A Grand Challenge for Humanity:
General Artificial Companions Hypothesis

This article is extracted from parts of the paper:
Chou, C. Y., Chan, T. W., Chen, Z. H., Liao, C. Y., Shih, J. L., Wu, Y. T., Chang, B., Yeh, Y. C., Cheng, H., & Hung. H. C. (in press). Defining AI companions: a research agenda—from artificial companions for learning to general artificial companions for Global Harwell. Research and Practice in Technology Enhanced Learning.

Abstract

Intelligent, caring, and patient, artificial companions——whether virtual or robotic, are designed to provide around-the-clock companionship to human beings. They are sophisticated AI entities aiming to establish positive and meaningful companionships with humans through human-like interactions and to fulfill the overarching goals of these companionships. As AI technologies continue to advance, these human-AI companionships incorporate ethical considerations, foster engagement, and advance human holistic development. Thus, in addition to transforming education, artificial companions can contribute to individual wellbeing as well as broader humanitarian objectives such as peace, diversity, equity, inclusivity, and environmental sustainability. At the outset of the massive exploration of artificial companions' potential by research communities, it is fundamentally important to go beyond mere contributions to economic growth and efficiency. This includes tackling the most challenging global problems or human crises of our time, fostering understanding and cooperation among researchers and practitioners in different fields, and paving the way toward a future world marked by Global 'Harwell' (a portmanteau of 'harmony' and 'wellbeing'). To advocate for and demonstrate the pivotal role of artificial companions in achieving these objectives, we propose the General Artificial Companions Hypothesis. This hypothesis posits that artificial companions have the potential to assist humans in pursuing Global Harwell, directing AI development efforts toward this grand challenge, while simultaneously alleviating concerns over AI's potential negative societal impacts. However, all these endeavors must begin with humans, nurturing individuals who embrace Global Harwell as their core value and life goal through education. Only these individuals can design genuine artificial companions for the betterment of humanity.

Keywords: artificial companions, AI companions, artificial general intelligence, general artificial companions, artificial intelligence in education, intelligent computer assisted learning

Introduction

Arguably, the most significant implication for education arising from the launch of ChatGPT in 2022 is its role as an artificial companion (or AI companion; we shall use these terms interchangeably) for learning, specifically as an AI learning companion. Clearly, the significance of artificial companions spans various domains, including health, entertainment, work, sports, and nearly all facets of our lives.

With the advent of generative AI technologies, such as ChatGPT, the public can now envision the Turing Test being passed—a scenario where a human interrogator cannot distinguish between a human and a machine in a text-only conversation. Rapid advancements in AI, the internet, the metaverse, and technologies like augmented reality, advanced robotics, the Internet of Things, quantum computing, blockchain and others are poised to expedite the development of artificial companions, both virtual or robotic. These companions are expected to demonstrate human-like intellectual, emotional, social, and even value-based behaviors. In various subtle ways, the power of such digital resemblance may make it challenging for us to discern whether the artificial companion we are interacting with is a real person or an artificial entity. Recently, artificial companions have garnered significant interest and, with the advancement of AI, have demonstrated considerable promise. An artificial companion could be a patient and caring 'superhuman' available around the clock, potentially approaching or even surpassing human experts in various domains. For example, Deep Blue defeated a chess master in 1997 and AlphaGo achieved victories against multiple human Go masters starting in 2015. ChatGPT has the capabilities to check syntax, rewrite sentences, translate into different languages, generate content and answer various questions. However, it sometimes provides inaccurate information or even fabricated 'hallucinations,' leaving room for improvement (Alkaissi & McFarlane, 2023; Rasmussen et al., 2023; Tlili et al., 2023). In the future, we will find numerous artificial companions around us.

AI tools, AI assistants, and AI companions

In adopting AI across various domains such as education, healthcare, entertainment, finance, retail, and transportation, we observe the spectrum from viewing AI merely as a tool to considering AI as an assistant and even as a companion (see Figure 1). The spectrum covers a wider range, and the three roles often become blurred. Nevertheless, we can roughly differentiate them by their complexity or depth of interaction with us. As a tool, AI acts as a piece of equipment that we manually operate to accomplish a task. As an assistant, AI aids us by performing certain tasks or providing intricate support to complete a task. Clearly, generative AI can now or soon serve as an AI assistant in many domains, such as developing programming code, creating animations, and providing legal advice.

The epoch of artificial companions is approaching. As a companion, the interactions between AI and the human user involve mixed initiative: sometimes the human initiates a dialogue or an activity to which the AI responds, and vice versa. Beyond the specific purpose of fulfilling the current task, offering sophisticated responses requires consideration of the human user's intellectual, emotional, social, and value dimensions. This requires the companion to acquire a 'memory' of the user's past experiences and beliefs; in other words, it necessitates a user model. We should also be cognizant that a companion's response, when incorporated with such nuanced and thoughtful consideration, will subtly yet significantly influence the human user's perspectives across various dimensions.

Figure 1. Spectrum of human-AI relationships Figure 1. Spectrum of human-AI relationships

AI companions and human-AI companionship

What is a companion, and what defines companionship? First, companionship is a type of positive dyadic social relationship, which involves the following six features: emotional support, joint activities, dependability and faith, mutual respect, dialogue and exchange, and enjoying togetherness. Emotional support refers to offering empathy, understanding, and comfort. Joint activities mean engaging in common interests or activities together. Dependability and faith imply being dependable and trustworthy. Mutual respect indicates valuing each other's opinions, feelings, and needs. Dialogue and exchange involve open and honest discussion about thoughts and feelings. Enjoying togetherness suggests finding pleasure in spending time together.

Now, can we consider our mother, spouse, child, friend, teacher, tutor, fellow classmate, or even our pet as companions? If yes, assuming a human or creature is our companion, our companionship can be defined by our 'relationship,' such as a parent-child relationship, and 'interactions,' which include both past and prospective interactions.

We may consider the parent-child relationship as an example to further explore the concept of 'relationship.' This relationship involves two actors, each assuming a specific 'role': one as the parent and the other as the child. The parent is responsible for everything related to the child's growth, while the child learns from the parent and follows their instructions. Thus, we can see that the roles within a relationship define the overarching goal of their interactions. By 'overarching goal', we mean a goal composed of many, or even numerous, sub-goals to achieve the overarching goal. 'Interactions' here are viewed as a series of action-and-reaction pairs between the two actors, continuing throughout their relationship, from past to present and into the future, all intended to fulfill their overarching goal. Perhaps we can define:

companionship of two actors = (relationship, overarching goal, interactions)

Involving just two actors, a human and an artificial companion that serve as companions to each other, this formulation represents the simplest and most basic form of companionship. Naturally, this definition can be extended to encompass companionship among a group of actors. Indeed, in our everyday lives, we engage in both dyadic and group companionships. In the future, we will have many artificial companions around us, some of them representing delegates of our human friends, relatives, or people we may or may not be acquainted with (as in the Metaverse environment).

In this article, we focus on dyadic companionship, where one actor is an AI learning companion and the other is a human learner. The overarching goals are mainly educational.

Nevertheless, with a bit of imagination, some challenging questions may arise from the definition of companionship. Suppose the group of actors includes all humans and all existing artificial companions in the world. Given the extremely complex relationships and interactions among all the actors, what would be the overarching goal of this global companionship? If we focus solely on education, then what would be the overall educational goal for global learning companionship?

The origin of artificial learning companions

Inspired by the potential applications of machine learning in education, Chan and Baskin (1988) proposed the concept of learning companions. This concept envisions the computer acting as a student's learning companion, akin to the Chinese proverb 'Studying with the Prince,' to enhance learning through such companionship (see Figure 2). They cited Vygotsky's 'zone of proximal development'—the gap between children's actual cognitive development and their potential development with the guidance and scaffolding of adults or more capable peers—as a theoretical basis for artificial learning companions (ALCs).

Figure 2. Learning with the Prince (Chan & Baskin, 1988) Figure 2. Learning with the Prince (Chan & Baskin, 1988)

Chan and Baskin (1988) described two approaches to designing ALCs: simulation and machine learning. "In the simulation approach, the companion's performance is controlled by the system in order to adapt to the student. A simulated companion may have deliberate sub-optimal behavior in order to match skill with the student. On the other hand, in the machine learning approach, the growing knowledge of the companion, which results in improved performance, is acquired through machine learning techniques. In this approach, the student's learning is more likely to benefit from observing how the companion learns. The companion explains his learning process, his discoveries and his hypotheses derived from what he has learned1 (p. 199).

They pointed out that the potential benefits to student learning are supported by the zone of proximal development, which refers to the distance between actual developmental level and potential developmental level. The former pertains to students' competence in solving problems independently, while the latter relates to their competence in solving problems with adult guidance or in collaboration with more capable peers (Vygotsky, 1978). This concept of the zone of proximal development explains why princes in ancient Chinese empires often learned more effectively when studying with their learning companions (Chan & Baskin, 1988).

Chan and Baskin also highlighted that the paradigm of the ALCs spans a wide spectrum of design possibilities, influenced by the potential variations in the number and identities of both human students and ALCs. For example, “it is possible to have no teacher involved…Then the student may observe how the companion solves the problems and improves performance. In this way, the student learns how to learn by teaching the learning companion…To the other extreme, it is possible to have multiple (virtual) teachers with different persona. For example, there may be a patient teacher and a demanding teacher. The student may choose one of them to adaptively respond to his own learning style…which means more than one ALC with different knowledge level or personas involved in the learning environment. (p.199)”

Subsequently, more ALCs were designed and implemented with diverse roles supporting various learning activities. For example, Distributed WEST facilitates collaborative and/or competitive learning among students using two connected computers with other students or ALCs (Chan et al., 1992). Reciprocal Tutoring Systems enable students to participate in reciprocal tutoring activities where ALCs act as peer tutors, tutees or competitors (Chan & Chou, 1997; Chou et al., 2002). EduAgents provide students heterogeneous ALCs, including two strong ALCs and two weak ALCs (Hietala & Niemirepo, 1998).

Chou et al. (2003) defined ALCs as “computer-simulated characters with human-like characteristics that plays a non-authoritative role in a social learning environment.” These human-like characteristics encompass competence, emotions, beliefs, behaviors, appearance, personality, and more, which can be expressed or displayed in text, images, animations, multimedia, virtual reality, augmented reality, natural language processing, speech recognition and synthesis, and image recognition or through robots. They categorized ALCs by roles, such as competitors, collaborators, tutees, peer tutors, troublemakers, critics, or clones, engaging students in various social learning activities like collaborative learning, reciprocal tutoring, and learning by teaching. It is noteworthy that this article broadens the definition to include not only non-authoritative roles but also authoritative roles like parents, tutors, and experts who possess greater knowledge and social status.

Numerous studies have demonstrated that ALCs enhance students' learning performance (e.g. Atkinson, 2002; Castro-Alonso et al., 2021; Graesser et al., 2005; Kim et al., 2007; Kulik & Fletcher, 2016; Lester et al., 1997; Li et al., 2022; Lusk & Atkinson, 2007; Ma et al., 2014; Moreno et al., 2001; VanLehn, 2011), motivation (e.g. Baylor, 2009; Kim et al., 2007; Lester et al., 1997; Liu et al., 2024; Moreno et al., 2001; Schroeder & Adesope, 2014), and self-regulation (e.g. Harley et al., 2018; Karaoğlan Yılmaz et al., 2018).

ALCs have been developed with various appearances or representations. Animated pedagogical agents provide human-computer interaction through face-to-face mixed-initiative dialogue (e.g. Blair et al., 2007; Clarebout et al., 2002; Davis, 2018; Gulz & Kakke, 2006; Johnson et al., 2000; Kim & Baylor, 2006; Kim et al., 2006; Lester & Stone, 1997; Sikström et al., 2022; Woo, 2009). Animal companions provide pet-like companionship to students (e.g. Chen, 2012; Chen et al., 2007, 2011). Learning companion robots feature a robotic appearance combined with human-like expressions, including voice, facial expressions, gestures, and motions (e.g. Cagiltay et al., 2022; Ho et al., 2021; Hsieh et al., 2015; Hsu et al., 2007; Kory & Breazeal, 2014; Lfelebuegu, 2024; Liu et al., 2024; Lubold et al., 2021; Michaelis & Mutlu, 2018; Spitale et al., 2022; Wang et al., 2009; Wei et al., 2011; Zinina et al., 2023). Chatbot companions utilize natural language processing techniques to communicate with students in natural language (e.g. Huang et al., 2022; Huang et al., 2008; Hwang & Chang, 2023; Kim et al., 2022; Kuhail et al., 2023; Liu et al., 2024; Okonkwo & Ade-Ibijola, 2021; Ruan et al., 2019; Skjuve et al., 2021).

Global Harwell

The world is on the brink of peril (Chan, 2023). We are confronting unparalleled challenges, including millions of deaths caused by COVID-19, climate change, resource depletion, environmental pollution, wealth disparities, and worries over AI's negative impact on humanity. Furthermore, escalating global conflicts intensify fears of a potential nuclear apocalypse. Resolving such human crises fundamentally hinges on education playing a critical role, and ALCs must extend their objectives beyond supporting learning.

'Global Harwell'—where the term 'Harwell' combines 'harmony' and 'wellbeing'—on one hand, represents what humankind aspires to (Chan, 2023; Global Harwell Group, 2024a). On the other hand, it refers to a set of overarching values or a collective ethos that defines societal norms, ethics, and goals on a global scale. This suggests that the pursuit of Global Harwell should be established as the core and basic values for humanity, indicating that the very purpose of human knowledge and technology is to support the achievement of Global Harwell. Also, since the education of our next generations determines the destiny of our future world, Global Harwell should be adopted as a shared, fundamental global educational goal.

From learning to the pursuit of Global Harwell in the Seamless AI World

In the Seamless AI World (SAIW) (Global Harwell Group, 2024b), a world seamlessly connected to, filled with and empowered by AI technology, our world is 'shrinking' or becoming 'smaller' in the sense that students who are physically far apart can easily and closely interact in their native languages through AI-supported instant interpretation. The exchange of ideas among students from different places and cultures is expected to significantly increase.

Obviously, in SAIW, there will be numerous artificial companions around. Challenging yet urgent, artificial companions play a crucial role not only in assisting with learning but also in pursuing the Global Harwell goal. Thus, if researchers and educators across the globe collaborate, and if Global Harwell is adopted as a global educational goal, we can anticipate a profound transformation in education.

However, involving many aspects of learning, such endeavor brings forth a set of challenges in SAIW:

  1. Where and when to learn in SAIW?
  2. With whom should the student learn, such as human teachers, classmates, or artificial companions in SAIW?
  3. Will there be a theory that informs the design of artificial companions for learning, as well as for attaining the Global Harwell goal in SAIW?
  4. What knowledge and skills should students acquire about Global Harwell, and how can such knowledge and skills be integrated into their subject matters in SAIW?

Lifelong artificial companions

If we predict there will be numerous artificial companions around us, then the advent of lifelong artificial companions for individuals is within reach. Chan, et al. (2001, p. 159) described a dream: "…… every student in the future has a set of lifelong learning companions. For a youngster, s/he may choose a set of 'animal' companions, and in her/his school years, s/he will be like a leading character in a Disney cartoon movie always surrounded by the companions. Every animal companion can assume a different role. For example, a dalmatian is a collaborator. Mushu dragon is a peer tutor. Monkey king is a troublemaker, who challenges the student from time to time. Tamagochi (a once popular electronic chick that can be incubated by kids) can be taught by the student to play against the other Tamagochies raised by other students in some network learning games. In other words, we can 'disneyficate' the learning environment.". For the creation of a lifelong learning companion, Chou, Chan, and Lin (2003, p. 266) further elaborated that "Such a learning companion is like a friend that stays with the student from childhood to adulthood. The companion stores the student's lifelong portfolio. Educational agents constructed in many domains can be combined into a lifelong learning companion, or the student's lifelong portfolio can at least be exchanged among the educational agents."

Lifelong artificial companions are designed to accompany individuals throughout their lives, offering continuous support and interaction. They adapt to changing needs and preferences, providing personalized assistance, information, and even emotional companionship across various life stages— for example, during years in school, working life, and retirement. These companions aim to build long-term relationships, fulfilling roles such as personal assistants or caregivers, while respecting ethical considerations such as privacy and autonomy. By integrating deeply into daily routines, lifelong artificial companions enhance wellbeing, illustrating a future where AI technology plays a supportive and integral role in enhancing human life from childhood through old age.

Quest for Artificial General Intelligence with ethical, emotional, and sociability awareness

AI has made remarkable strides in recent years, outperforming humans in specific tasks such as medical diagnostics, natural language processing, financial trading, industrial automation, playing games, and analyzing data. However, when it comes to overall cognitive abilities and understanding the world as humans do, AI has not yet surpassed human-level intelligence. This distinction between narrow AI and Artificial General Intelligence (AGI) (or Strong AI2) is crucial: while narrow AI excels in specific domains, AGI aims to achieve human-like cognitive abilities across a wide range of tasks, akin to how humans think and learn. This includes capabilities such as reasoning, problem-solving, abstract thinking, and adapting to new situations.

While AI has transformative potential, its integration into society must be approached cautiously, with careful consideration of ethical implications and societal impact. This includes considerations of fairness, transparency, accountability, and the ability to make decisions that respect human autonomy and dignity.

In terms of values and attitudes, AI systems can exhibit behaviors that mimic values or attitudes, following ethical guidelines or rules that govern their interactions with humans, such as simulating empathy or politeness. However, they still lack the deeper emotional or value-based comprehension that humans possess. The challenge lies in developing AI systems that not only perform tasks effectively but also align with human values, ethical principles, and factors of emotion and sociability when interacting with humans in decision-making. Current AI systems still lack the intuitive understanding of complex human values and social nuances, as well as emotional intelligence, which are essential for navigating ethical dilemmas and interpersonal interactions in various contexts.

As AI technology continues to evolve, the quest for AGI and AI systems that are ethically, emotionally, and sociability aware remains ongoing. Achieving these goals will require interdisciplinary collaboration and continued advancements in AI research and development. The pursuit of Global Harwell, in such a situation, may serve as a beacon illuminating the path forward.

General Artificial Companions Hypothesis

Our proposition of the General Artificial Companions Hypothesis (GACH) sets forth a visionary goal for AI, envisioning its role in advancing the future world towards the concept of Global Harwell. This hypothesis posits that AI, when developed with a focus on human-AI companionship and societal betterment, can profoundly impact our global community. By fostering AI technologies that prioritize ethical companionship and contribute positively to societal progress, we strive not only to enhance human harmony and wellbeing but also to cultivate a harmonious coexistence between technology and humanity. Ultimately, our vision is to harness AI's potential to create a future where innovation, empathy, and sustainable development define our collective journey towards Global Harwell.

General Artificial Companions (GACs) refer to lifelong artificial companions for individual humans, powered by either AGI technology or technology nearing AGI capabilities. Additionally, all the 'overarching goals' of human-AI companionships, defined by their roles at various stages of the human actor's life, must include Global Harwell as a primary subgoal.

The General Artificial Companions Hypothesis (GACH) posits that in SAIW, the support provided by human individuals' GACs will contribute 50% or more to the development of Global Harwell as their primary values in their lives, as evidenced by their daily behaviors.

To elucidate, the GACH suggests that in a future scenario where Strong AI and GACs are prevalent and deeply integrated into society (SAIW), GACs will provide substantial and noteworthy support to human individuals. This support is expected not only to influence but to shape the development of Global Harwell as their primary values in lives.

According to the GACH, the contribution of these GACs to the formation of these values is substantial, potentially constituting 50% or more of the factors that influence how individuals prioritize and manifest these values in their daily lives. This influence is observable through the behaviors and interactions of individuals with their GACs.

In essence, the GACH implies that as artificial companions become more integrated and sophisticated, they will play a crucial role in shaping human values and behaviors within the framework of Global Harwell, reflecting a symbiotic relationship where technology not only assists but actively participates in the cultural and ethical evolution of society.

It is also pertinent, important, and imperative to highlight that as more people adopt Global Harwell as their life values, fewer people, out of individual or institutional selfishness, or even unconsciously, will use AI for detrimental purposes against humankind. This will, at the very least, impose more constraints on those who seek to develop AI-empowered entities harmful to society (Liu, 2024).

Clearly, there are numerous issues to clarify and questions to answer regarding this hypothesis. For example, to what extent can we qualify our world as a SAIW? Can we limit the scope of a world to a well-defined community during an experiment, such as within a family, a group of families, a school including students' parents, company staff, occupational unions, and so on? If this hypothesis can be proven within such a limited scope, could it serve as a model and be rapidly disseminated worldwide? At what stage of life is it easier to develop Global Harwell as people's values which affects their life goals? All these and other issues and questions need to be addressed by researchers and practitioners adopting AI in any aspect of our human lives.

In sum, the proposal for the GACH aims to offer AI researchers and practitioners a goal for creating an everlasting and peaceful world with this technology—ensuring that our future generations thrive on this planet and continue to grow in a Global Harwell future. The world, however, cannot passively await the advent of Global Harwell; our future generations' education cannot hinge on the eventual arrival of GACs. We must act now, collectively and collaboratively!

Undoubtedly, Global Harwell represents the paramount goal of AI technology for humanity!

Conclusion

ALCs hold the potential to revolutionize education by providing personalized, adaptive learning experiences that cater to individual student needs. The approach, integrating emergent AI technologies, learning theories, and educational strategies, sustains the development of effective and meaningful use in education. More importantly, in the SAIW, everything changes. Aligning with the broader goal of artificial companions to foster Global Harwell as a shared value worldwide not only promotes academic success but also enables students to achieve harmony and wellbeing. In responding to these demands—for education, for Global Harwell, and for all aspects of life—GACs address challenging issues related to ethical and well-rounded development. While championing the pivotal role of artificial companions in cultivating a Global Harwell era in a seamless AI world, the GACH emphatically highlights the crucial need for further investigation. Designing AI means designing the future—a future that ensures our future generations continue to survive and flourish in this century and beyond.

Note

Note 1: Learning through learning machine techniques is challenging if we adopt deep learning.
Note 2: We do not distinguish AGI and Strong AI in this article.

References

Alkaissi, H., & McFarlane, S. I. (2023). Artificial hallucinations in ChatGPT: Implications in scientific writing. Cureus, 15(2): e35179.
Atkinson, R. K. (2002). Optimizing learning from examples using animated pedagogical agents. Journal of Educational Psychology, 94(2), 416-427.
Baylor, A. L. (2009). Promoting motivation with virtual agents and avatars: Role of visual presence and appearance. Philosophical Transactions of the Royal Society B: Biological Sciences, 364(1535), 3559-3565.
Blair, K., Schwartz, D. L., Biswas, G., & Leelawong, K. (2007). Pedagogical agents for learning by teaching: Teachable agents. Educational Technology, 47(1), 56-61.
Cagiltay, B., White, N. T., Ibtasar, R., Mutlu, B., & Michaelis, J. E. (2022). Understanding factors that shape children's long term engagement with an in-home learning companion robot: A case study. In Proceeding of the 21st Annual ACM Interaction Design and Children Conference, 362-373.
Castro-Alonso, J. C., Wong, R. M., Adesope, O. O., & Paas, F. (2021). Effectiveness of multimedia pedagogical agents predicted by diverse theories: A meta-analysis. Educational Psychology Review, 33, 989-1015.
Chan, T. W. (2023). 'Global Harwell' as the ultimate educational goal: Can and should we pursue this goal through SIDC research and practice in the 'Seamless AI World?' Keynote speech at the 31st International Conference on Computers in Education (ICCE 2023), https://www.youtube.com/watch?v=EIYTxNuStrc
Chan, T. W. & Baskin, A. B. (1988). Studying with the prince: The computer as a learning companion. In Proceedings of the International Conference on Intelligent Tutoring Systems (Vol. 194200), ITS'88, 194-200.
Chan, T. W. & Chou, C. Y. (1997). Exploring the design of computer supports for reciprocal tutoring. International Journal of Artificial Intelligence in Education, 8, 1-29.
Chan, T. W., Chung, Y. L., Ho, R. G., Hou, W. J., & Lin, G. L. (1992). Distributed learning companion systems - WEST revisited. In C. Frasson, G. Gauthier & G. McCalla (Eds.). In Proceedings of the 2nd international conference of intelligent tutoring systems, Lecture notes in computer science, Vol. 608, 643-650. Springer-Verlag.
Chan, T.W., Hue, C.W., Chou, C.Y., & Tzeng, O.J.L. (2001). Four Spaces of Network Learning Models, Computers & Education, 37, 141-161.
Chen, Z. H. (2012). We care about you: Incorporating pet characteristics with educational agents through reciprocal caring approach. Computers & Education, 59(4), 1081-1088.
Chen, Z. H., Chou, C. Y., Deng, C. Y., & Chan, T. W. (2007). Active open learner models as animal companions: Motivating children to learn through interaction with My-Pet and Our-Pet. International Journal of Artificial Intelligence in Education, 17(2), 145-167.
Chen, Z. H., Liao, C., Chien, T. C., & Chan, T. W. (2011). Animal companions: Fostering children's effort-making by nurturing virtual pets. British Journal of Educational Technology, 42(1), 166-180.
Chou, C. Y., Chan, T. W., & Lin, C. J. (2002). An Approach of Implementing General Learning Companions for Problem Solving, IEEE Transactions on Knowledge and Data Engineering, 14(6), 1376-1386.
Chou, C. Y., Chan, T. W., & Lin, C. J. (2003). Redefining the learning companion: The past, present, and future of educational agents. Computers & Education, 40(3), 255-269.
Clarebout, G., Elen, J., Johnson, W. L., & Shaw, E. (2002). Animated pedagogical agents: An opportunity to be grasped? Journal of Educational Multimedia and Hypermedia, 11(3), 267-286.
Davis, R. O. (2018). The impact of pedagogical agent gesturing in multimedia learning environments: A meta-analysis. Educational Research Review, 24, 193-209.
Global Harwell Group (2024a, July). Challenging Educational Problem (Part 1) at Global Harwell Forums, https://globalharwellgoal.org/intro/part1.html
Global Harwell Group (2024b, July). Challenging educational problem (Part 2). https://globalharwellgoal.org/intro/part2.html
Graesser, A. C., Chipman, P., Haynes, B. C., & Olney, A. (2005). AutoTutor: An intelligent tutoring system with mixed-initiative dialogue. IEEE Transactions on Education, 48(4)
Gulz, A., & Haake, M. (2006). Design of animated pedagogical agents—A look at their look. International Journal of Human-Computer Studies, 64(4), 322-339.
Harley, J. M., Taub, M., Azevedo, R., & Bouchet, F. (2018). Let's set up some subgoals: Understanding human-pedagogical agent collaborations and their implications for learning and prompt and feedback compliance. IEEE Transactions on Learning Technologies, 11(1), 54-66.
Hietala, P., & Niemirepo, T. (1998). The competence of learning companion agents. International Journal of Artificial Intelligence in Education, 9(3-4), 178-192.
Ho, H. R., Cagiltay, B., White, N. T., Hubbard, E. M., & Mutlu, B. (2021). Robomath: Designing a learning companion robot to support children's numerical skills. In Proceedings of the 20th Annual ACM Interaction Design and Children Conference, 283-293.
Hsieh, Y. Z., Su, M. C., Chen, S. Y., & Chen, G. D. (2015). The development of a robot-based learning companion: A user-centered design approach. Interactive Learning Environments, 23(3), 356-372.
Hsu, S. H., Chou, C. Y., Chen, F. C., Wang, Y. K., & Chan, T. W. (2007). An investigation of the differences between robot and virtual learning companions' influences on students' engagement. In 2007 First IEEE International Workshop on Digital Game and Intelligent Toy Enhanced Learning (DIGITEL'07), 41-48.
Huang, W., Hew, K. F., & Fryer, L. K. (2022). Chatbots for language learning—Are they really useful? A systematic review of chatbot-supported language learning. Journal of Computer Assisted Learning, 38(1), 237-257.
Huang, Y. T., Yang, J. C., & Wu, Y. C. (2008, July). The development and evaluation of English dialogue companion system. In 2008 Eighth IEEE International Conference on Advanced Learning Technologies, 864-868.
Hwang, G. J., & Chang, C. Y. (2023). A review of opportunities and challenges of chatbots in education. Interactive Learning Environments, 31(7), 4099-4112.
Johnson, W. L., Rickel, J. W., & Lester, J. C. (2000). Animated Pedagogical Agents: Face-to-Face Interaction in Interactive Learning Environments. International Journal of Artificial Intelligence in Education, 11, 47-78.
Karaoğlan Yılmaz, F. G., Olpak, Y. Z., & Yılmaz, R. (2018). The effect of the metacognitive support via pedagogical agent on self-regulation skills. Journal of Educational Computing Research, 56(2), 159-180.
Kim, H., Yang, H., Shin, D., & Lee, J. H. (2022). Design principles and architecture of a second language learning chatbot. Language Learning & Technology, 26(1), 1-18.
Kim, Y., & Baylor, A. L. (2006). A social-cognitive framework for pedagogical agents as learning companions. Educational Technology Research and Development, 54, 569-596.
Kim, Y., Baylor, A. L., & Pals Group. (2006). Pedagogical agents as learning companions: The role of agent competency and type of interaction. Educational Technology Research and Development, 54, 223-243.
Kim, Y., Baylor, A. L., & Shen, E. (2007). Pedagogical agents as learning companions: the impact of agent emotion and gender. Journal of Computer Assisted Learning, 23(3), 220-234.
Kory, J., & Breazeal, C. (2014). Storytelling with robots: Learning companions for preschool children's language development. In The 23rd IEEE International Symposium on Robot and Human Interactive Communication, 643-648.
Kuhail, M. A., Alturki, N., Alramlawi, S., & Alhejori, K. (2023). Interacting with educational chatbots: A systematic review. Education and Information Technologies, 28(1), 973-1018.
Kulik, J. A., & Fletcher, J. D. (2016). Effectiveness of intelligent tutoring systems: a meta-analytic review. Review of Educational Research, 86(1), 42-78.
Lester, J. C., & Stone, B. A. (1997, February). Increasing believability in animated pedagogical agents. In Proceedings of the First International Conference on Autonomous agents, 16-21.
Lester, J. C., Converse, S. A., Kahler, S. E., Barlow, S. T., Stone, B. A., & Bhogal, R. S. (1997, March). The persona effect: affective impact of animated pedagogical agents. In Proceedings of the ACM SIGCHI Conference on Human factors in computing systems, 359-366.
Lfelebuegu, A. O. (2024). Rise of the robots: What It means for educators. Journal of Applied Learning and Teaching, 7(1), 713-720.
Li, W., Wang, F., Mayer, R. E., & Liu, T. (2022). Animated pedagogical agents enhance learning outcomes and brain activity during learning. Journal of Computer Assisted Learning, 38(3), 621-637.
Liu, C. C., Chiu, C. W., Chang, C. H., & Lo, F. Y. (2024). Analysis of a chatbot as a dialogic reading facilitator: its influence on learning interest and learner interactions. Educational Technology Research and Development., 1-29.
Liu, C. H. (2024). Personal exchange.
Liu, X., Ma, J., & Wang, Q. (2024). A social robot as your reading companion: exploring the relationships between gaze patterns and knowledge gains. Journal on Multimodal User Interfaces, 18(1), 21-41.
Lubold, N., Walker, E., & Pon-Barry, H. (2021). Effects of adapting to user pitch on rapport perception, behavior, and state with a social robotic learning companion. User Modeling and User-Adapted Interaction, 31, 35-73.
Lusk, M. M., & Atkinson, R. K. (2007). Animated pedagogical agents: Does their degree of embodiment impact learning from static or animated worked examples? Applied Cognitive Psychology, 21(6), 747-764.
Ma, W., Adesope, O. O., Nesbit, J. C., & Liu, Q. (2014). Intelligent tutoring systems and learning outcomes: A meta-analysis. Journal of Educational Psychology, 106(4), 901-918.
Michaelis, J. E., & Mutlu, B. (2018). Reading socially: Transforming the in-home reading experience with a learning-companion robot. Science Robotics, 3(21), eaat5999.
Moreno, R., Mayer, R. E., Spires, H. A., & Lester, J. C. (2001). The case for social agency in computer-based teaching: Do students learn more deeply when they interact with animated pedagogical agents? Cognition and Instruction, 19(2), 177-213.
Okonkwo, C. W., & Ade-Ibijola, A. (2021). Chatbots applications in education: A systematic review. Computers and Education: Artificial Intelligence, 2, 100033.
Rasmussen, M. L. R., Larsen, A. C., Subhi, Y., & Potapenko, I. (2023). Artificial intelligence-based ChatGPT chatbot responses for patient and parent questions on vernal keratoconjunctivitis. Graefe's Archive for Clinical and Experimental Ophthalmology, 261(10).
Ruan, S., Willis, A., Xu, Q., Davis, G. M., Jiang, L., Brunskill, E., & Landay, J. A. (2019, June). Bookbuddy: Turning digital materials into interactive foreign language lessons through a voice chatbot. In Proceedings of the sixth (2019) ACM conference on learning@ scale, 1-4.
Schroeder, N. L., & Adesope, O. O. (2014). A systematic review of pedagogical agents' persona, motivation, and cognitive load implications for learners. Journal of Research on Technology in Education, 46(3), 229-251.
Skjuve, M., Følstad, A., Fostervold, K. I., & Brandtzaeg, P. B. (2021). My chatbot companion-a study of human-chatbot relationships. International Journal of Human-Computer Studies, 149, 102601.
Sikström, P., Valentini, C., Sivunen, A., & Kärkkäinen, T. (2022). How pedagogical agents communicate with students: A two-phase systematic review. Computers & Education, 188, 104564.
Spitale, M., Okamoto, S., Gupta, M., Xi, H., & Matarić, M. J. (2022). Socially assistive robots as storytellers that elicit empathy. ACM Transactions on Human-Robot Interaction (THRI), 11(4), 1-29.
Tlili, A., Shehata, B., Adarkwah, M. A., Bozkurt, A., Hickey, D. T., Huang, R., & Agyemang, B. (2023). What if the devil is my guardian angel: ChatGPT as a case study of using chatbots in education. Smart Learning Environments, 10:15.
VanLehn, K. (2011). The relative effectiveness of human tutoring, intelligent tutoring systems, and other tutoring systems. Educational Psychologist, 46(4), 197-221.
Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes. Harvard University Press.
Wang, Y. H., Young, S. S., & Jang, J. S. R. (2009). Evaluation of tangible learning companion/robot for English language learning. In 2009 Ninth IEEE International Conference on Advanced Learning Technologies, 322-326.
Wei, C. W., Hung, I., Lee, L., & Chen, N. S. (2011). A joyful classroom learning system with robot learning companion for children to learn mathematics multiplication. Turkish Online Journal of Educational Technology-TOJET, 10(2), 11-23.
Woo, H. L. (2009). Designing multimedia learning environments using animated pedagogical agents: factors and issues. Journal of Computer Assisted Learning, 25(3), 203-218.
Zinina, A., Kotov, A., Arinkin, N., & Zaidelman, L. (2023). Learning a foreign language vocabulary with a companion robot. Cognitive Systems Research, 77, 110-114.