Using Social Robotics to Identify Educational Behavior: A Survey

Abstract

The advancement of social robots in recent years has opened a promising avenue for providing users with more accessible and personalized attention. These robots have been integrated into various aspects of human life, particularly in activities geared toward students, such as entertainment, education, and companionship, with the assistance of artificial intelligence (AI). AI plays a crucial role in enhancing these experiences by enabling social and educational robots to interact and adapt intelligently to their environment. In social robotics, AI is used to develop systems capable of understanding human emotions and responding to them, thereby facilitating interaction and collaboration between humans and robots in social settings. This article aims to present a survey of the use of robots in education, highlighting the degree of integration of social robots in this field worldwide. It also explores the robotic technologies applied according to the students’ educational level. This study provides an overview of the technical literature in social robotics and behavior recognition systems applied to education at various educational levels, especially in recent years. Additionally, it reviews the range of social robots in the market involved in these activities. The objects of study, techniques, and tools used, as well as the resources and results, are described to offer a view of the current state of the reviewed areas and to contribute to future research.

1. Introduction

Artificial intelligence (AI) is defined as the ability of machines to perform tasks that normally require human intelligence, such as learning, decision making, and pattern recognition [1]. Today, AI encompasses multiple subfields, including machine learning, computer vision, fuzzy logic, data science, natural language processing, heuristics, and intelligent agents [2]. These advances have facilitated the development of AI-based robots capable of using complex algorithms to adapt to different situations and perform tasks autonomously and efficiently.

Within the field of robotics, social robots have emerged as a specialized class that interacts intuitively and socially competent with humans [3]. These robots not only execute preprogrammed tasks, but also understand and respond to the complexities of human interactions, using AI to process information and make decisions in real time. This makes them valuable tools in a number of fields, including healthcare, elderly care, and, more recently, education [4]. In education, AI-based social robots represent an innovative approach to enhance student interaction and learning. These robots are designed to encourage active participation, personalize learning, and support students’ socioemotional development by adapting to their individual needs through data analysis and tailoring their responses [5]. Unlike traditional approaches that focus on mere academic instruction, educational social robots integrate the affective dimension, providing emotional support that can be crucial for student motivation and engagement [6]. Likewise, the implementation of social robots in the educational setting marks a fascinating technological breakthrough that redefines the interaction between students and technology [7]. These robots, designed with the ability to interact intuitively and adaptively, offer an innovative approach to enhance learning and engagement in the classroom [8]. Their presence not only adds a playful dimension to the educational environment but also provides the opportunity to personalize instruction and facilitate more individualized learning experiences [9]. By combining AI and social interaction, robots in the classroom are not only presented as educational tools but also as learning companions capable of stimulating curiosity, fostering collaboration, and contributing to development.

In recent years, psychology, medicine, therapy, and other areas have combined with advancements in engineering and robotics to offer an alternative in the treatment of children and adults. However, the last decade has seen an increase in the incorporation of robots that enhance and help improve education [10].

The use of social robots in early childhood education is crucial for several reasons. First, the education of young people is critical to preserve and promote the culture and traditions of each nation [11]. Second, specialized educational services for children with disabilities facilitate their integration into society, ensuring equal opportunities and fighting against social exclusion [12]. Moreover, in exceptional situations, such as the COVID-19 pandemic, social bots can provide effective distance education for children [13]. In this context, this paper focuses specifically on the analysis of AI-based social robots in education. In contrast to a general review on robots in education, our aim is to highlight how these specific robots are being implemented to enrich the educational environment, with an emphasis on their effectiveness and the challenges they present. In particular, we review recent studies (2019–2023) to assess the impact of these robots at different educational levels, with research questions focusing on the relationship between social robots and education, the AI techniques used, the educational levels where these robots have been implemented, and their specific goals.

The following research questions were analyzed:

-

RQ1: What is the relationship between social robots and education?

-

RQ2: What are the AI techniques used by social robots in education?

-

RQ3: What are the educational levels where social robots have been implemented and with what objectives?

Finally, the contributions of this review are as follows:

-

We discuss the different relationships between social robots and education.

-

We present the most commonly used AI techniques in education and draw conclusions about the suitability of social robots implemented at different educational levels with the aim of defining a user guide for future research, considering the following data:

-

The age range of the target group.

-

The role of the robot in the interaction, i.e., whether the interaction occurred alone with the robot or with the cooperation of persons, i.e., an instructor, a tutor, researchers, other children for group sessions, or a family member.

-

The type of interaction (games, workshops, homework, lessons, or other) between the target group and the robot.

-

The name of the robot that was used.

-

The type of robot (humanoid, non-humanoid).

-

The educational level at which the robot was implemented.

The above data were extracted from each study and organized for later focus.

The first two sections set the stage for an enriching discussion in Section 3, focusing on identifying the main challenges pointed out by researchers or deduced from this study. Finally, Section 4 closes the analysis, synthesizing the answers to the issues raised and proposing the next steps needed to advance the integration of social robots in the educational setting toward more advanced levels of preparedness.

2. Materials and Methods

2.1. Search Strategy

A search for full-text and peer-reviewed articles was conducted in multiple electronic databases. In addition, the search was limited to research articles in each of the databases, and searches were carried out considering different keywords,

Table 1. Data sources.

Database	Article Type	Keywords
ScienceDirect	Research article	Education AND affective computing
ResearchGate		Affective computing AND prediction depression
IEEE Xplore		Social robots AND education
Google Scholar		Artificial intelligence AND education

Databases and research keywords used.

contains the main search terms used.

2.2. Selection Criteria

The electronic databases that collect information on the most important publications in the scientific field were consulted; this search was carried out from September 2022 to November 2023. Mendeley was used to create a database with the publications under study. For this, specific filters were applied according to the inclusion criteria capable of extracting all the data needed for this article. In addition, the Preferred Reporting Items for Systematic reviews and Meta Analyses (PRISMA) methodology was used to describe aspects of a systematic search of the publications that have been developed on this topic. Inclusion criteria were developed, including the type of article, since only research articles that were within the established year and the key words that were related to the objective of the research were integrated.

Table 2. Study inclusion and exclusion criteria.

Criteria	Included	Excluded
Keywords	“education” “affective computing”, “social robotics” “classification”, depression “robotics”, “students” “humanoids”, “robots”	“Autonomous mobile robots (AMR)”, “automated guided vehicles (AGV)”, “articulated robots”, “hybrids”
Type of article	Research	Review
Year of publication	2019–2023	Another year
Application area	Education, engineering, AI	Another area
Solution	Affective computing	NA

Criteria used for the research and the data included and excluded.

also shows the exclusion criteria, where articles outside the year range were excluded, as well as articles that only talk about autonomous robots and review articles.

2.3. Data Extraction

A literature review was conducted between 2019 and 2023, considering the most recent works in social robotics implemented in education. The search was based on the PRISMA methodology for systematic reviews/meta-analysis, and a flowchart was created, showing the initial review process to the end. As shown in Figure 1, when the term “social robots in education” was searched, 1400 results were found. Virtually all these papers were references to scientific articles, books, book chapters or book reviews, but when we added the inclusion criteria starting with the year, from 2019 to 2023, the number of articles decreased to 778. Meanwhile, when we added the type of article, it decreased to fewer than 300. When filtering by subject areas, such as “informatics”, the number of eligible articles decreased to 169. The number of articles decreased to fewer than 100 when adding the area of “engineering” and decreased even more when selecting by publication title.

2.4. Statistical Results

The aforementioned methodology was used to collect an adequate amount of published research within the scope of this study. The statistical analysis of all the publications collected is detailed here.

2.4.1. Demographics

Figure 2 shows that most of the studies were conducted in Europe (46.15%, N = 18), followed by Asia (33.33%, N = 13), America (17.94%, N = 7), and Oceania (2.56%, N = 1). No studies were found in Africa.

A detailed map of the countries where the studies were conducted is illustrated in Figure 3. The majority of the studies presented were conducted in the Netherlands (17.9%, N = 7), followed by China (12.8%, N = 5), the United States (10.2%, N = 4), Greece (10.2%, N = 4), Germany (8.3%, N = 3), Iran (8. 3%, N = 3), Colombia (2.5%, N = 1), Ecuador (2.5%, N = 1), Brazil (2.5%, N = 1), Japan (2.5%, N = 1), the United Arab Emirates (2.5%, N = 1), Malaysia (2.5%, N = 1), Qatar (2. 5%, N = 1), Kuwait (2.5%, N = 1), the United Kingdom (2.5%, N = 1), Switzerland (2.5%, N = 1), Poland (2.5%, N = 1), Austria (2.5%, N = 1), and Australia (2.5%, N = 1).

2.4.2. Timeline

Figure 4 shows the number of publications per year in the evaluated period. It can be seen that 2019 and 2020 had the highest number of publications, with 11 articles each, followed by 2022 with 8 publications, 2021 with 5, and 2023 with 4. On average, about 7.8 articles were published per year during this period.

It is important to note that in 2020, there was a notable increase in research activity, probably due to the conditions generated by the COVID-19 pandemic. Although a slight decrease in the number of publications was observed in the last three years evaluated (2021–2023), this could be due to multiple factors, such as restrictions in R&D activities during the pandemic [14] and the fact that the total number of publications was not fully reflected in the indexed databases by the time this work was addressed.

2.5. Analysis of the Information

This subsection shows the selection of articles that responded to each of the research questions.

2.5.1. Social Robots

According to the study by [15], social robots can be divided into four groups according to the complexity of the scenario in which the interaction takes place, including socially educational, social interface, socially responsive, and sociable robots. Socially educational robots may possess a different appearance to encourage the learning process and the ability to show emotions through anthropomorphic and non-anthropomorphic characteristics. They manage to encourage people to interact with technology, such as robotic pets in video games [16]. In contrast to [17], anthropomorphism and human–robot interaction are features that play a very important role in robot design because anthropomorphism refers to a social phenomenon that describes the tendency of people to attribute real qualities to objects and other nonreal artifacts.

Social interface robots use a human-like form of communication to facilitate interaction with people. This is the case of robots with some social intelligence; to convey messages in an appropriate way, they rely on gestures and expressions [18]. Receptive robots, like the previous robots, benefit from interactions with people. Falling into this type are robots that learn through either motion or speech training by means of simple auditory or verbal features of spoken language, such as voice features and speech cues, which influence the quality of interaction with a social robot [19]. Finally, the social robots on which this review article is focused are those that, through their own goals and internal motivations, try to engage humans in a social way, not only for the benefit of humans but also to benefit each other, e.g., to instruct each other. These robots try to adapt to people in social and cognitive terms to relate to them [20]. At present, social robots are still in the research stage; however, they are beginning to appear in certain circumstances, such as robots intended for rehabilitation, companionship, and assistance and care for the sick and elderly with different degrees of dependence. These robots are indispensable auxiliaries of care both at home and in homes for the elderly [21].

2.5.2. Social Robot and Educational Applications

Social robots have been implemented in the educational field with various applications. Among the objectives that have been studied in children, we found that the NAO robot developed by the company Softbank Robotics has been used in various educational institutions, including special education institutions to encourage participation and interactive learning in elementary school children.

Table 3. Social robots implemented in education in 2019.

Ref.	Country	Robot Name	Robot Type	Robot Role	Interaction	Objectives	Participant Total (Female, Male)	Age Range (Years)	Educational Level
[22]	Australia	Cozmo	Non- humanoid	Main interaction	Test	Give the participants a sense of independence to communicate with each other	6 (1, 5)	24–42	Special education
[23]	Iran	NAO/NIMA (Persian name, Persa)	Humanoid	Main interaction	A pre-test, immediate post-test to measure the English vocabulary learning gains	Investigate the effects that using a humanoid robot as a teacher’s assistant can have on English vocabulary learning development and retention among individuals with Down syndrome	10 (4, 6)	24–36	Special education
[24]	Japan	NAO	Humanoid	Main interaction	Test	Teaching English to native European children, as well as teaching Dutch and German to immigrants	80 (-, -)	18–24	University
[25]	United Arab Emirates	NAO	Humanoid	Main interaction	Test	Measure and understand the perception of humanoid robots by students in the Arab world, specifically in the United Arab Emirates (UAE)	44 (-, -)	10–12	Elementary School
[26]	United Kingdom	Kaspar	Humanoid	Main interaction	Games	Stimulate interaction and collaboration between children while teaching the robot	30 (19, 11)	7–8	Special education
[27]	Netherlands	NAO	Humanoid	Main interaction	Imitation	Examine the contribution of the robot ZORA to achieving therapeutic and educational goals in rehabilitation and special education for children with severe physical disabilities	33 (11, 22)	2–8	Special education
[28]	China	Xiao	Non- humanoid	Main interaction, parent present	Communication with parents and private conversation	Evaluate an innovative approach in rural areas	156 (-, -)	6–12	Elementary School
[29]	China	NAO	Humanoid	Main interaction	Games	Teach complex social rules	40 (5, 35)	5–8	Special education
[30]	China	NAO	Humanoid	Main interaction	Training session	Assess narrative skills	26 (3, 23)	4–6	Special education
[31]	China	NAO	Humanoid	Main interaction	Training session	Examine robot intervention in comparison to human intervention	23 (3, 20)	6–12	Elementary School
[32]	Iran	NAO	Humanoid	Main interaction	Music-based scenario	Teach fundamentals of music, improve social/cognitive skills	4 (0, 4)	6	Elementary School

During 2019, the implementation of social robots in educational settings highlighted the prominence of the NAO robot over Xiao. NAO stood out for its ability to interact naturally and effectively with students, offering an enriching and personalized educational experience. Its humanoid design and advanced artificial intelligence capabilities made it a versatile and highly effective educational companion. Although Xiao was also present in some educational settings, it was the NAO robot’s renowned performance that positioned it as a leader on the educational scene in 2019, demonstrating its value as an innovative and effective educational tool.

presents a synthesis of relevant quantitative data from the studies examined in this section during 2019. It is important to note that in this and the following 4 tables, the symbol “-” indicates that the corresponding information is not available in the cited reference.

Since the NAO social robot has the ability to interact emotionally and adapt to diverse learning styles, it has been widely implemented in specialized educational programs to provide support to children with special needs. This implementation has contributed to the development of social and cognitive skills in these children, making it one of the most widely used social robots in education in 2019, especially in the field of special education. However, some of these implementations face limitations, such as a small number of participants, the absence of a control group, the severity of autism problems, and a lack of access to valid tools to accurately measure the children’s behavior.

Based on the implementation trends in education, it is possible to observe in

Table 4. Social robots implemented in education in 2020.

Ref.	Country	Robot Name	Robot Type	Robot Role	Interaction	Objectives	Participant Total (Female, Male)	Age Range (Years)	Educational Level
[33]	Switzerland	Pepper	Humanoid	Main interaction	Test	Predict student behavioral intention	462 (163, 299)	19–21	University
[34]	Iran	Nima and Arash	Humanoid	Main interaction	Teaching scenarios devised to enhance interaction between the kids and the robot	Assess impact of robot on interrelated areas of religious and ethical features in education in an Islamic society	42 (42, -)	8–9	Elementary School
[35]	Malaysia	PvBOT	Non- humanoid	Main interaction, teacher present	Games	Examine the effectiveness of robots as learning tools	8 (4, 4)	10–13	Special education
[36]	Netherlands	NAO	Humanoid	Main interaction, individual or group sessions, teacher, and parents present	Social interaction and exercises	Obtain attitudes toward robots	118 (61, 56)	9–12	Elementary School
[37]	Netherlands	NAO	Humanoid	Main interaction	Social interaction and exercises	Improve performance in STEM subjects	86 (41, 45)	8–10	Elementary School
[38]	Poland	Fanuc LR Mate 200 iD Sputnik NAO	Non- humanoid and Humanoid	Main interaction, researcher present	Social interaction and exercises	Determine attitudes toward robots	195 (110, 85)	19–58	High School
[39]	Austria	Spiderino	Non- humanoid	Main interac-tion, teacher present	Workshops	Increase interest in STEM subjects	69 (-, -)	14–18	University
[40]	United States	Tega	Non-humanoid	Main interaction	Games	Compare roles (tutor/mentee/peer) and impact of learning	64 (-, -)	5–7	Elementary School
[41]	Colombia	ONO	Humanoid	Main interaction	Therapy sessions	Assess autism factors for diagnosis	45 (32, 13)	6–11	Special education
[42]	Ecuador	NAR	Humanoid	Main interaction	Games	Strengthen acquired knowledge	25 (-, -)	3–5	Kindergarten
[43]	Qatar	NAO	Humanoid	Main interaction, teacher present	Distrust and deception games	Test and analyze impact of robot	15 (-, -)	7–11	Special education

During 2020, the implementation of social robots in education experienced a significant boost as educational institutions were forced to adapt to new teaching models due to the COVID-19 pandemic. These robots not only offered a technological solution to maintain student–teacher interaction but also provided emotional and motivational support at a time of uncertainty and change. Their ability to personalize the learning experience and facilitate active student participation was critical to maintaining educational continuity in a virtual environment. As technology advanced, social bots became valuable allies in overcoming educational challenges during 2020.

the implementations of robots in education in 2020, when, once again, the NAO robot was the most used, highlighting that during this year it was not only used at the primary level but also at the secondary, middle, and high school levels.

NAO’s friendly and expressive interaction makes it suitable for fostering the development of social and emotional skills in students, fundamental aspects of a well-rounded education. Given the technological nature of the robot, Figure 5 shows NAO was used to stimulate students’ interest in science and technology, encouraging them to explore STEM-related careers from an early age, which is why, in 2020, it was the most widely implemented social robot in education.

Designed to interact naturally with people, Jibo and Reeti were made to communicate effectively with humans, with expressive face and facial recognition capabilities. In the educational field, they were implemented at higher and primary levels in 2021, as shown in

Table 5. Social robots implemented in education in 2021.

Ref.	Country	Robot Name	Robot type	Robot Role	Interaction	Objectives	Participant Total (Female, Male)	Age Range (Years)	Educational Level
[44]	Greece	Pepper	Humanoid	Main interaction, researcher present	Games and teaching coins	Enhance short-term and long-term memory	3 (-, -)	6–12	Special education
[45]	Netherlands	NAO	Humanoid	Main interaction	Dutch and Turkish receptive vocabulary tests	Investigate whether providing L1 translations during an L2 vocabulary training by a social robot facilitated L2 word learning	67 (34, 33)	4–6	Kindergarten
[46]	Greece	NAO	Humanoid	Main interaction	Test	Assess the performance of social robots in place of university professors in the field of engineering	89 (83, 6)	19–28	University
[47]	Germany	Reeti	Non- humanoid	Main interaction teacher present	Exercises	Assess the learning environment	80 (70, 10)	18–38	University
[48]	United States	Jibo	Non- humanoid	Main interaction	Games	Stimulate creativity	79 (40, 39)	5–10	Elementary School

During 2021, the implementation of social robots such as Reeti and Jibo in education was highlighted in various educational settings. Recognized for their ability to interact naturally and emotionally with students, they became valuable tools to facilitate personalized learning and active participation in the classroom. Their innovative design and advanced artificial intelligence capabilities allowed for more dynamic and stimulating participation in educational activities. These robots also stood out for their versatility, providing support in both formal teaching and individualized tutoring. The presence of social robots, such as Reeti and Jibo, marked an important step toward the successful integration of technology in education, enriching the learning experience for students and promoting a more interactive and collaborative educational environment.

.

Due to Reeti’s ability to mimic human facial expressions and behaviors, during 2021, it was implemented in activities aimed at improving the social and emotional skills of students in therapeutic and educational environments to support children with special needs, contributing to the development of social, emotional, and cognitive skills. Jibo was implemented as an interactive tool to enhance the learning experience in distance learning environments, providing a more personalized and emotional interaction, which is why they were the most used robots in education in this year.

The year 2022 witnessed an exciting expansion in the implementations of robots such as NAO, Zenbo, and Childbot in education. These robots not only demonstrated their ability to enhance teaching and learning but also paved the way for future innovations in the integration of technology in education; see

Table 6. Social robots implemented in education in 2022.

Ref.	Country	Robot Name	Robot Type	Robot Role	Interaction	Objectives	Participant Total (Female, Male)	Age Range	Educational Level
[49]	Netherlands	NAO	Humanoid	Main interaction, teacher, and parents present	Test	Provide education on sleep hygiene in an interactive and playful way through a social robot	28 (14, 14)	8–12	Elementary School
[50]	Greece	NAO	Humanoid	Main interaction, researcher present	Test	Intervention in improving the learning performance of elementary school children with specific learning disorders	134 (45, 89)	8–10	Elementary School
[51]	Netherlands	SAMBusddy	Non- humanoid	Main interaction, researcher present	Exploring and reducing children’s stress levels using a social robot	Build trustworthy interactions with children and lower children’s stress levels	115 (59, 56)	3–6	Kindergarten
[52]	United States	Pepper, Milo and NAO	Humanoid	Main interaction	Robot musical theater	Motivate audience to participate in actions to prevent climate change	14 (-, -)	Mixed age	Elementary School
[53]	Brazil	NAO and Zenbo	Humanoid and non-humanoid	Main interaction, researcher present	Social interaction and exercises	Music education	20 (9, 11)	9–11	Elementary School
[54]	United States	Childbot	Humanoid	Main interaction, individual or group sessions, teacher present	Games	Child–robot interaction	21 (-, -)	Mixed age	Elementary School
[8]	Germany	NAO	Humanoid	Main interaction	Social interaction and exercises	Adaptive robotic tutor in a university environment	58 (52, 6)	M = 19	University
[55]	Netherlands	NAO	Humanoid	Main interaction, researcher present	Social interaction and exercises	Assist in learning another language	63 (24, 39)	4–6	Elementary School

NAO was able to engage in educational activities ranging from language teaching to programming. ASUS Zenbo became an ideal learning companion for students of all ages, and Childbot provided an immersive and enriching educational experience for preschoolers. In some studies represented in the above table, the ages of the participants were not a specific range; instead, they opted to identify a mean age or a mixed age, referring to various ages.

. With their ability to inspire curiosity, facilitate engagement, and personalize the learning experience, these robots continue to play a crucial role in transforming the educational landscape toward a more technology-centric and interactive future.

NAO has been adapted to facilitate student participation in virtual environments, offering online interaction and remote educational activities. It also stimulates students’ interest in science and technology, encouraging them to explore STEM-related careers from an early age, and supporting subject-specific teaching by providing hands-on demonstrations and activities that reinforce key concepts in science, mathematics, and other areas. It enhances language teaching and language skills development in students of different ages and was the most widely used robot in 2022.

It is important to emphasize that in some of the studies, the robots were not implemented in educational institutions, were pilot tests, or were hardware with supporting web platforms. Furthermore, some did not have quantitative results that validated the reliability of the information; therefore, they were excluded from this research.

Table 7. Social robots implemented in education in 2023.

Ref.	Country	Robot Name	Robot Type	Robot Role	Interaction	Objectives	Participant Total (Female, Male)	Age Range	Educational Level
[56]	Greece	NAO	Humanoid	Main interaction	Imitation games	Evaluate the added pedagogical value of the humanoid Softbank NAO 6 and its impact on students’ cognitive and SEL outcomes	115 (-, -)	(-, -)	Special education and elementary School
[57]	China	HUMANE	Non- humanoid	Main interaction	Test	Compare the learning effectiveness of robot-based intervention (RBI) with that of content-matched human-based intervention (HBI)	38 (7, 31)	6–9	Elementary School
[58]	Kuwait	NAO	Humanoid	Main interaction	Social interaction and exercises	Combine the humanoid NAO robot with a mobile application to enhance the educational experiences of children with autism spectrum disorder (ASD)	12 (-, -)	3–6	Kindergarten
[59]	Germany	NAO	Humanoid	Main interaction	Games	Enhance in-person social learning experiences through robot-supported collaborative learning facilitated by an NAO social robot	48 (-, -)	18–24	University

NAO robots are used in educational institutions for teaching science, technology, engineering, and mathematics (STEM). Their ability to program themselves and perform complex tasks makes them valuable tools for students of all ages.

presents a summary of the main implementations of the NAO humanoid robot in 2023, highlighting its applications due to its versatility and interaction capabilities. It was used at various academic levels, which are described below.

Germany, as a leader in technological innovation, uses NAO in advanced research projects in robotics and artificial intelligence, exploring areas such as human–robot interaction and social robotics. Meanwhile, in China, NAO is widely used in language learning, especially to teach English to students, thanks to its ability to hold conversations and perform linguistic activities. Likewise, in Greece, NAO supports inclusive education by facilitating the participation of students with diverse abilities in integrated learning activities.

2.5.3. Recognition Systems in Social Robots

Facial expressions and body language, a fundamental way of communicating human emotions, find applications in human–machine interactions, surveillance, security, and deception detection, among others [60]. The face conveys most of the emotions experienced by humans through facial muscles; furthermore, it can often convey what words cannot [61]. This is why several studies have been conducted on the recognition, interpretation, and simulation of human behavior. Therefore, a large part of the studies and projects carried out in this context have used data from videos or sensors with devices such as Kinect [62]. The recognition systems that have been implemented in social robots in the educational field with various applications are presented in

Table 8. Recognition systems in robotics.

Robot	Robotic Work	AI Method	Educational Level	Ref.
Kismet	- Object detection - Facial recognition - Emotion recognition	- Artificial neural networks (ANNs) and fuzzy logic models - Prosodic analysis algorithms (intonation, pitch, and rhythm patterns) - Acoustic models and machine learning algorithms - Natural language processing (NLP) algorithms - Reinforcement learning and emotional regulation models	University	[63]
KASPAR	- Multiple-point touch detection - Object recognition - Gesture recognition	- Convolutional neural networks (CNNs) - Machine learning algorithms for emotion classification - NLP algorithms - Supervised learning - Computer vision algorithms	Kindergarten, special education, and elementary school
ASIMO	- Facial recognition - Human tracking - Touch detection - Sound discrimination	- Natural language processing (NLP) algorithms - Convolutional neural networks (CNNs) - Computer vision algorithms - Simultaneous localization and mapping algorithms (SLAMs) - Model-based predictive control (MPC)	University
iCub	- Human detection - Object tracking - Touch detection - Sound localization	- Convolutional neural networks (CNNs) - Natural language processing (NLP) algorithms - Simultaneous localization and mapping algorithms (SLAMs) - Artificial neural network (ANN)	University
Albert Einstein	- Face tracking - Facial recognition - Facial expression recognition - Object tracking - Speech recognition	- Convolutional neural networks (CNNs) - Natural language processing (NLP) algorithms - Machine learning algorithms for emotion classification	University
NAO Pepper	- Emotion recognition - Autonomous movement - Simultaneous localization and mapping (SLAM) - Object tracking and recognition - Facial recognition - Voice recognition - Face tracking - Facial recognition - Facial expression recognition - Object tracking - Speech recognition	- Support vector regression (SVR) - Deep neural networks (DNNs) - Support vector machines (SVMs)	Kindergarten, elementary school, middle school, high school, and university	[64]
Candide-3	- Facial recognition - Facial expression recognition - Facial recognition - Voice recognition - Face tracking	- Fuzzy weighted average algorithm	Elementary school	[65]
Candide-3	- Object tracking - Speech recognition - Facial recognition - Facial expression recognition - Voice recognition - Face tracking	- Artificial neural network (ANN) - Hidden Markov model (HMM) - Summarized vector machines (SVMs) - K-nearest neighbors (KNN) - AdaBoost algorithm	Middle school	[66]
WACNN	- Object recognition - Facial recognition - Facial expression recognition - Emotion recognition	- Convolution neural networks (CNNs) - Hybrid genetic algorithm (HGA)	Elementary school, middle school, and high school	[67]

Social robots used in education often incorporate AI techniques and algorithms for facial recognition, a crucial capability for facilitating interaction and personalization in educational environments. The table above describes some of the main techniques and algorithms used for facial recognition in social robots.

. Among the objectives that have been studied in children, we found that the main robot used was the NAO robot, whose main use is as a tutor and teacher.

3. Discussion

In the previous sections, we have highlighted the growing interest in the integration of social robots in educational contexts, especially between 2019 and 2023, emphasizing how artificial intelligence (AI) techniques have facilitated these developments. However, it is essential to delve deeper into how these technologies have been successfully implemented and what advantages they offer compared to other types of educational robots.

The incorporation of AI in social robots, such as the NAO, has proven to be instrumental in enabling richer and more personalized interactions with students. For example, through the use of convolutional neural networks (CNNs), robots can perform facial recognition and emotional expression detection, thus allowing them to identify students and adjust their interactions in real time according to each student’s emotional state, resulting in a much more personalized teaching experience [68]. A prominent study by [69] showed that students who interacted with AI-equipped robots showed higher engagement and retention of information compared to those using non-AI robots. This underscores the ability of AI to provide an educational experience that goes beyond the mere repetition of information by being able to adapt and respond to individual student needs.

In addition, the use of AI allows robots to handle complex situations and adapt to changing student behaviors. For example, the use of machine learning techniques has enabled robots to learn from previous interactions and adjust their responses to meet the individual needs of each student [70]. In a study by [71], researchers found that AI-equipped robots significantly improved the motivation and learning ability of students with autism spectrum disorder (ASD). The ability of AI to identify patterns in student responses and adjust robot behavior has resulted in a more effective and personalized form of teaching, which would be difficult to achieve with traditional algorithms.

Another notable case is the use of the social robot Pepper, in which AI has been incorporated to enhance its natural language processing and emotion detection capabilities [72]. In a recent study, Pepper was able to conduct prolonged interactions with elementary school students, adjusting its tone and language according to the children’s level of understanding and emotional feedback. The results showed improved engagement and reduced boredom during learning sessions, suggesting that AI not only enables more personalized interactions but is also more effective in terms of information retention and comprehension.

Despite these advances, the implementation of AI in social bots presents several challenges. One of the main ones is the computational burden associated with real-time information processing. However, with advances in processing power and the availability of cloud-based solutions, many of these challenges are being addressed. Robots such as NAO and Pepper have been able to effectively handle the computational load required for tasks such as natural language processing and emotion detection, making them more viable for use in educational settings [73].

However, there remains a lack of standardization in intervention methods and the need to customize algorithms for each educational setting. A study by [74] showed that although social bots are effective in certain contexts, their performance varies significantly depending on the setting and student population. This requires algorithms to continuously adapt, which presents a challenge for large-scale implementation. Differences in student behaviors and emotions, especially at younger ages, mean that AI solutions must be developed that are flexible and able to adapt to different educational needs.

In conclusion, the integration of AI into social bots has shown significant potential to improve educational experiences. However, to achieve impact on a larger scale, additional research is needed to address current limitations and establish more standardized methodologies. The future of AI implementation in education will depend on our ability to balance technical complexity with adaptability and accessibility in different learning contexts.

4. Conclusions

This article presents a survey of the application of social robots in education during the period of 2019–2023, with a particular focus on the artificial intelligence techniques that support their functions. Throughout the analysis, it is shown that, although there are diverse applications and approaches, the key to success lies in the proper integration of AI technologies that enable adaptive and personalized interaction.

The most common functions, such as student identification, emotion recognition, and object detection, have been implemented using advanced AI techniques, such as convolutional neural networks, deep learning, and machine learning. These technologies have proven particularly effective in special education, where social robots, such as NAO, have played a crucial role in teaching children with ASD, improving their social and communication skills.

Despite these advances, this study has also identified important challenges that need to be addressed in future research. Among these challenges are the need to customize interaction algorithms according to the specific characteristics of the students and the lack of consistency in the intervention methodologies used in the reviewed studies. These problems limit the ability to generalize the results and apply the findings to different educational contexts.

We recommend that future research focuses on developing more adaptive and personalized AI algorithms and standardizing intervention methodologies to ensure the validity of the results. In addition, it is crucial to explore new applications of social bots in areas such as distance education and support for marginalized communities, where AI has the potential to transform teaching and learning in significant ways.

The integration of advanced artificial intelligence technologies in social robots not only enhances human–robot interaction but also offers new opportunities to create more inclusive and effective learning experiences, thus fulfilling the goal of facilitating human–robot collaboration in educational settings.