Ronghuai Huang, Ying Chen and Ahmed Tlili: typical applications and governance of educational robots

This research article by Ronghuai Huang, Ying Chen and Ahmed Tlili appeared in China Modern Educational Equipment, issue 1 of 2024, and principally discusses material from 2022–2023. Its full scope is retained, with editorial corrections and research-context notes below. The original journal English abstract is separately archived.

Ronghuai Huang, Ying Chen, Ahmed Tlili

Beijing Normal University

Abstract and background

Abstract: Rapid progress in generative AI, including generative algorithms and large language models, has renewed concern about integrating intelligent technologies into education. Educational robots, developed specifically for educational settings, show considerable potential to support teaching and learning but introduce ethical and safety risks, requiring further research, regulation and governance. Drawing on international developments in AI and education, this article identifies seven typical scenarios, including robotics in STEM education, programming for robots and social-robot-supported language learning. Using ChatGPT as an example and analysing social-media discourse, it discusses chatbots’ educational value and suggestions for school use.

It proposes a trustworthy AI framework for education incorporating robustness, legality, compliance and ethical acceptability, and governance to support high-quality, sustainable education.

Keywords: artificial intelligence; ChatGPT; educational robots; application scenarios; technology governance

A new generation of AI is a driving force for technological advances, industrial upgrading and productivity growth. ChatGPT, released in late 2022, brought renewed attention to AI [1], raised concern about machines replacing people and prompted educators to reconsider AI’s prospects and the future direction of education.

Educational robots are an important part of intelligent educational environments. Researchers and institutions worldwide have examined their educational applications. Rapid technological iteration and the expanding presence of robots make it necessary to track international AI-in-education developments, study typical scenarios, assess their value in schools and identify potential risks and threats. Such analysis should underpin a human-centred technology governance framework.

1. Advances in AI and international developments in AIED

Stanford University’s Institute for Human-Centered Artificial Intelligence (Stanford HAI) published the Artificial Intelligence Index Report 2023 [2], examining global developments and trends. The original article highlighted three issues: industry moving ahead of academia, growing support from policymakers and differences in public attitudes across countries. In the cited 2022 Ipsos survey, 78% of Chinese respondents agreed that products and services using AI had more benefits than drawbacks, compared with 35% of US respondents. These are sample agreement rates for that specific question. The Chinese sample had the highest agreement rate among the countries surveyed.

The article discusses generative algorithms, large language models (LLMs) and multimodality as three technological aspects of generative AI. Transformer, used in models underlying ChatGPT, is an attention-based architecture, rather than the name of a particular LLM. Self-attention relates positions within a sequence to compute representations [3]. The authors argue that chatbots can supply information not readily found by traditional dictionary lookup and illustrate contextual disambiguation with “the doors of the jaguar were open”, where jaguar refers to the car brand, and “the jaguar made a howling noise”, where it refers to the animal. These are the article’s illustrative examples.

The GPT-4 Technical Report [4] describes results on professional and academic benchmarks, including approximately top-decile performance on a simulated US bar examination. That result concerns GPT-4 on a particular test.

UNESCO’s April 2021 AI and Education: Guidance for Policy-makers [5] explains that AI applications rely on complex underlying technologies, including traditional AI, machine learning, artificial neural networks and deep learning. They enable applications increasingly offered as AI-as-a-service: natural language processing, speech recognition, image recognition and processing, autonomous agents, emotion detection, predictive data mining and artificial creativity.

Educational applications are discussed in three areas: educational provision and management; learning and assessment, including intelligent tutoring, AI-supported collaborative learning, educational virtual and augmented reality and exploratory environments; and teacher empowerment and improved teaching, including AI-driven forum monitoring, human–AI dual-teacher arrangements and AI teaching assistants.

As artificial intelligence in education (AIED) develops, educational robots and other AI applications remain contested. ChatGPT’s potential to help teachers prepare lessons and learners study independently also raises ethical and safety issues, making regulation and governance necessary. UNESCO’s November 2021 Recommendation on the Ethics of Artificial Intelligence [6] identifies four values: respecting, protecting and promoting human rights, fundamental freedoms and human dignity; environmental and ecosystem flourishing; diversity and inclusiveness; and peaceful, just and interconnected societies. It sets out ten principles and eleven policy action areas. As a global AI ethics framework, it reaffirms human-centred values and offers practical guidance on using AI for humanity’s benefit.

2. Typical applications of educational robots

Educational robots are developed for education to support learners’ analytical, creative and practical abilities. Their characteristics include suitability for teaching, openness, extensibility and friendly human–robot interaction. The article distinguishes robotics education from educational service robots [7]. Robotics education encompasses educational activities, courses, physical platforms, resources and educational philosophies. Educational service robots have teaching and learning functions and are used to assist teaching and management in areas such as STEM, languages and learning among particular groups.

Using high citation, representativeness and relevance to emerging developments as selection principles, the authors clustered 100 influential educational-robotics articles from the preceding decade. They identified seven research themes, treated here as typical application scenarios [8].

2.1 Robots in STEM education

Research on robots in STEM education covers four main areas: developing robotics-based STEM curricula, including learning situations, models, groups and outcomes; exploring how robots can develop computational thinking, metacognition and problem-solving; providing a useful, practical environment for STEM learning; and examining learners’ and teachers’ perceptions and experiences, including acceptance, trust, attitudes, confidence, interest and anxiety.

2.2 Programming for robots

Programming for robots can be studied from early childhood through adolescence and beyond. Early-childhood research emphasizes experiencing computational thinking; primary education, developing it; secondary education, creativity, practical skills, collaboration and problem-solving; and university education, greater classroom engagement.

2.3 Language learning supported by social robots

Social robots can support early language learning and adults’ multilingual learning. Research asks how different forms of social robot might develop children’s vocabulary, reading, speaking, grammar and sign-language skills. It also examines support for adult second-language learning and ways to extend classroom provision into opportunities available at different times and places.

2.4 Robots in classroom teaching

Classroom research investigates how robotics can change teaching, visualize abstract knowledge and develop systems thinking. In K–12 education, robots are commonly studied in foundational subjects to improve understanding of basic concepts and engagement. In higher education, they are often used in engineering courses across disciplines to address limitations of existing methods and support understanding of abstract concepts and systems design.

2.5 Robot-assisted special education

Robot-assisted special education research considers learners including autistic children and asks how robots might support communication and social interaction in particular tasks. The article identifies four uses and research aims: robots as tutors or instructional tools for skills such as music and drama; support for verbal and non-verbal social skills, imitation and relationships; exploration of effects on behavioural and emotional difficulties; and robots as a platform or mediator between children and human tutors, with assessment of intervention outcomes. These are research directions and potential applications, rather than a guarantee of benefit for every child.

2.6 Cognitive-training robots for older adults

Research on cognitive-training robots for older adults examines possible effects on memory, language and emotional state among people with cognitive impairment, alongside acceptance by older adults and stakeholders. Two strands are identified: empirical examination of effects on cognitive abilities and research on acceptance among older adults and relevant groups.

2.7 Personalized tutoring through conversational agents

Conversational robots are a form of intelligent tutoring system; conversational functions appear in language-learning software, subject-learning systems and social platforms. The authors anticipate attention to a new generation of chatbots based on natural-language generation, including AIGC and LLM technologies, ethics, inclusion, usability, user experience, academic integrity and human–AI collaboration.

3. Chatbots’ value in school education: the case of ChatGPT

ChatGPT is a chatbot based on generative pre-trained Transformer models that interacts through natural language and generates content in response to instructions. The original also considers the potential of multimodal content. The case and user-experience observations below concern the early research setting of late 2022 and early 2023. The authors see potential to improve learning efficiency alongside issues of academic integrity, privacy and data security. Debates at that time included two areas: insufficient resources, mismatches between chatbot technology and teaching tasks, regulation and liability, data security, neglected user needs, dialogue design and content quality; and schools blocking access on educational networks because of concerns about cheating.

Emergency restrictions could avoid some risks while also excluding convenience and opportunities [9].

The authors examined cases of ChatGPT use in education in China, the United States and other settings. Across three stages—social-platform observation, analysis of interview content and analysis of user experience—they investigated chatbot use, identified concerns and offered suggestions.

3.1 Public discourse and attitudes

Using ChatGPT with education, teaching or learning as search terms, the study collected 2330 tweets from 1530 Twitter users between 23 December 2022 and 6 January 2023 and applied sentiment and t-SNE analyses. The original identifies dispersed discussion topics and a lack of collective consensus, critical reflection and concern, and terms such as cheating and information security. The cited study categorized 5% of tweets as positive and 2.5% as negative; 92.5% were not assigned to either category. This is not a finding that 92.5% of users were neutral. The researchers also interpreted word clusters as showing some optimism while urging caution. That interpretation and the tweet sentiment categories should be understood separately and not generalized to the whole public.

3.2 Analysis of interview content

Interview participants generally regarded ChatGPT as significant for educational change but also expressed concern. The authors analysed the interviews through five lenses: educational transformation, response quality, usefulness, personality and emotion, and ethics.

Five observations emerged. First, participants thought natural-language access to varied knowledge could support learning, changes in teaching and educational digitalization. Second, most were satisfied with information quality and accuracy, while noting the early system’s knowledge time frame and erroneous answers; quality also depended on instructions. Third, most found dialogue fluent and enjoyable, but the text interface at the time did not recognize body language or gestures, leaving anthropomorphism and social interaction for further investigation. Fourth, topic-specific information and immediate feedback might reduce teachers’ workloads, but limited contextual understanding and contradictory answers led the authors to suggest better annotation and retrieval of earlier information to support research.

Fifth, participants and authors worried that overreliance could weaken active learning, creativity and critical thinking and increase plagiarism or cheating, alongside biased or false information and threats to personal data and privacy. These are historical interview observations and interpretations, not causal measurements of outcomes for all students.

3.3 Suggestions for educational use

On the basis of their analysis, the authors propose six guiding suggestions for educational use of ChatGPT:

1. Adopt with consideration rather than simply prohibit. Recognize positive and negative effects and investigate how chatbots can be used well in schools. Educational, safety and psychology specialists should collaborate, while education authorities develop guidance and policies for safe, effective use.
2. Update educational thinking. With ChatGPT integrated into teaching and learning, rote memorization aimed at knowledge acquisition and simple text processing should receive less emphasis. Teachers should develop logical and critical thinking. Oral debate, as in the ancient Stoic tradition, can assess thinking, the reasonableness and accuracy of arguments and persuasiveness. Human–AI teaching designs should include practical tasks and oral presentations and adopt new assessment methods.
3. Do not accept outputs uncritically. Because chatbots cannot guarantee accuracy, teachers should develop information discernment and guide learners in assessing and using generated texts.
4. Improve the ability to use these tools. Interaction affects information access and learning efficiency. Schools and teachers should develop thinking and questioning skills and teach learners how to use LLMs and data sources for learning.
5. Investigate more human-oriented robots. Most chatbots at the time were task-oriented and lacked an emotional dimension, limiting emotional value and immersion. The authors propose drawing on social interaction theory and studying effects on psychological well-being and academic work.
6. Give close attention to ethics. False and harmful chatbot information can hurt students. Design and development should be learner-centred, respect law and ethical norms, pursue algorithmic transparency and explainability, and consider social, emotional, cognitive and pedagogical factors.

4. Scenario standards and governance for robots in education

Widespread educational robotics can introduce new safety and ethical risks. Integrating AI into education in a regulated, orderly way is a shared concern. The authors call for objective analysis of three aspects of effects—needs, potential and challenges—and of integration with teaching and learning and trustworthiness. They propose a trustworthy AI framework for education with four basic criteria: robustness, legality, compliance and ethical acceptability. Governance for an intelligent society should use AI to support precise and sustained educational governance and high-quality, sustainable development [10].

The authors identify five core concerns for AI technology governance in education:

1. Develop teachers’ and learners’ digital literacy and skills in an orderly way. Improve their ability to use and control intelligent robots, especially to obtain, distinguish, verify, select and integrate information, so that they can learn and work well in human–AI collaboration.
2. Establish a system for admitting AI products to schools. Institutions should study classification and grading of products; education authorities should develop arrangements for review, oversight and accountability, including law, policy, institutional rules and management methods, to support healthy development.
3. Strengthen social experiments on AI in education. Use evidence to examine long-term effects across teaching and learning and identify patterns of educational change. Further develop experimental methodology and norms and progressively build an educational AI social-experiment system [11].
4. Clarify educational application scenarios. Analyse how real needs connect with technological provision, and relationships between learners’ cognitive characteristics and intelligent environments. Examine time, place, participants, resources, functions and subjects across product use, identify appropriate knowledge points and connect educational robots with subject content.
5. Protect teachers’ and learners’ privacy and data. The authors warned that incomplete laws and oversight could allow privacy intrusion, data leaks, discrimination and wider educational divides. They called for ethical, safe and controllable generative AI products suitable for teachers and learners, with privacy and data safeguards [12].

References

1. 吴砥,李环,陈旭.人工智能通用大模型教育应用影响探析[J].开放教育研究,2023,29(2):19-25,45.
2. MASLEJ N, FATTORINI L, BRYNJOLFSSON E, et al. Artificial Intelligence Index Report 2023[EB/OL].(2023-04-03)[2023-11-02].https://aiindex.stanford.edu/wp-content/uploads/2023/04/HAI_AI-Index-Report_2023.pdf.
3. VASWANI A,SHAZEER N,PARMAR N, et al. Attention Is All You Need[EB/OL].(2017-12-06)[2023-11-02].https://arxiv.org/pdf/1706.03762v5.pdf.
4. OpenAI. GPT-4 Technical Report[EB/OL].(2023-03-27)[2023-11-02].https://arxiv.org/pdf/2303.08774.pdf.
5. MIAO F, HOLMES W, HUANG R, et al. AI and education: guidance for policy-makers[EB/OL].(2021-04-09)[2023-11-02]. https://unesdoc.unesco.org/ark:/48223/pf0000376709.
6. UNESCO. Recommendation on the Ethics of Artificial Intelligence[EB/OL]. (2021-11-23)[2023-11-02].https://unesdoc.unesco.org/ark:/48223/pf0000381137_eng.
7. 黄荣怀,刘德建,徐晶晶等.教育机器人的发展现状与趋势[J].现代教育技术,2017,27(1):13-20.
8. HUANG R H, LIU D, CHEN Y, et al. Learning for All with AI? 100 Influential Academic Articles of Educational Robots[R]. Beijing: Smart Learning Institute of Beijing Normal University, 2023.
9. TLILI A, SHEHATA B, ADARKWAH M, et al. What if the devil is my guardian angel: ChatGPT as a case study of using chatbots in education[J]. Smart Learn. Environ. 2023,10:15. https://doi.org/10.1186/s40561-023-00237-x.
10. 夏立新,杨宗凯,黄荣怀,等.教育数字化与新时代教育变革(笔谈)[J].华中师范大学学报(人文社会科学版),2023,62(5):1-22.
11. 黄荣怀,王欢欢,张慕华,等.面向智能时代的教育社会实验研究[J].电化教育研究,2020,41(10):5-14.
12. 张绒.生成式人工智能技术对教育领域的影响:关于ChatGPT的专访[J].电化教育研究,2023,44(2):5-14.

Original journal English metadata

Typical Scenarios and Governance of Educational Robot Application

Huang Ronghuai, Chen Ying, Ahmed Tlili

Beijing Normal University

Abstract: The rapid development of generative AI has triggered new concerns about the integration of intelligent technologies into education. Educational robot shows great potential in assisting teaching and learning, but it also brings new ethical and safety risks, so the research, regulation and governance of educational robots should be strengthened. We sort out seven typical application scenarios, including educational robots in STEM education, robot-oriented programming, language learning supported by social robots, etc., and propose the values and application suggestions of chatbot in school education through social media discourse analysis about ChatGPT. We also argue that a credible AI framework in education should be constructed, and intelligent technology governance should be carried out to provide guarantee for the high quality and sustainable development of education.

Keywords: AI; ChatGPT; educational robot; application scenarios; technological governance

Editorial and research-context notes

1. Journal record: 2024(1), pages 1–4 and 8; DOI 10.13492/j.cnki.cmee.2024.01.006. BNU’s author profile and SLI bibliography corroborate it. Reference [9] corrects the surname to Adarkwah and formats the volume and article number as 10, 15. BNU author profile; SLI bibliography; Cited 2023 study
2. UNESCO’s Recommendation has four values, ten principles and eleven policy areas. The original count of seven is corrected, as is the reference to natural ecosystems rather than AI systems. UNESCO Recommendation
3. The 78% and 35% figures concern a specific 2022 Ipsos benefits-versus-drawbacks question, not present-day overall support. Stanford 2023 report
4. Transformer is an architecture. GPT-4’s simulated-test results and image/text input with text output do not describe every ChatGPT version. Transformer paper; GPT-4 technical report
5. The sample comprised 2330 tweets from 1530 users. The 92.5% uncategorized-tweet share is not a neutral-user share; word-cluster interpretation is not a public census. Cited 2023 study
6. Interview observations, the 2021 knowledge boundary, the text interface and school restrictions belong to early research settings, not current product specifications or universal causal findings. Cited 2023 study; SLI full text
7. Special-education and older-adult sections retain all uses and research aims as potential applications, not individual or universal clinical efficacy claims. SLI full text
8. The four criteria, six suggestions and five governance concerns are the authors’ analysis and proposals. Wording on not trusting outputs has been clarified; proposals are not presented as implemented standards. SLI full text

Journal and republication credit

Original publication: China Modern Educational Equipment, 2024(1), pp.1–4,8. Authors: Ronghuai Huang, Ying Chen and Ahmed Tlili, Beijing Normal University. The original requests attribution on republication; author and journal credits are retained. DOI: 10.13492/j.cnki.cmee.2024.01.006

Related research and resources

Continue reading: the 2023 educational-robotics handbook report, the 2024 Forum on Robotics and Artificial Intelligence in Education, and special and inclusive education resources.