Human-AI Rapport from the Perspective of Media Naturalness

Su Di¹², Liao Jiangqun³¹, Peng Kaiping²

¹ Department of Psychology and Cognitive Science, Tsinghua University, Beijing, 100084
² Student Mental Health Counseling Center, Ningxia University, Yinchuan, 750021
³ Business School, Beijing Technology and Business University, Beijing, 100048

Abstract

As Artificial Intelligence (AI) becomes increasingly pervasive, establishing high-quality human-AI interactive relationships has emerged as a critical research frontier. This paper provides an in-depth analysis of the possibilities for harmonious coexistence between humans and AI, introducing the concept of "Human-AI Rapport" (HAR) and its three constituent components: harmonious relationship, mutual understanding, and tacit cooperation. By integrating this concept with Media Naturalness Theory (MNT), we innovatively propose a Human-AI Rapport Promotion Model from the perspective of constructing naturalness in human-AI interaction. This model not only reveals the critical roles and implementation pathways of cognitive, social, and emotional intelligence in AI design, providing a novel theoretical lens for human-AI interaction research, but also identifies urgent questions requiring deeper investigation. We call for harnessing AI's full potential to expand its application scenarios and value boundaries, thereby promoting mutual adaptation between human society and AI technology to realize the vision of harmonious human-AI symbiosis.

Keywords: Artificial Intelligence, Human-AI Rapport, Media Naturalness

Within the theoretical framework of Marxism, productive tools serve as important indicators of productivity development levels, and their evolution directly drives transformations in social production relations. Each leap in productive tools is accompanied by tremendous gains in productivity, subsequently triggering adjustments in production relations and profound changes in social structure. Today, Artificial Intelligence (AI) has penetrated every aspect of human production and life as a new generation of productive tools. From smart homes to medical diagnosis, from educational tutoring to customer service, AI is gradually shedding its purely instrumental role to participate in social production as a means of assistance, collaboration, and human capability enhancement (Kim et al., 2022; G. Park et al., 2023; Pasternak et al., 2022; Steele et al., 2022), becoming a crucial driving force in the development of new quality productive forces (Feng, 2024). Against this important backdrop, harmonious symbiosis between humans and AI is essential, and several core questions have become increasingly prominent: What kind of relationship should humans and AI adopt for dialogue and cooperation? How can we establish high-quality interactive relationships between humans and AI? This inquiry not only concerns the user experience and acceptance of AI technology but also directly affects the practical effectiveness and future prospects of AI applications (Pentina, Xie, et al., 2023).

Previous research has universally emphasized the importance of rapport between human users and AI, advocating for the promotion of such rapport as a core orientation in AI design (Lee et al., 2024; Mai et al., 2021; Nichols et al., 2022; G. Park et al., 2023). However, these discussions have largely remained within the traditional frameworks of interpersonal rapport or professional rapport (Tickle-Degnen & Rosenthal, 1990; Gabbert et al., 2021), failing to construct a unique conceptual system or theoretical model specifically for rapport between humans and AI. Clearly, simply applying interpersonal rapport theories to interpret human-AI relationships has limitations and is not entirely appropriate. In view of this, this study innovatively proposes the concept of "Human-AI Rapport" (HAR) to describe the quality of the relationship between humans and AI when humans accept AI services or engage in professional interactions with AI. This concept aims to provide a more solid and targeted theoretical foundation for subsequent related research, thereby promoting deeper development in this field.

Currently, research on human-AI rapport involves multiple disciplines including psychology and computer science. Psychology, as a science that deeply analyzes the cognition, behavior, emotions, and affect of interacting parties, provides the theoretical foundation required for data collection, annotation, and model construction in human-AI rapport research. Computer science, meanwhile, provides the technical means (devices, algorithms, programs) to achieve comprehensive collection, precise computation, and innovative generation of multimodal data. The two complement each other and jointly promote the development of human-AI rapport theory and practice. Against this background, this study aims to explore the connotation and establishment strategies of human-AI rapport, hoping to provide useful references and insights for the human-centered design of AI technology, optimization of user experience, and harmonious construction of human-AI relationships.

2.1 Interpersonal Rapport and Professional Rapport

In interpersonal interactions, relationships characterized by harmony, smoothness, and synchronization are called rapport, which represents good relationship quality and tacit understanding, and is one of the typical features of many successful interactions (Gratch et al., 2007). Tickle-Degnen and Rosenthal (1990) propose that rapport includes three components: positivity, which refers to friendliness and warmth between both parties; attention, which refers to mutual interest and concern; and coordination, which refers to the perception of balance and harmony during interaction. From a third-party observation perspective, rapport also has three corresponding manifestations: head, facial, and vocal expressions (such as nodding, smiling, intonation changes, etc.) reflect positivity between both parties; static body movements represent attention; and dynamic body movements (such as adjusting posture to adapt to the other) increase coordination (Hendrick, 1990). Past research has demonstrated that rapport in interpersonal interactions can produce positive effects in fields such as negotiation, management, psychological counseling and therapy, teaching, and nursing (Gabbert et al., 2021; Gratch et al., 2007; Kim et al., 2022).

In contrast to general social environments, after systematically reviewing research on the use and measurement of rapport in professional information-gathering contexts, Gabbert et al. (2021) proposed the concept of professional rapport-building, which refers to the intentional use of rapport-facilitating behaviors to promote positive interactions with work objects in professional settings (such as negotiations, investigative interviews, etc.). Therefore, while rapport discussed in social domains represents an interactive state to be achieved in ideal and high-quality interpersonal interactions, in time-constrained professional domains, rapport is a task-oriented practical strategy and means used by interviewers/consultants/interviewers to establish connections, facilitate dialogue, and achieve goals with work objects, though it does not necessarily establish genuine and long-term rapport (Brouillard et al., 2024; Gabbert et al., 2021; Neequaye, 2023). Functionally, there are three main ways to build rapport: (1) customizing interviews through self-disclosure and discussing personal issues; (2) demonstrating approachability through smiling, open body language, and appropriate tone; and (3) showing positive attention to interviewees through active listening, eye contact, understanding, and affirmation (Gabbert et al., 2021).

2.2 The Concept and Components of Human-AI Rapport

Similar to the scenarios where professionals conduct work mentioned earlier, under various forms of interaction, AI's functions have also evolved from simply executing commands to more intelligent applications such as collecting user information, analyzing user intentions, achieving understanding and empathy, and providing services and companionship (Hou et al., 2024; Kim et al., 2022; Ranjbartabar et al., 2021). Therefore, it is necessary to discuss whether the above rapport and building strategies are equally applicable to human-AI interaction in professional scenarios.

Although embodied AI agents can mimic users' nonverbal behaviors to "fabricate" positivity, attention, and coordination in human-AI interaction (Huang et al., 2011; Lubold et al., 2021), they cannot achieve physiological synchronization and neural synchronization (McNaughton & Redcay, 2020). Excessive attention from machines may also cause users to feel that their safety and privacy are threatened, leading to panic and aversion (Koller et al., 2023). Moreover, a large number of non-embodied AI agents are also helping professional workers or users achieve task goals (Chattaraman et al., 2019; G. Park et al., 2023), but they cannot use behavioral performance strategies (such as changing body posture and expressions) to build rapport with humans. Therefore, traditional rapport concepts and frameworks may be poorly adapted to human-AI interaction.

Combining the concept of traditional interpersonal rapport with the particularities of human-AI interaction mentioned above (Diederich et al., 2022), this study proposes the concept of "Human-AI Rapport" (HAR), referring to the degree to which users experience harmonious relationship, mutual understanding, and tacit cooperation with AI in the process of using AI to achieve various professional work goals. It is users' subjective perception of the depth and quality of the interactive relationship. The three components in this concept are related to but distinct from the three dimensions of traditional interpersonal rapport (see Table 1 [TABLE:1]): harmonious relationship is built upon "positivity" but emphasizes users' good experience of the overall cooperative atmosphere (an experience that does not merely come from AI's consistently positive feedback). Mutual understanding is built upon "attention," where AI's attention is not only reflected in collecting user behavioral performance data but also requires understanding the intentions behind user behaviors and expressing this understanding in ways acceptable to users. Tacit cooperation is built upon "coordination," specifically referring to the ability to tacitly maintain synchronization with users while achieving complementarity with them in cooperation without requiring user requests or emphasis (such as personalized recommendations, memory assistance, etc.).

A review of the literature reveals that concepts similar to human-AI rapport include AI usage, trust in AI, attitudes toward AI, and relationship with AI. To clearly delineate the differences between human-AI rapport and these related concepts and highlight its unique theoretical value, it is necessary to provide a detailed and in-depth explanation (see Table 2 [TABLE:2]). First, the connotations and evaluation contents of each concept differ. AI usage refers to individuals' behaviors or willingness to accept and adopt AI (Wang et al., 2024; B. Li et al., 2023). Attitudes toward AI refer to individuals' psychological tendencies to approve of or resist AI being used in work, life, and decision-making (De Freitas et al., 2023; J. Park et al., 2024). Relationship with AI refers to the degree of liking and emotional dependence that gradually develops between users and AI as contact time and AI roles change (Pentina, Hancock, et al., 2023; Tschopp et al., 2023). Trust in AI refers to users' positive expectations when interacting with AI amidst uncertainty and unpredictability, including both evaluations of AI's functional reliability and recognition of the values and ethics behind AI algorithms, as well as corresponding psychological security (Asan et al., 2020; Choung et al., 2023). Second, the concepts differ in their generalization degree when evaluating AI and its application scenarios. Human-AI rapport focuses on evaluations of users' interactions with specific AI agents in professional contexts, with clear directionality; other concepts are not limited to evaluations of specific AI agents or interaction scenarios and may generalize to evaluations of broad AI technology and application scenarios. These evaluations stem from both actual usage experience and external information such as others' experiences, viewpoints, and news reports. Finally, human-AI rapport differs from other concepts in the dimension of evaluation objects. Other concepts typically focus evaluation from humans onto AI, emphasizing unilateral consideration of AI; human-AI rapport focuses on evaluating the bilateral relationship during users' professional interactions, meaning its evaluation content is rooted in factors elicited by the interaction itself rather than solely focusing on AI's characteristics.

3.1 Application of Interpersonal Rapport Models in the Human-AI Domain

Previous research on human-AI relationships has largely held that theories related to human relationships can be used to further understand the development of human-AI relationships (Seymour & Van Kleek, 2021; Xie & Pentina, 2022), and scholars have widely applied Tickle-Degnen and Rosenthal's (1990) interpersonal rapport model (i.e., positivity, attention, and coordination) to research on building and predicting human-AI rapport (Gratch et al., 2007; Maxim et al., 2023; Pasternak et al., 2022). For example, Huang et al. (2011) developed Virtual Rapport 2.0 based on the Virtual Rapport 1.0 virtual agent (Gratch et al., 2006) using the interpersonal rapport model as a framework. They employed a data-driven approach to enhance the attentiveness and coordination of the interactive relationship and used emotional response and reciprocity enhancement techniques to strengthen the positivity of emotional communication. Results showed that compared to version 1.0, version 2.0 effectively improved users' perception of rapport (Huang et al., 2011).

3.2 Media Naturalness Theory

Kock's (2004) Media Naturalness Theory (MNT) posits that humans are biologically predisposed to communicate through voice, facial expressions, and body language. This natural communication includes five elements: (1) sharing the same environment where people can see and hear each other; (2) obtaining real-time and synchronous feedback from partners; (3) being able to convey and observe facial expressions; (4) being able to convey and observe body language; and (5) being able to convey and hear voice. However, as technology develops, technology-mediated interactions have become increasingly common, but these interactions often fail to meet the above elements and inhibit many natural features of face-to-face communication. Therefore, naturalness in interaction—the degree to which interactive media achieves similarity with face-to-face communication through technology (Kock, 2004)—becomes key to solving this problem. It emphasizes creating a sense of synchronicity and collocation in interaction by simulating face-to-face voice, facial expressions, and body movements, making technology-mediated interactions as close as possible to natural face-to-face communication (Kock, 2004, 2005). MNT further identifies three core mechanisms for enhancing interaction naturalness: decrease in cognitive effort, reduction in communication ambiguities, and increase in physiological arousal (Kock, 2004).

3.3 A New Human-AI Rapport Promotion Model

Although theoretical models of interpersonal rapport have been widely applied in human-AI interaction research, inherent differences between the two determine that the direct applicability of interpersonal rapport-building strategies in human-AI interaction contexts is not absolute (Ter Stal et al., 2021). Moreover, previous research on human-AI rapport has been relatively fragmented, and a systematic and complete theoretical model for how to promote human-AI rapport has not yet been formed. MNT provides an effective and specific implementation path for building and promoting human-AI rapport: the more natural AI is, the more humans tend to view it as a synchronous and collocated communication partner, accompanied by positive perceptions and attitudes. Therefore, building a harmonious and rapport relationship between humans and AI, and making AI be perceived as "human-like," hinges on AI design that closely centers on human experience and focuses on enhancing the naturalness of its interaction with humans. This study attempts to combine the newly proposed concept and structure of human-AI rapport with the core ideas of MNT to propose a novel Human-AI Rapport Promotion Model (see Figure 1 [FIGURE:1]).

The model has two layers. The outer layer comprises the design objectives for human-AI interaction, including human-AI rapport objectives (three vertices representing the three evaluation dimensions of human-AI rapport) and media naturalness objectives (lines connecting any two dimensions, representing that achieving these objectives can enhance corresponding rapport perceptions). The inner layer comprises the pathways to achieve these objectives, including three types of AI intelligence: cognitive intelligence, social intelligence, and emotional intelligence. From inside to outside, through technological improvements, AI's enhancement in these three intelligences will effectively achieve the three core interaction objectives in MNT and significantly improve AI's naturalness (Chandra et al., 2022), ensuring that humans remain fluent and engaged when communicating with AI, thereby making the human-AI interaction process closer to the authentic experience of interpersonal interaction. Ultimately, human-AI rapport promotion is achieved through these three pathways.

4 Strategies for Promoting Human-AI Rapport

Based on the above Human-AI Rapport Promotion Model, AI design aimed at enhancing naturalness is the fundamental strategy for improving rapport between users and AI. Researchers and designers need to explore how to effectively promote human-AI rapport from three perspectives: cognitive intelligence, social intelligence, and emotional intelligence.

4.1 Enhancing Cognitive Intelligence

The goals of mutual understanding and tacit cooperation between humans and AI place higher demands on AI's cognitive intelligence. Thanks to evolution, the human brain can easily recognize and process interpersonal communication signals such as facial expressions, body movements, and voice without excessive cognitive effort (Kock, 2005). However, when communicating with AI of low naturalness, humans may need to expend more cognitive effort and learn new communication strategies to understand AI or make themselves understood by AI, which is not conducive to building rapport. Therefore, AI should not merely be a passive tool that answers user questions or provides calculation results; it needs to demonstrate cognitive intelligence, particularly the ability to understand, summarize, and reason about users' instructions and intentions, and make correct, adaptive, and appropriate responses. This reduces users' cognitive load when interacting with AI, specifically decreasing the cognitive effort required for repetition, explanation, and clarification (Hudon et al., 2021), thereby making users perceive that they can easily achieve mutual understanding with AI and accomplish synchronized and complementary tacit cooperation.

Today's deep learning technology has endowed AI systems with pattern recognition and feature extraction capabilities that demonstrate cognitive intelligence, enabling them to perceive user needs and execute complex tasks in more adaptive ways, thereby enhancing user experience. For example, topic spotting technology proposed by researchers can help dialogue systems automatically infer conversation topics, making dialogue systems more engaging and efficient, which helps maintain long-term conversations with users and, based on this, facilitates mutual understanding and rapport building between humans and AI (Chitkara et al., 2019). Memory and recall capabilities are also important cognitive abilities (Richards & Bransky, 2014). If AI can "remember" previous conversations with users and appropriately "recall" relevant "memories" during conversations, it can effectively promote their interaction (Kasap & Magnenat-Thalmann, 2012). This is not only a manifestation of intelligence but also allows the intelligent agent to better understand users through their background knowledge (Richards, 2017; Richards & Bransky, 2014). Moreover, in interaction, AI's recall of information must be natural and context-appropriate; otherwise, it will damage users' perception of tacit cooperation (Richards & Bransky, 2014). Conversely, if AI lacks memory and recall capabilities, it forces humans to invest more cognitive resources in repetition and information prompting during interaction, creating a negative interactive experience of not being understood.

Nichols et al. (2022) established a dialogue generation strategy called Tiers of Friendship, enabling intelligent agents to build deeper conversation topics with users based on historical dialogues, thereby enhancing rapport with users (Nichols et al., 2022). Corresponding to memory is the question of how AI should "tacitly" demonstrate "forgetting" (Ellwart & Kluge, 2019), because humans may not want all information to be remembered. This places higher demands on the intelligent agent's sensitivity and adaptability (Richards & Bransky, 2014).

4.2 Exhibiting Social Intelligence

In human-AI interaction with low naturalness, the lack of nonverbal social cues often triggers communication ambiguities and negative user impressions, which is detrimental to relationship building and creates obstacles to tacit cooperation (Feine et al., 2019; I. Park et al., 2023). Similar to strategies for building rapport in interpersonal interactions (Cheung et al., 2015; Kulesza et al., 2016; Lin & Lin, 2017), researchers have endowed AI with nonverbal communication capabilities through animated characters, mechanical devices, and voice output, enabling users to perceive naturalness in communication and cooperation through AI's posture (Riek et al., 2010), gestures (Cade et al., 2011; Lubold et al., 2021), facial expressions (Pasternak et al., 2022), and voice intonation (Lubold et al., 2021), thereby establishing good human-AI rapport.

Beyond social cues, providing identity cues is also a social strategy that helps individuals build good relationships with others and enhance social influence (Taylor et al., 2023). For AI, merely mimicking nonverbal behaviors is insufficient to provide enough social intelligence to enhance its naturalness. AI needs to proactively provide identity cues and background information for interaction to reduce communication ambiguities and increase its social presence in users' eyes (Chen et al., 2023; Go & Sundar, 2019), which triggers users to respond to technology in the same way they treat humans (Westerman et al., 2020). For example, researchers made social improvements to robots assisting convenience store employees, such as proactively greeting employees through text or voice, introducing themselves, and explaining role assignments, which made employees view the robot as a friendly social other, effectively promoting task cooperation, improving employees' perception of the work atmosphere, and forming human-AI rapport (Takahashi et al., 2022).

Similarly, AI self-disclosure can make users more willing to open up (Neef et al., 2022). This strategy has been proven to promote coordination and positive rapport between both parties and improve AI's effectiveness in completing corresponding tasks in health, teaching, and negotiation AI applications (Mai et al., 2021; Neef et al., 2022; Zhao et al., 2018). Additionally, AI addressing users by name and engaging in small talk has also been proven to be an effective strategy for building rapport with users as a manifestation of social intelligence (Lee et al., 2024).

4.3 Optimizing Emotional Intelligence

Elements unique to human face-to-face communication (such as facial expressions) contribute to physiological arousal, bring positive emotional experiences, and promote direct mutual understanding (Krumhuber et al., 2013). When low-naturalness AI interacts with humans, these elements are easily suppressed, resulting in lower perceived excitement and loss of many emotional experiences beneficial to interaction (Kock, 2004, 2005). AI with emotional intelligence will enhance users' physiological arousal, thereby increasing naturalness, making users experience that AI's emotional expressions come not from machines' uniform, undifferentiated reactions but from perception and understanding of human emotions, and responding in ways that express understanding and empathy, thereby promoting rapport building (Tickle-Degnen & Rosenthal, 1990).

Therefore, emotional intelligence is an indispensable component in building human-AI rapport, representing AI's comprehensive ability to perceive, use, understand, and manage emotions (Hou et al., 2024). First, AI captures subtle emotional changes from users' facial expressions and even physiological signals, which is the foundation for understanding user needs and predicting user reactions. For example, by integrating the Computer Expression Recognition Toolbox, AI can analyze users' expressions and emotional states in real time and adjust its response strategies accordingly to communicate with users in more appropriate ways (Cerekovic et al., 2017). Second, although humans often express emotions through nonverbal behaviors (facial expressions, posture, tone, etc.), AI that interacts with users through language can also possess the ability to identify, analyze, and interpret users' emotional states from conversational context through deep learning, natural language processing, and machine learning technologies (Anzum & Gavrilova, 2023; Chen et al., 2024; D & Juliet, 2023).

On the basis of recognition and understanding, responding to users with empathy is an effective means to enhance coordination and attentiveness in AI-human interaction, and related technologies for empathetic dialogue generation have gradually matured, providing guarantees for improving human-AI interaction experience and relationship quality (Hou et al., 2024). When AI makes responsive movements or expressions, or demonstrates resonance with users' emotions through dialogue, users feel understood and supported, thereby establishing closer emotional bonds and rapport (Abdulrahman et al., 2021; Chen et al., 2024; Namkoong et al., 2024; Ranjbartabar et al., 2021).

5 Discussion and Outlook

This study innovatively proposes the concept of human-AI rapport, deeply explores its connotation in combination with relevant literature in the human-AI relationship field, and specifically discusses how to effectively establish human-AI rapport under an integrative framework of MNT and human-AI rapport, further enriching the theoretical framework in this field, providing a new perspective for human-AI relationship research, and offering suggestions for future AI design and optimization. Future research should conduct deeper investigations into the following questions in this field.

5.1 Measurement of Human-AI Rapport

Previous studies have used different questionnaires to measure users' perception of human-AI rapport. The main measurement approaches fall into three categories: (1) measurements based on the content and dimensions of the three-factor model proposed by Tickle-Degnen and Rosenthal (1990) (Lubold et al., 2018; Steele et al., 2022); (2) measurements that integrate indicators such as attitudes, relationship, and trust as previously described, with Gratch et al.'s (2007) rapport scale being more common—this 10-item scale evaluates rapport between users and AI entities from 10 perspectives including connectedness, understanding, helpfulness, credibility, likability, naturalness, favorability, anthropomorphism, persuasiveness, and recommendation, and has been adapted and adopted by many researchers (Acosta & Ward, 2011; Lubold et al., 2021; Ter Stal et al., 2021); and (3) adoption of other concepts and corresponding scales, such as working alliance (Mai et al., 2021) and social presence (Huang et al., 2011).

In the future, based on a deeper understanding of the concept and structure of human-AI rapport, there is an urgent need to develop more authoritative and valid self-report scales to ensure research coherence and accuracy, thereby avoiding biases caused by measurement differences and laying a solid foundation for research progress in the field. In addition, some objective indicators can be used to assist in measuring human-AI rapport during experiments. Previous studies have often used behavioral indicators; for example, Wong and McGee (2012) used the length and fluency of participants' speech as indicators of experienced human-AI rapport, while Lubold et al. (2021) recorded the frequency of participants' use of polite language, praise for AI, speaking AI's name, and using "inclusive" language (such as "we") to comprehensively measure human-AI rapport, supplemented by self-report scales. Users' facial expressions can also be recognized and used to determine the presence of rapport (Sharma et al., 2021). With in-depth research on the brain and physiological mechanisms of interpersonal rapport (Ellingsen et al., 2022), physiological indicators such as EEG, ECG, and EMG can also be collected in the future to reflect users' perception of rapport during interaction with AI.

5.2 Direction of AI Intelligence Enhancement

AI's computing power and intelligence are becoming increasingly high with technological development and can gradually assist or replace humans in completing some complex tasks. However, from the perspective of human-AI rapport goals, does AI's intelligence level absolutely positively affect human-AI rapport? According to media naturalness theory, media that differ greatly from face-to-face communication may require more cognitive effort even if they have rich nonverbal cues and fast feedback mechanisms, because they do not conform to humans' biological communication mechanisms (Kock, 2004). This further demonstrates that the key factor in building human-AI rapport is not AI's cognitive, social, and emotional intelligence per se, but how intelligence enhancement affects users' perception of naturalness in human-AI interaction. Existing research has also shown that in social intelligence (Chattaraman et al., 2019; Pecune et al., 2020) and emotional intelligence (Ranjbartabar et al., 2021; Ter Stal et al., 2021), if intelligence enhancement conflicts with users' interaction purposes, it may have negative effects on human-AI rapport. However, when AI uses algorithms to understand users' needs and goals and intelligently adjusts its social and emotional expression, it can effectively improve human-AI rapport (Pecune et al., 2020). Therefore, future research can further explore how to improve the direction of AI intelligence enhancement to better match the goals of human-AI relationship development.

5.3 Exploring More Symbiotic Human-AI Relationships

Previous research on human-AI rapport has mostly explored how AI applied in professional service fields or interacting with users in single, short-term sessions in laboratory settings is perceived and evaluated by users (Huang et al., 2011; Pasternak et al., 2022; Ranjbartabar et al., 2021). However, with social and technological development, AI will assume more roles that require building rapport (such as personal assistants, intelligent life companions, virtual lovers, etc.), will spend longer time with users, and may have fixed names, stable personality traits and identity, and even experiences formed through long-term cohabitation (W. Li et al., 2023). This will continuously expand AI's applicable scenarios and application value and give rise to more symbiotic human-AI interaction models that emphasize rapport, also bringing more challenges to future human society's acceptance and adaptation to AI technology. Therefore, based on the new concept and theoretical model proposed in this study, future research should focus on more complex types of human-AI relationships in terms of symbiosis (such as multi-stakeholder perspectives; Law et al., 2024) to stimulate more discussion and practice on future visions of human-AI cooperation and promote the development of AI technology toward greater naturalness, pleasantness, and sustainability, truly realizing the vision of harmonious human-AI symbiosis.

References

Abdulrahman, A., Richards, D., Ranjbartabar, H., & Mascarenhas, S. (2021). Verbal empathy and explanation to encourage behaviour change intention. Journal on Multimodal User Interfaces, 15(2), 189–199.

Acosta, J. C., & Ward, N. G. (2011). Achieving rapport with turn-by-turn, user-responsive emotional coloring. Speech Communication, 53(9), 1137–1148.

Anzum, F., & Gavrilova, M. L. (2023). Emotion detection from micro-blogs using novel input representation. IEEE Access, 11, 19512–19522.

Asan, O., Bayrak, A. E., & Choudhury, A. (2020). Artificial intelligence and human trust in healthcare: Focus on clinicians. Journal of Medical Internet Research, 22(6), e15154.

Brouillard, C., Gabbert, F., & Scott, A. J. (2024). Addressing current issues in assessing professional rapport: A systematic review and synthesis of existing measures. Applied Cognitive Psychology, 38(3), e4205.

Cade, W. L., Olney, A., Hays, P., & Lovel, J. (2011). Building rapport with a 3D conversational agent. In S. D’Mello, A. Graesser, B. Schuller, & J.-C. Martin (Eds.), Affective Computing and Intelligent Interaction (pp. 305–306). Springer.

Cerekovic, A., Aran, O., & Gatica-Perez, D. (2017). Rapport with virtual agents: What do human social cues and personality explain? IEEE Transactions on Affective Computing, 8(3), 335–349.

Chandra, S., Shirish, A., & Srivastava, S. C. (2022). To be or not to be …human? Theorizing the role of human-like competencies in conversational artificial intelligence agents. Journal of Management Information Systems, 39(4), 969–1005.

Chattaraman, V., Kwon, W.-S., Gilbert, J. E., & Ross, K. (2019). Should AI-based, conversational digital assistants employ social- or task-oriented interaction style? A task-competency and reciprocity perspective for older adults. Computers in Human Behavior, 90, 315–330.

Chen, K., Lian, H., Gao, Y., & Li, Y. (2024). Emotional support dialog system through recursive interactions among large language models. In National Conference on Man-Machine Speech Communication (pp. 151-163). Singapore: Springer Nature Singapore.

Chen, Y.-C., Yeh, S.-L., Lin, W., Yueh, H.-P., & Fu, L.-C. (2023). The effects of social presence and familiarity on children–robot interactions. Sensors, 23(9), 4231.

Cheung, E. O., Slotter, E. B., & Gardner, W. L. (2015). Are you feeling what I’m feeling? The role of facial mimicry in facilitating reconnection following social exclusion. Motivation and Emotion, 39(4), 613–630.

Chitkara, P., Modi, A., Avvaru, P., Janghorbani, S., & Kapadia, M. (2019). Topic spotting using hierarchical networks with self attention. arXiv preprint arXiv:1904.02815.

Choung, H., David, P., & Ross, A. (2023). Trust in AI and its role in the acceptance of AI technologies. International Journal of Human–Computer Interaction, 39(9), 1727–1739.

D, D., & Juliet, S. (2023). Sentimental analysis based on user emotions using machine learning. In 2023 International Conference on Circuit Power and Computing Technologies (ICCPCT).

De Freitas, J., Agarwal, S., Schmitt, B., & Haslam, N. (2023). Psychological factors underlying attitudes toward AI tools. Nature Human Behaviour, 7(11), 1845–1854.

Diederich, S., Brendel, A., Morana, S., & Kolbe, L. (2022). On the design of and interaction with conversational agents: An organizing and assessing review of human-computer interaction research. Journal of the Association for Information Systems, 23(1), 96–138.

Ellingsen, D.-M., Duggento, A., Isenburg, K., Jung, C., Lee, J., Gerber, J., Mawla, I., Sclocco, R., Edwards, R. R., Kelley, J. M., Kirsch, I., Kaptchuk, T. J., Toschi, N., & Napadow, V. (2022). Patient–clinician brain concordance underlies causal dynamics in nonverbal communication and negative affective expressivity. Translational Psychiatry, 12(1), 1–9.

Ellwart, T., & Kluge, A. (2019). Psychological perspectives on intentional forgetting: An overview of concepts and literature. KI - Künstliche Intelligenz, 33(1), 79–84.

Feine, J., Gnewuch, U., Morana, S., & Maedche, A. (2019). A taxonomy of social cues for conversational agents. International Journal of Human-Computer Studies, 132, 138–161.

Feng, D. (2024). The contemporary implications of artificial intelligence advancing the development of new quality productive forces: An exploration based on Marx's view of machinery. Studies in Dialectics of Nature, 09, 77-82.

Gabbert, F., Hope, L., Luther, K., Wright, G., Ng, M., & Oxburgh, G. (2021). Exploring the use of rapport in professional information-gathering contexts by systematically mapping the evidence base. Applied Cognitive Psychology, 35(2), 329–341.

Go, E., & Sundar, S. S. (2019). Humanizing chatbots: The effects of visual, identity and conversational cues on humanness perceptions. Computers in Human Behavior, 97, 304–316.

Gratch, J., Wang, N., Gerten, J., Fast, E., & Duffy, R. (2007). Creating rapport with virtual agents. In C. Pelachaud, J.-C. Martin, E. André, G. Chollet, K. Karpouzis, & D. Pelé (Eds.), Intelligent Virtual Agents (pp. 125–138). Springer.

Hendrick, C. (1990). The nature of rapport. Psychological Inquiry, 1(4), 312–315.

Hou, H., Ni, S., Lin, S., & Wang, P. (2024). When AI learns empathy: Themes, scenarios, and optimization of empathic computing from a psychological perspective. Advances in Psychological Science, 05, 845-858.

Huang, L., Morency, L.-P., & Gratch, J. (2011). Virtual rapport 2.0. In H. H. Vilhjálmsson, S. Kopp, S. Marsella, & K. R. Thórisson (Eds.), Intelligent Virtual Agents (pp. 68–79). Springer.

Hudon, A., Demazure, T., Karran, A., Léger, P.-M., & Sénécal, S. (2021). Explainable artificial intelligence (XAI): How the visualization of AI predictions affects user cognitive load and confidence. In F. D. Davis, R. Riedl, J. Vom Brocke, P.-M. Léger, A. B. Randolph, & G. Müller-Putz (Eds.), Information Systems and Neuroscience (Vol. 52, pp. 237–246). Springer International Publishing.

Kasap, Z., & Magnenat-Thalmann, N. (2012). Building long-term relationships with virtual and robotic characters: The role of remembering. The Visual Computer, 28(1), 87–97.

Kim, H., So, K. K. F., & Wirtz, J. (2022). Service robots: Applying social exchange theory to better understand human–robot interactions. Tourism Management, 92, 104537.

Kock, N. (2004). The psychobiological model: Towards a new theory of computer-mediated communication based on Darwinian evolution. Organization Science, 15(3), 327–348.

Kock, N. (2005). Media richness or media naturalness? The evolution of our biological communication apparatus and its influence on our behavior toward E-communication tools. IEEE Transactions on Professional Communication, 48(2), 117–130.

Koller, M., Weiss, A., Hirschmanner, M., & Vincze, M. (2023). Robotic gaze and human views: A systematic exploration of robotic gaze aversion and its effects on human behaviors and attitudes. Frontiers in Robotics and AI, 10.

Krumhuber, E. G., Kappas, A., & Manstead, A. S. R. (2013). Effects of dynamic aspects of facial expressions: A review. Emotion Review, 5(1), 41–46.

Kulesza, W., Dolinski, D., & Wicher, P. (2016). Knowing that you mimic me: The link between mimicry, awareness and liking. Social Influence, 11(1), 68-74.

Law, R., Ye, H., & Lei, S. S. I. (2024). Ethical artificial intelligence (AI): Principles and practices. International Journal of Contemporary Hospitality Management, 37(1), 279–295.

Lee, J., Lee, D., & Lee, J. (2024). Influence of rapport and social presence with an AI psychotherapy chatbot on users’ self-disclosure. International Journal of Human–Computer Interaction, 40(7), 1620–1631.

Li, B., Chen, Y., Liu, L., & Zheng, B. (2023). Users’ intention to adopt artificial intelligence-based chatbot: A meta-analysis. The Service Industries Journal, 43(15–16), 1117–1139.

Li, W., Chen, S., Sun, L., & Yang, C. (2023). What makes virtual intimacy...intimate? Understanding the phenomenon and practice of computer-mediated paid companionship. Proc. ACM Hum.-Comput. Interact., 7(CSCW1), 112:1-23.

Lin, C.-Y., & Lin, J.-S. C. (2017). The influence of service employees’ nonverbal communication on customer-employee rapport in the service encounter. Journal of Service Management, 28(1), 107–132.

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(1), 35–73.

Lubold, N., Walker, E., Pon-Barry, H., Flores, Y., & Ogan, A. (2018). Using iterative design to create efficacy-building social experiences with a teachable robot. International Society of the Learning Sciences, Inc. [ISLS], 737–744.

Mai, V., Wolff, A., Richert, A., & Preusser, I. (2021). Accompanying reflection processes by an AI-based StudiCoachBot: A study on rapport building in human-machine coaching using self disclosure. In HCI International 2021—Late Breaking Papers: Cognition, Inclusion, Learning, and Culture (pp. 439–457). Springer International Publishing.

Maxim, A., Zalake, M., & Lok, B. (2023). The impact of virtual human vocal personality on establishing rapport: A study on promoting mental wellness through extroversion and vocalics. Proceedings of the 23rd ACM International Conference on Intelligent Virtual Agents, 1–8.

McNaughton, K. A., & Redcay, E. (2020). Interpersonal synchrony in autism. Current Psychiatry Reports, 22(3), 12.

Namkoong, M., Park, Y., Lee, S., & Park, G. (2024). Effect of AI counsellor’s nonverbal immediacy behaviour on user satisfaction: Serial mediation of positive expectancy violation and rapport. Behaviour & Information Technology, 1–17.

Neef, C., Mai, V., & Richert, A. (2022). “I am scared of viruses, too”—Studying the impact of self-disclosure in chatbots for health-related applications. In M. Kurosu (Ed.), Human-Computer Interaction. User Experience and Behavior (pp. 515–530). Springer International Publishing.

Neequaye, D. A. (2023). Why rapport seems challenging to define and what to do about the challenge. Collabra: Psychology, 9(1), 90789.

Nichols, E., Siskind, S. R., Ivanchuk, L., Pérez, G., Kamino, W., Šabanović, S., & Gomez, R. (2022). Hey Haru, let's be friends! using the tiers of friendship to build rapport through small talk with the tabletop robot Haru. 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 6101–6108.

Park, G., Lee, S., & Chung, J. (2023). Do anthropomorphic chatbots increase counseling satisfaction and reuse intention? The moderated mediation of social rapport and social anxiety. Cyberpsychology, Behavior, and Social Networking, 26(5), 357–365.

Park, I., Lee, S., & Lee, D. (2023). Virtual audience providing AI-generated emotional reactions to enhance self-disclosure in self-introduction. International Journal of Human–Computer Interaction, 39(13), 2702-2713.

Park, J., Woo, S. E., & Kim, J. (2024). Attitudes towards artificial intelligence at work: Scale development and validation. Journal of Occupational and Organizational Psychology, 97(3), 920–951.

Pasternak, K., Wu, Z., Visser, U., & Lisetti, C. (2022). Towards building rapport with a human support robot. In R. Alami, J. Biswas, M. Cakmak, & O. Obst (Eds.), RoboCup 2021: Robot World Cup XXIV (pp. 214–225). Springer International Publishing.

Pecune, F., Marsella, S., & Jain, A. (2020). A framework to co-optimize task and social dialogue policies using Reinforcement Learning. Proceedings of the 20th ACM International Conference on Intelligent Virtual Agents, 1–8.

Pentina, I., Hancock, T., & Xie, T. (2023). Exploring relationship development with social chatbots: A mixed-method study of replika. Computers in Human Behavior, 140, 107600.

Pentina, I., Xie, T., Hancock, T., & Bailey, A. (2023). Consumer–machine relationships in the age of artificial intelligence: Systematic literature review and research directions. Psychology & Marketing, 40(8), 1593–1614.

Ranjbartabar, H., Richards, D., Bilgin, A. A., & Kutay, C. (2021). First impressions count! The role of the human’s emotional state on rapport established with an empathic versus neutral virtual therapist. IEEE Transactions on Affective Computing, 12(3), 788–800.

Richards, D. (2017). Intimately intelligent virtual agents: Knowing the human beyond sensory input. Proceedings of the 1st ACM SIGCHI International Workshop on Investigating Social Interactions with Artificial Agents, 39–40.

Richards, D., & Bransky, K. (2014). ForgetMeNot: What and how users expect intelligent virtual agents to recall and forget personal conversational content. International Journal of Human-Computer Studies, 72(5), 460–476.

Riek, L. D., Paul, P. C., & Robinson, P. (2010). When my robot smiles at me: Enabling human-robot rapport via real-time head gesture mimicry. Journal on Multimodal User Interfaces, 3(1), 99–108.

Seymour, W., & Van Kleek, M. (2021). Exploring interactions between trust, anthropomorphism, and relationship development in voice assistants. Proceedings of the ACM on Human-Computer Interaction, 5(CSCW2), 1-16.

Sharma, S., Gangadhara, K. G., Xu, F., Slowe, A. S., Frank, M. G., & Nwogu, I. (2021). Coupled systems for modeling rapport between interlocutors. 2021 16th IEEE International Conference on Automatic Face and Gesture Recognition (FG 2021), 1–8.

Steele, C., Lobczowski, N., Davison, T., Yu, M., Diamond, M., Kovashka, A., Litman, D., Nokes-Malach, T., & Walker, E. (2022). It takes two: Examining the effects of collaborative teaching of a robot learner. International Conference on Artificial Intelligence in Education (pp. 604–607). Springer International Publishing.

Takahashi, T., Song, S., Baba, J., Nakanishi, J., Yoshikawa, Y., & Ishiguro, H. (2022). Can an empathetic teleoperated robot be a working mate that supports operator’s mentality? 2022 17th ACM/IEEE International Conference on Human-Robot Interaction (HRI), 1059–1061.

Taylor, S. J., Muchnik, L., Kumar, M., & Aral, S. (2023). Identity effects in social media. Nature Human Behaviour, 7(1), 27–37.

Ter Stal, S., Jongbloed, G., & Tabak, M. (2021). Embodied conversational agents in eHealth: How facial and textual expressions of positive and neutral emotions influence perceptions of mutual understanding. Interacting with Computers, 33(2), 167–176.

Tickle-Degnen, L., & Rosenthal, R. (1990). The nature of rapport and its nonverbal correlates. Psychological Inquiry, 1(4), 285–293.

Tschopp, M., Gieselmann, M., & Sassenberg, K. (2023). Servant by default? How humans perceive their relationship with conversational AI. Cyberpsychology: Journal of Psychosocial Research on Cyberspace, 17(3), Article 3.

Wang, T., Zhan, X., & Yu, W. (2024). The impact of AI perception on employees' psychology and behavior and its theoretical explanations. Advances in Psychological Science, 07.

Westerman, D., Edwards, A. P., Edwards, C., Luo, Z., & Spence, P. R. (2020). I-it, I-thou, I-robot: The perceived humanness of AI in human-machine communication. Communication Studies, 71(3), 393–408.

Wong, J. W.-E., & McGee, K. (2012). Frown more, talk more: Effects of facial expressions in establishing conversational rapport with virtual agents. In Y. Nakano, M. Neff, A. Paiva, & M. Walker (Eds.), Intelligent Virtual Agents (pp. 419–425). Springer.

Xie, T., & Pentina, I. (2022). Attachment theory as a framework to understand relationships with social chatbots: A case study of replika. Proceedings of the 55th Hawaii International Conference on System Sciences, 2046–2055.

Zhao, R., Romero, O. J., & Rudnicky, A. (2018). SOGO: A social intelligent negotiation dialogue system. Proceedings of the 18th International Conference on Intelligent Virtual Agents, 1–8.