuikkaniemi et al., 2014).

• Projection-based AR: This type of AR typically uses advanced projectors or smart glasses to project digital images onto real-world surfaces, creating a mixed reality experience. Changing the movement on the surface of the object activates the display of images. Projection-based AR is used to project digital keyboards onto a desk surface. In some cases, the image produced by projection may not be interactive (Billinghurst & Kato, 2002).

• Superimposition-based AR: In this type of AR overlay technology replaces an object with a virtual one using visual object recognition. This process usually occurs by partially or completely replacing the view of an object with an augmented view. First Person Shooter (FPS) games are the best example of augmented reality based on superimposition (Billinghurst & Kato, 2002).

It's important to note that these categories are not mutually exclusive, and some AR applications may use a combination of these types.

Mobile augmented reality has gained popularity in recent years, thanks to advancements in smartphones and more powerful mobile processors. It has opened up new possibilities for augmented reality experiences on mobile devices (Tang et al., 2015). Mobile AR is a technology that allows digital information to be overlaid on the real-world environment through a mobile device, such as a smartphone or tablet. This technology uses the camera and sensors of the mobile device to track the user's surroundings and overlay digital content in real-time. Mobile augmented reality applications can range from simple experiences, such as adding filters to a camera app, to more complex ones, such as interactive games or educational tools that allow users to explore and learn about their environment in a new way. Mobile AR app downloads have been increasing worldwide since 2016 (Fig. 2). The global AR market size is projected to reach USD 88.4 billion by 2026 (Markets & Markets, 2023).

Fig. 2

Consumer mobile device augmented reality applications (embedded/standalone) worldwide from 2016 to 2022 (in millions) [Source: Statista, 2023a, 2023b]

Technological developments have brought about rapid changes in the educational world, providing opportunities for new learning experiences and quality teaching (Voogt & Knezek, 2018). It is no surprise that the field of education is increasingly gaining popularity for the suitability of Augmented Reality applications (Dunleavy et al., 2009; Radu, 2014). In recent years, many researches have been published that highlight the use and effect of AR in various aspects of the educational process, enhancing the pedagogical value of this technology (Dede, 2009).

It is worth mentioning the interest observed in recent years by Internet users in the Google search engine, regarding the term "augmented reality in education". According to the Google tool (Google Trends), the chart below shows the number of searches on the Google search engine for Augmented Reality in education from 2015 to the present.

Compared to the past, the use of AR has become considerably more accessible, enabling its application across all levels of education, from preschool to university (Bacca et al., 2014; Ferrer-Torregrosa et al., 2015). AR has greatly improved the user's perception of space and time, and allows for the simultaneous visualization of the relationship between the real and virtual world (Dunleavy & Dede, 2014; Sin & Zaman, 2010). Cheng and Tsai (2014) also noted that AR applications facilitate a deeper understanding of abstract concepts and their interrelationships. Klopfer and Squire (2008) highlighted the novel digital opportunities offered to students to explore phenomena that may be difficult to access in real-life situations. Consequently, AR applications have become a powerful tool in the hands of educators (Martin et al., 2011).

Augmented reality applications provide numerous opportunities for individuals of all ages to interact with both the real and augmented environment in real-time, thereby creating an engaging and interesting learning environment for students (Akçayır & Akçayır, 2017). AR apps are received positively by students, as they introduce educational content in playful ways, enabling them to relate what they have learned to reality and encouraging them to take initiatives for their own applications (Jerry & Aaron, 2010). The international educational literature highlights several uses of AR, which have been designed and implemented in the teaching of various subjects, including Mathematics, Natural Sciences, Biology, Astronomy, Environmental Education, language skills (Billingurst et al., 2001; Klopfer & Squire, 2008; Wang & Wang, 2021), and even the development of a virtual perspective of poetry or "visual poetry" (Bower et al., 2014).

The increasing interest in augmented reality and creating effective learning experiences has led to the exploration of various learning theories that can serve as a guide and advisor for educators considering implementing AR technologies in their classrooms (Klopfer & Squire, 2019; Li et al., 2020). The pedagogical approaches recorded through the use of appropriate AR educational applications include game-based learning, situated learning, constructivism, and investigative learning, as reported in the literature (Lee, 2012; Yuen & Yaoyuneyong, 2020).

By examining relevant literature and synthesizing research findings, a systematic review can provide valuable insights into the current state of AR applications in education, their characteristics, and the challenges associated with their implementation in several axes:

• Identifying trends and characteristics: It can explore the different types of AR technologies used, their educational purposes, and the target subjects or disciplines. This can provide an overview of the current landscape and inform educators, researchers, and developers about the range of possibilities and potential benefits of AR in education (Liu et al., 2019).

• Assessing effectiveness: A systematic review can evaluate the effectiveness of AR applications in enhancing learning outcomes. By analyzing empirical studies, it can identify the impact of AR on student engagement, motivation, knowledge acquisition, and retention. This evidence-based assessment can guide educators in making informed decisions about incorporating AR technologies into their teaching practices (Chen et al., 2020; Radu, 2014).

• Examining implementation challenges: AR implementation in educational settings may pose various challenges. These challenges can include technical issues, teacher training, cost considerations, and pedagogical integration. A systematic review can highlight these challenges, providing insights into the barriers and facilitating factors for successful implementation (Bacca et al., 2014; Cao et al., 2019).

• Informing design and development: Understanding the characteristics and challenges of AR applications in education can inform the design and development of new AR tools and instructional strategies. It can help developers and instructional designers address the identified challenges and create more effective and user-friendly AR applications tailored to the specific needs of educational contexts (Kaufmann & Schmalstieg, 2018; Klopfer et al., 2008).

This paper concludes by offering researchers guidance in the examined domain, presenting the latest trends, future perspectives, and potential gaps or challenges associated with the utilization of augmented reality (AR) in education. Supported by a series of research questions, the paper delves into diverse facets of AR applications, encompassing target audience, educational focus, assessment methods, outcomes, limitations, technological approaches, publication channels, and the evolving landscape of research studies over time. By addressing these questions, the study endeavors to provide a comprehensive understanding of the unique characteristics and trends surrounding AR applications in the educational context.

The paper is structured for easy readability, with the following organization: The "Material and Methods" section outlines the systematic review's methodology, inclusion/exclusion criteria, research questions guiding the analysis, and a list of quality criteria for chosen articles. In the subsequent "Results" section, the selection process results are detailed, aligning with the prior research questions. This section specifically delves into the technological approach, assessment methodology, quality outcomes, and key findings (including scope, outcomes, limitations, and future plans) of each study. Following this, the "Discussion" section offers a thorough analysis of the findings, unveiling opportunities, gaps, obstacles, and trends in AR in education. Lastly, the "Conclusion" section summarizes the systematic review's major findings and offers guidance to researchers pursuing further work in the field.

Materials and methods

In this scientific paper, a systematic literature review was conducted for the period 2016–2020 to determine the characteristics of augmented reality educational applications and whether they can be effectively utilized in various aspects of the educational process. The study followed a Systematic Literature Review (SLR) protocol, which involves identifying, evaluating, and interpreting all available research related to a specific research question, topic, or phenomenon of interest (Kitchenham, 2004). The paper is structured according to the PRISMA Checklist review protocol (Moher et al., 2009), which outlines the stages of the systematic literature review. The stages of the systematic literature review are framed by the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses), which has a wide application in research that aims to study a topic in depth by examining the research that has already been done and published (Grant & Booth, 2009).

The electronic databases Science Direct, Scopus, Google Scholar, Web of Science, MDPI, PubMED, IEEExplore, and ACM Digital Library were searched for scientific articles using keywords (employing Boolean phrases) such as augmented reality, AR, application, education, training, learning, mobile, app, etc., according to PICO (Stone, 2002). The keywords used in the queries were as follows: (AR OR “augmented reality”) AND (application OR education OR educational OR teaching OR app OR training OR learning OR mobile OR ICT OR “Information and Communication Technologies” OR tablet OR desktop OR curriculum). The selection of the aforementioned databases was based on considerations of comprehensiveness, interdisciplinarity, quality, international coverage, and accessibility. These databases collectively offer access to peer-reviewed journals and conference proceedings from diverse academic disciplines, ensuring a broad and reliable coverage of AR in education research. Additionally, the inclusion of Google Scholar allows for the identification of open access literature. Their reputation, interdisciplinary nature, and search capabilities further support a comprehensive and credible examination of the topic. The selected databases are known for their frequent updates, enabling the review to capture the latest research and stay up-to-date with the rapidly evolving field of AR in education. Data collection began in January 2021, and inclusion and exclusion criteria for the study are presented below.

Inclusion criteria

• Articles involving the use of Augmented Reality applications for educational purpose

• Studies published in English

• Scientific research from peer-reviewed journals and conferences

• Articles published between 2016 and 2020

Exclusion criteria

Research studies that were excluded from this review include theses, theoretical papers, reviews, and summaries that do not provide the entire articles. Additionally, studies that are "locked" and require a subscription or on-site payment for access were also excluded.

At the beginning of the data extraction process, a set of eight research questions was identified to guide the analysis:

RQ1. What is the target audience of the AR application?

RQ2. What educational areas or subjects are being targeted by the application?

RQ3. What type of assessment methods were utilized for the final solution?

RQ4. What were the outcomes achieved through the application of the proposed solution?

RQ5. What limitations or obstacles were noted in relation to the use of the application?

RQ6. What technological approaches were employed in the application's development?

RQ7. What are the primary channels for publishing research articles on AR educational interventions?

RQ8. How has the frequency of research studies on this topic changed over time?

The quality of the finally processed articles was assessed according to a series of criteria (Table 1). The CORE Conference Ranking (CORE Rankings Portal—Computing Research and Education, n.d.) and the Journal Citation Reports (JCR) (Ipscience-help.- thomsonreuters.com, 2022) were used for ranking conferences and journals accordingly. The maximum score for an article could be 10 points.

Table 1 Quality criteria

Results

Initially, a total of 3,416 articles were retrieved from the searches. A "clearing" stage was then conducted, consisting of several steps. First, duplicates and non-English articles were removed, resulting in 2731 articles. Second, titles and abstracts were screened, yielding 1363 potentially relevant studies. Third, articles that were not available, as well as reviews and theoretical papers not related to the topic, were eliminated. Finally, the studies that met the inclusion criteria were isolated, resulting in a total of 73 articles. The entire process is illustrated in Fig. 3. Figure 4 illustrates the quantity of Google searches conducted for the phrase “Augmented reality in education.”

Fig. 3

PRISMA flowchart

Fig. 4

Number of Google searches for the term "Augmented reality in education"

Table 2 illustrates the outcomes of the review process of the selected papers in terms of the technological methodology utilized and the characteristics of the assessment phase for the final solution. The analysis of the quality assurance results of the selected papers are presented in Table 4 (see Annex). According to the quality assurance criteria, 52.05% of the selected papers received a score above half of the total score, with a significant number of them (23.29%) scoring above 7.5. One paper achieved the maximum score, three papers scored 9.5, and one paper scored 9. Notably, 6.85% of the examined articles scored within the maximum 10% (total score = 9 to 10) of the rating scale.

Table 2 Review results

Most studies employed a combination of diverse methodologies to evaluate the final solution, with 83.56% of the studies employing a questionnaire, 16.44% employing observation techniques, 16.44% interviewing the participants, and only 4.11% utilizing focus groups for subjective assessment. Objective assessments were developed in only 6.85% of the studies (Andriyandi et al., 2020; Bauer et al., 2017; Karambakhsh et al., 2019; Mendes et al., 2020), with two studies utilizing automatic detection of correct results (Andriyandi et al., 2020; Karambakhsh et al., 2019), and one study using task completion time (Mendes et al., 2020). Approximately one third (31.51%) used achievement tests pre- and post-study to evaluate users' performance after the applied intervention. One study used an achievement test solely in the initial phase (Aladin et al., 2020), and another (Scaravetti & Doroszewski, 2019) only at the end. Concerning subjective assessment, each study employed various instruments depending on the application's characteristics, with custom-made questionnaires being used in almost two-thirds (61.90%) of the articles. The SUS was the most widely used well-known instrument (n = 7, 11.11%), followed by the IMMS (n = 4, 6.35%) and the QUIS (n = 3, 4.76%). The UES, TAM, SoASSES, QLIS, PEURA-E, NASA-TLX, MSAR, IMI, HARUS, and CLS were used in one study each.

Scientific journals were the primary source of publication (98.6%, n = 72), with only one paper (1.4%) presented at a conference. A significant proportion (38.82%) of the articles was published in computer-related sources. The publishing focus was almost equally divided between the education and learning field (18.82%) and engineering (16.47%). The health domain was slightly addressed, with only eight journals (9.41%), followed by sources representing the environment (2.35%). Procedia Computer Science dominated the publishing sector, with 16 articles (21.92%), followed by Computers in Human Behavior (6.85%), the International Journal of Emerging Technologies in Learning (5.48%), and the IOP Conference Series: Materials Science and Engineering (4.11%). The remaining articles (n = 45) were distributed across 39 journals. Notably, over one-third (n = 28, 38.36%) of the studies lacked a JCR ranking. More than half (52.1%) of the reviewed papers were published after 2019 (see Fig. 5).

Fig. 5

Frequency of papers per year

Table 3 provides a taxonomy for the classification and analysis of the studies included, which aids in the synthesis of findings and the detection of research patterns and gaps. This taxonomy can also function as a structured framework, assisting educators and researchers in categorizing, arranging, and comprehending the diverse aspects of applying AR technology in educational contexts. Tables 5, 6, and 7 subsequently (see Annex) presents the outcomes of the present study, built upon this taxonomy. The “Article id” in Tables 5, 6, and 7 is associated to the one presented in Table 2.

Table 3 Taxonomy of AR in education

Table 2 presents the technological approach followed by each project. Almost two thirds (68.49%) of the published studies exploited marker-based AR, superimposition-based was found in 9.59% of the articles, while 5.48% followed the location-based approach. As far as the devices are concerned, the majority uses a smartphone (n = 37, 50.68%) or a tablet (n = 35, 47.95%), while 13.70% (n = 10) exploits a head mounted display. Two studies (2.35%) used an interactive board, one a smart TV, and two a Kinect camera. Almost half of the papers (45.21%) worked on an Android operating system, while 28.77% used the iOS and only 9.59% the Windows one. A great percentage (32.88%) did not report the used operating platform. It has to be noticed, that a study may have use more than one of the mentioned devices or operating systems during the experiments. Regarding the used platform and tools for developing the final solution, Unity was the most common one (n = 35, 47.95%), followed by Vuforia (n = 31, 42.47%), Aurasma (n = 5, 6.85%), ARKit or formerly Metaio (n = 5, 6.85%), and Blippar (n = 2, 2.74%). A great percentage (n = 11, 15.07%) did not provide any details on the used platform and tools. As seen in Table 8 (see Annex), the topics covered by the reviewed articles were widely dispersed.

The majority of the reviewed studies (n = 31, 42.47%) focused on the university level, followed by 26.03% (n = 19) that targeted secondary education, 21.92% (n = 16) primary education, 6.85% (n = 5) early childhood education, 1.37% (n = 1) nursery school, and 1.37% (n = 1) health professionals. Special education was addressed in only six papers (8.22%), while 6.85% (n = 5) did not specify the target population.

For a comprehensive overview, Table 9 (see Annex) outlines the primary outcomes, limitations, and future steps of the reviewed studies concerning the utilized applications.

Discussion

The present study involved the analysis of both qualitative and quantitative data obtained from a collection of articles. The qualitative data obtained allowed for the identification of the decisions and actions taken by authors in designing and developing educational AR applications, as well as the extent to which these applications have been utilized. Notably, the study's analysis of educational AR applications was not restricted to any specific age group, subject area, or educational context. Rather, the study aimed to examine the full spectrum of educational AR applications, both within formal and informal education settings. Unlike prior investigations, the current study provides a comprehensive overview of research conducted between 2016 and 2020, exploring a diverse range of study designs and methodologies.

Based on the findings, it was discovered that almost all research studies pertaining to the topic at hand were published in scientific journals. Nonetheless, upon closer examination and analysis of the publications, it was noted that 25 of the studies that were published in journals were, in fact, conference proceedings that were later categorized as journals (e.g., Procedia Computer Science, Procedia CIRP, etc.) with no ranking, making up 38.89% of the total. Roughly 43.03% of the journals that were included in the review were of top-quality and ranked Q1. Collectively, 61.11% of the journals had a ranking score (Q1–Q4), and were thus considered as reputable sources. The wide variety of publishing sources (43 in total for the 73 papers examined) suggests that there is no specialized journal or conference dedicated to the area of interest. Additionally, it signifies that there are various ways in which AR can be employed in educational settings, ranging from simple applications such as labeling objects in a classroom to more intricate applications such as simulations. The following examples illustrate the diverse range of AR applications in education:

• Visualizing Concepts: AR can be used to visualize abstract concepts such as the solar system, anatomy, and physics. By using AR, learners can see these concepts in 3D, making it easier to understand and remember.

• Gamification: AR can be used to create interactive games that teach learners various skills such as problem-solving, critical thinking, and collaboration. These games can be used to make learning more fun and engaging.

• Virtual Field Trips: AR can be used to take learners on virtual field trips, allowing them to explore various places and learn about different cultures, history, and geography.

• Simulations: AR can be used to create simulations that allow learners to practice real-world scenarios and develop skills such as decision-making and problem-solving. For example, medical students can use AR to simulate surgeries and practice various procedures or to operate a microscope. Engineers also use AR to simulate experiments in mechanical engineering, electronics, electrical engineering and constructions.

The advent of emerging technologies and the development of low-cost devices and mobile phones with high computing power have created opportunities for innovative AR solutions in education. Researchers tend to prefer publishing their studies in journals, which are considered the most prestigious and impactful sources, even though it may take years to publish compared to only a few months in a conference.

The distribution of published articles per year (Fig. 5) can be attributed to the appearance of the first commercially available AR glasses in 2014 (Google Glasses), followed by the release of Microsoft's Hololens AR headset in 2016. As a result, a greater number of AR applications in retail emerged after 2017, and the AR industry has continued to develop as the cost of required devices has become more affordable. Based on the results, research related to the use of AR and mobile technology for educational purposes is expected to increase significantly in the coming years. According to a recent report by ResearchAndMarkets.com, the global market for Augmented Reality in education and training is projected to grow from 10.37 billion USD in 2022 to 68.71 billion USD in 2026 at a CAGR of 60.4% (Research & Markets, 2023).

In terms of the technological background of the provided solutions, the Android operating system dominated the market in the second quarter of 2018, accounting for 88% of all smartphone sales (Statista, 2023a, 2023b). This finding is consistent with the research results, which indicated that almost half of the studies developed the application for the Android system. This can be attributed in part to the fact that Android is widely adopted, particularly among children and teachers in most countries, who tend to own cheaper Android smartphones rather than iPhones. However, it is now becoming a trend for any commercial application to target both iOS and Android phones, which explains the 28.77% of apps developed for the iOS operating system. Only a small percentage of the studies (9.58%, n = 7) worked with Windows, indicating a strong trend towards mobile AR technologies. One third of the studies (32.88%) did not specify any operating system.

The augmented reality industry is experiencing significant growth, which can be attributed to the increasing number of mobile users who are adopting this technology. Snap Inc. predicts that by 2025, around 75% of the world's population will be active users of AR technology. In addition, Deloitte Digital x Snap Inc. has reported that 200 million users actively engage with augmented reality on Snapchat on a daily basis, primarily through mobile applications. This trend is supported by the modern citizen profile, which is characterized by continuous mobility, limited free time, and greater reliance on mobile phones than PCs or laptops. According to a Statcounter study ("Desktop vs mobile", 2023), 50.48% of web traffic comes from mobile devices. Furthermore, mobile learning is increasingly popular, as evidenced by various studies (Ferriman, 2023).

With respect to development platforms and tools, the market is dominated by Unity (47.95%) and Vuforia (42.47%). This can be attributed to the fact that Unity's AR Foundation is a cross-platform framework that allows developers to create AR experiences and then build cross-platform applications for both Android and iOS devices without additional effort. Additionally, Unity is a leading platform for creating real-time 3D content. Vuforia is a software development kit (SDK) that facilitates the creation of AR applications by enabling the addition of computer vision functionalities, which allow the application to recognize objects, images, and spaces.

Marker-based AR was utilized in 68.49% of the studies, as it is simple and effective in providing a seamless user experience. This technology involves using a camera to detect a specific visual marker, such as a QR code, and overlaying digital content onto the marker in real-time. This allows users to interact with the digital content in a more intuitive way, as they can physically move the marker and see the digital content move along with it. Furthermore, marker-based AR has been in use for longer than other forms of AR and has a more established user base. Its popularity has been further enhanced by many companies and brands integrating it into their marketing campaigns and products. Additionally, its accessibility is a contributing factor, as it requires less processing power and hardware compared to other forms of AR, making it easier for users to access and experience on their mobile devices. Markerless AR, which uses GPS and other location data to place virtual content in the real world based on the user's location, is gaining popularity, but only 2.74% of the examined studies used it. There are also markerless AR systems that use machine learning and computer vision to track and overlay digital content onto real-world objects without the need for markers. While marker-based AR is currently the most common type of AR, other forms of AR are rapidly evolving and gaining traction. Nonetheless, the review indicates that markerless AR applications are still in the early stages of development. As AI, machine learning, and computer vision techniques continue to advance, researchers will need to adopt them to improve AR applications in several ways:

• Object recognition and tracking: AI algorithms can be used to improve the accuracy of object recognition and tracking in AR applications. Machine learning can be used to train algorithms to recognize specific objects and track their movements in real-time. This can improve the stability of AR overlays and create a more immersive user experience.

• Content generation and personalization: Machine learning can be used to generate and personalize AR content for individual users. Algorithms can analyze user behavior and preferences to generate relevant and engaging content in real-time.

• Real-time language translation: AI-powered language translation can be integrated into AR applications to enable real-time translation of text and speech.

• Spatial mapping: Machine learning algorithms can be used to create detailed 3D maps of the user's environment. This can be used to improve the accuracy and stability of AR overlays and enable more sophisticated AR applications, such as indoor navigation.

• Predictive analytics: Machine learning algorithms can be used to provide users with contextual information based on their location, time of day, and other factors, while AI can predict user behavior. This can be used to create a more personalized and relevant AR experience.

The aforementioned aspects can potentially lead to new opportunities for innovation in the field of AR educational applications. These opportunities can be expanded by developing and utilizing virtual assistants and digital avatars within the educational context. Digital avatars and characters created by artificial intelligence can be designed to respond more naturally to users' behavior and emotions, thereby enhancing engagement and interactions and improving the user experience. AI-powered avatars can also facilitate realistic interactions, leading to more immersive and enjoyable learning experiences. Additionally, AI-powered platforms can be used to create interactive training sessions that provide stimulating and engaging learning experiences. For example, a virtual environment can simulate real-life job situations to aid in employee training. Likewise, AI-powered tools can create interactive experiences in which students can explore virtual objects and concepts in real-time.

Based on the research findings, the process of technology assessment is an arduous, challenging, and time-consuming task, but it is necessary in any research endeavor. However, there is no established gold standard for the subjective evaluation of Augmented Reality applications, which creates a vague landscape that forces most researchers (61.90%) to use custom-made scales. Consequently, this renders research results non-comparable. Moreover, many studies do not utilize reliable and valid instruments, making their findings questionable and not generalizable. Out of the examined pool, 35 cases used non-valid scales, 33 cases used non-reliable scales, and 33 cases used neither reliable nor valid scales. The System Usability Scale (SUS) was used seven times, the Intrinsic Motivation Measurement Scale (IMMS) four times, the Questionnaire for User Interaction Satisfaction (QUIS) three times, and all other scales (Unified Theory of Acceptance and Use of Technology – UTAUT, Extension Scale—UES, Technology Acceptance Model—TAM, Socially Adaptive Systems Evaluation Scale—SoASSES, Quality of Life Impact Scale—QLIS, Perceived Usability, and User Experience of Augmented Reality Environments—PEURA-E, National Aeronautics and Space Administration Task Load Index—NASA-TLX, Mixed Reality Simulator Sickness Assessment Questionnaire—MSAR, Intrinsic Motivation Inventory—IMI, Holistic Acceptance Readiness for Use Scale—HARUS, and Collaborative Learning Scale—CLS) were used only once each. In two studies (Conley et al., 2020; Saundarajan et al., 2020), even though the researchers tested the reliability of the questionnaires used, they did not assess their validity or use any established methodology to evaluate those questionnaires. Based on the presented results, the subjective satisfaction and assessment of AR solutions appear to be a daunting and challenging task. Therefore, there is a pressing need for the development of instruments that can capture the different aspects of a user's satisfaction (Koumpouros, 2016). In addition, it is essential to report users' experiences with the technologies used to enhance the completeness of research papers. Privacy protection and confidentiality, ethics approval and informed consent, and transparency of data collection and management are also essential. Legal and policy attention is required to ensure proper protection of user data and to prevent unwanted sharing of sensitive information with third parties (Bielecki, 2012). Conducting research involving children or other special categories (such as pupils with disabilities) requires great attention to the aforementioned issues and should follow all recent legislations and regulations, such as the General Data Protection Regulation (European Commission, 2012), Directive 95/46/EC (European Parliament, 1995), Directive 2002/58/EC (European Parliament, 2002), and Charter of Fundamental Right (European Parliament, 2000). The study also found that the number of end users participating in the assessment of the final solution is critical in obtaining valid results (Riihiaho, 2000). However, this remains a challenge, as only 19.18% of studies used 1 to 20 end users to evaluate the application, 20.55% used 21 to 40, 16.44% used 41 to 60, 9.59% used 61 to 80, and 21.92% used more than 80 end users. Only in four studies did both teachers and students evaluate the provided solution, although it is crucial for both parties to assess the solution used, particularly in the educational context, as they observe and assess the same thing from different perspectives.

In the examined projects, insufficient attention was given to primary and secondary education subjects, with only 21.92% and 26.03% of the efforts analyzed targeting these levels, respectively. Additionally, researchers should focus on subjects that are typically known for being information-intensive and requiring rote memory. The examined projects encountered several issues and limitations, including:

• small sample sizes,

• short evaluation phases,

• lack of generalizable results,

• need for end-user training,

• absence of control groups and random sampling,

• difficulty in determining if the solution has ultimately helped,

• considerations of technology-related factors (e.g., cost, size, weight, battery life, compatibility issues, limited field of view from the headset, difficulty in wearing the head-mounted displays, accuracy, internet connection, etc.),

• limited number of choices and scenarios offered to end users,

• subjective assessment difficulty,

• heterogeneity in the evaluation (e.g., different knowledge levels of the end users),

• poor quality of graphics,

• environmental factors affecting the quality of the application (e.g., light and sound),

• quick movements affecting the quality and accuracy of the provided solution,

• image and marker detection issues, and

• lack of examination of long-term retention of the studied subjects.

In terms of future steps, it is essential to obtain statistically accepted results, which requires a significant number of end users in any research effort. Additionally, it is crucial to carefully examine user subjective and objective satisfaction using existing valid and reliable scales that can capture users' satisfaction in an early research stage (Koumpouros, 2016). Researchers should aim to simulate an environment that closely resembles the real one to enable students to generalize and apply their acquired skills and knowledge easily. Other key findings from the examined studies include the need for:

• experiments with wider cohorts of participants and subjects,

• examination of different age groups and levels,

• use of smart glasses,

• integration of speech recognition techniques,

• examination of reproducibility of results,

• use of markerless techniques,

• enrichment of AR applications with more multimedia content,

• consideration of more factors during evaluation (e.g., collaboration and personal features),

• implementation of human avatars in AR experiences,

• integration of gesture recognition and brain activity detection,

• implementation of eye tracking techniques,

• use of smart glasses instead of tablets or smartphones, and

• further investigation of the relationship between learning gains, embodiment, and collaboration.

In addition, achieving an advanced Technology Readiness Level (TRL) (European Commission, 2014) is always desirable. An interdisciplinary team is considered to be extremely important in effectively meeting the needs of various end users, which can be supported by an iterative strategy of design, evaluation, and redesign (Nielsen, 1993). Usability testing and subjective evaluation are challenging but critical tasks in any research project (Koumpouros, 2016; Koumpouros et al., 2016). The user-friendliness of the provided solution is also a significant concern. Additionally, the involvement of behavioral sciences could greatly assist in the development of a successful project in the field with better adoption rates by end users (Spruijt-Metz et al., 2015).

Table 9 (see Annex) shows that AR technologies have been utilized in a variety of disciplines, educational levels, and target groups, including for supporting and enhancing social and communication skills in special education settings. Preliminary results suggest that AR may be beneficial for these target groups, although the limited number of participants, short intervention duration, and non-random selection of participants make generalization of the results challenging. Furthermore, the long-term retention of learning gains remains unclear. Nevertheless, students appear to enjoy using AR for learning and engaging with course material, and AR supports experiential learning, which emphasizes learning through experience, activity, and reflection. This approach to teaching can lead to increased engagement and motivation, improved retention and understanding, development of practical skills, and enhanced critical thinking and problem-solving abilities. In summary, AR has the potential to be a valuable tool for developing a range of skills and knowledge in learners.

An area of interest that warrants further investigation is the amount of time learners spend on each topic when utilizing augmented reality tools as opposed to conventional learning methods. This inquiry may yield valuable insights regarding the efficacy of AR-based

• (1) The ease with which students learn the material delivered through AR.

• (2) The amount of time required to learn the material when compared to conventional education.

• (3) Whether the use of AR enhances students' interest in the topic.

• (4) Whether students enjoy studying with AR more than they do with traditional methods.

• (5) Whether AR amplifies students' motivation to learn.

interventions. Researchers ought to explore the following five key issues when providing AR-based educational solutions:

It is evident that the aforementioned parameters require at least a control group in order to compare the outcomes of the intervention with those of conventional learning. Additionally, it is essential to consider the duration of the initial intervention and the retesting interval to assess the retention of learning gains. Finally, it is crucial to expand research into the realm of special education and other domains. For example, innovative IT interventions could greatly benefit individuals with autism spectrum disorders and students with intellectual disabilities (Koumpouros & Kafazis, 2019). Augmented reality could be proved valuable in minimizing attention deficit during training and improve learning for the specific target groups (Goharinejad et al., 2022; Nor Azlina & Kamarulzaman, 2020; Tosto et al., 2021).

As far as the educational advantages and benefits of AR in education are concerned, AR holds immense potential for enhancing educational outcomes across various educational levels and subject areas:

• Enhanced Engagement: AR creates highly interactive and engaging learning experiences. Learners are actively involved in the educational content, which can lead to increased motivation and interest in the subject matter.

• Visualization of Complex Concepts: AR enables the visualization of abstract and complex concepts, making them more tangible and understandable. Learners can explore 3D models of objects, organisms, and phenomena, facilitating deeper comprehension.

• Experiential Learning: AR supports experiential learning by allowing students to engage with virtual objects, conduct experiments, and simulate real-world scenarios. This hands-on approach enhances practical skills and problem-solving abilities.

• Gamification and Game-Based Learning: AR can be used to gamify educational content, turning lessons into interactive games. This approach fosters critical thinking, decision-making, and collaborative skills while making learning enjoyable.

• Virtual Field Trips: AR-based virtual field trips transport students to different places and historical eras, providing immersive cultural, historical, and geographical learning experiences.

• Simulation-Based Training: Medical and engineering students can benefit from AR simulations that allow them to practice surgeries, experiments, and procedures in a risk-free environment, leading to better skill development.

• Personalization of Learning: AR applications can personalize learning experiences based on individual student needs, adapting content and pacing to optimize comprehension and retention.

• Enhanced Accessibility: AR can assist learners with disabilities by providing tailored support, such as audio descriptions, text-to-speech functionality, and interactive adaptations to suit various learning styles.

To provide a more comprehensive understanding of AR in education, it is essential to connect it with related research areas:

• Gamification and Game-Based Learning: Drawing parallels between AR and gamification/game-based learning can shed light on how game elements, such as challenges and rewards, can be integrated into AR applications to enhance learning experiences.

• Virtual Reality (VR) in Education: Contrasting AR with VR can elucidate the strengths and weaknesses of both technologies in educational contexts, helping educators make informed decisions about their integration.

• Cross-Disciplinary Approaches: Collaborative research involving experts in AR, gamification, game-based learning, VR, and educational psychology can yield innovative approaches to educational technology, benefiting both learners and educators.

• Learning Outcomes and Age-Level Effects: Future studies should delve into the specific learning outcomes facilitated by AR applications in different age groups and educational settings. Understanding the nuanced impact of AR on various learner demographics is crucial.

• Subject-Specific Applications: Exploring subject-specific AR applications and their effectiveness can reveal how AR can be tailored to the unique requirements of diverse academic disciplines.

In conclusion, AR in education offers a myriad of educational advantages, including enhanced engagement, visualization of complex concepts, experiential learning, gamification, virtual field trips, and personalized learning. By linking AR research with related fields and investigating its impact on learning outcomes, age-level effects, and subject-specific applications, we can harness the full potential of AR technology to revolutionize education.

Summarizing, AR has positive indications and could significantly help the educational process of different levels and target groups. The innovation of various AR applications lies in the property of 3D visualization of objects—models. In this way, in the field of education, 3D visualization can be used for the in-depth understanding of phenomena by students, in whom the knowledge will be better impressed (Lamanauskas et al., 2007). Game-based learning, the Kinect camera or other similar tools and markerless AR should be further exploited in the future. Finally, it should be noted that in order to effectively achieve the design of an educational AR application, it is necessary to take into account the learning environment, the particularities of each student, the axioms of the psychology of the learner and of course all the theories that have been formulated for learning (Cuendet et al., 2013). In simpler terms, the use of AR applications in education makes learning experiential for learners and mainly aims to bridge the gap between the classroom and the external environment as well as to increase the ability to perceive reality on the part of students.

Research limitations

Our systematic literature review on AR in education, while comprehensive within its defined scope, has certain limitations that must be acknowledged. Firstly, the review was confined to articles published between 2016 and 2020, which may have excluded some recent developments in the field. Additionally, our focus on English-language publications introduces a potential bias, as valuable research in other languages may have been omitted. These limitations, though recognized, were necessary to streamline the study's scope and maintain a manageable dataset. We acknowledge the significance of incorporating more recent data, and already working to expand our research in future endeavors to encompass the latest developments, ensuring the timeliness and relevance of our findings. However, we believe that the period we examined is crucial, particularly due to the emergence of COVID-19, which significantly accelerated the proliferation of educational apps across various contexts. Hence, we consider this timeframe as a distinct era that warrants separate investigation.

Conclusion

The use of AR interventions shows promise for improving educational outcomes. However, to maximize its practical application, several aspects require further scrutiny. Drawing from an analysis of qualitative and quantitative data on educational AR applications, several recommendations for future research and implementation can be proposed. Firstly, there is a need to explore the impact of AR in special education, considering specific age groups, subject areas, and educational contexts. Additionally, studying the effectiveness of different methodologies and study designs in AR education is crucial. It is important to identify areas where AR can have the greatest impact and design targeted applications accordingly. Investigating the long-term effects of AR in education is essential, including how it influences learning outcomes, knowledge retention, and student engagement over an extended period. Understanding how AR can support students with diverse learning needs and disabilities and developing tailored AR applications for special education settings is also vital. Researchers should adopt appropriate methodologies for studying the impact of AR in education. This includes conducting comparative studies to evaluate the effectiveness of AR applications compared to traditional teaching methods or other educational technologies. Longitudinal studies should be conducted to examine the sustained impact of AR on learning outcomes and engagement by following students over an extended period. Mixed-methods research combining qualitative and quantitative approaches should be employed to gain a deeper understanding of the experiences and perceptions of students and educators using AR in educational settings, using interviews, observations, surveys, and performance assessments to gather comprehensive data. Integration strategies for incorporating AR into existing educational frameworks should be investigated to ensure seamless implementation. This involves exploring strategies for integrating AR into existing curriculum frameworks and enhancing traditional teaching methods and learning activities across various subjects. Providing teacher training and professional development programs to support educators in effectively integrating AR into their teaching practices is important. Additionally, exploring pedagogical approaches that leverage the unique affordances of AR can facilitate active learning, problem-solving, collaboration, and critical thinking skills development. The lack of specialized journals or conferences dedicated to educational AR suggests the need for a platform specifically focused on this area. The diverse range of AR applications in education, such as visualizing concepts, gamification, virtual field trips, and simulations, should be further explored and expanded. With the projected growth of the AR market in education, more research is expected in the coming years. Technological advancements should be leveraged, considering the dominance of the Android operating system, to develop applications that cater to both Android and iOS platforms. Furthermore, leveraging advancements in AI, machine learning, and computer vision can enhance object recognition and tracking, content generation and personalization, real-time language translation, spatial mapping, and predictive analytics in AR applications. Integrating virtual assistants, digital avatars, and AI-powered platforms can provide innovative and engaging learning experiences. Improving AR technology and applications can be achieved by investigating compatibility with different mobile devices and operating systems, exploring emerging AR technologies, and developing reliable evaluation instruments and methodologies to assess user experience and satisfaction. These recommendations aim to address research gaps, enhance the effectiveness of AR in education, and guide future developments and implementations in the field. By focusing on specific areas of investigation and considering the integration of AR within educational frameworks, researchers and practitioners can advance the understanding and application of AR in educational settings.

In conclusion, the utilization of AR interventions in education holds significant practical implications for enhancing teaching and learning processes. The adoption of AR has the potential to transform traditional educational approaches by offering interactive and personalized learning experiences. By incorporating AR technology, educators can engage students in immersive and dynamic learning environments, promoting their active participation and motivation. AR can facilitate the visualization of complex concepts, making abstract ideas more tangible and accessible. Moreover, AR applications can provide real-world simulations, virtual field trips, and gamified experiences, enabling students to explore and interact with subject matter in a way that traditional methods cannot replicate. These practical benefits of AR in education indicate its potential to revolutionize the learning landscape. However, it is important to acknowledge and address the limitations and challenges associated with AR interventions in education. Technical constraints, such as the need for compatible devices and stable connectivity, may hinder the widespread implementation of AR. Moreover, ethical considerations surrounding data privacy and security must be carefully addressed to ensure the responsible use of AR technology in educational settings. Additionally, potential barriers, such as the cost of AR devices and the need for appropriate training for educators, may pose challenges to the seamless integration of AR in classrooms. Understanding and mitigating these limitations and challenges are essential for effectively harnessing the benefits of AR interventions in education. While AR interventions offer tremendous potential to enhance education by promoting engagement, personalization, and interactive learning experiences, it is crucial to navigate the associated limitations and challenges in order to fully realize their practical benefits. By addressing these concerns and continuing to explore innovative ways to integrate AR into educational contexts, we can pave the way for a more immersive, effective, and inclusive educational landscape. Our systematic review highlights the substantial potential of AR in reshaping educational practices and outcomes. By harnessing the educational advantages of AR and forging connections with related research areas such as gamification, game-based learning, and virtual reality in education, educators and researchers can collaboratively pave the way for more engaging, interactive, and personalized learning experiences. As the educational landscape continues to evolve, embracing AR technology represents a promising avenue for enhancing the quality and effectiveness of education across diverse domains.

Annex

See Table 

Table 4 Quality assurance results

4,

Table 5 Taxonomy-based review results (I)

5,

Table 6 Taxonomy-based review results (II)

6,

Table 7 Taxonomy-based review results (III)

7,

Table 8 Domains covered

8, and

Table 9 Outcomes, limitations, future plans

9.