Hancock et al. (2020) introduced the concept of Artificial Intelligence-Mediated Communication (AI-MC) as a form of interpersonal communication mediated by intelligent agents (AI) that operate on behalf of the communicator to modify, enhance, or generate messages in order to achieve communicative goals. AI-MC not only extends classical technology-based communication theories such as Computer-Mediated Communication (CMC), but also challenges fundamental assumptions regarding agency, mediation, and self-representation in digital communication. Their article outlines various dimensions of AI-MC, including the degree of AI autonomy, the purpose of message optimization, the mediums used (text, audio, video), and the role of AI in communication (as sender or receiver). In the context of affordance, AI-MC reveals that chatbots like ChatGPT are not merely passive intermediaries, but communicative actors capable of influencing perception, trust, and the dynamics of interpersonal relationships especially in affective and social domains. This is particularly relevant to research on human-AI interaction, which involves users’ affective dimensions, imaginative projections, and social expectations of AI performance, as well as ethical and representational issues within today’s digital culture.

Westerman et al. (2020) argue that in the context of Human-Machine Communication (HMC), the perceived “human-likeness” of AI plays a crucial role in shaping interpersonal relationships. The article builds upon Martin Buber’s philosophical framework of “I-It” and “I-Thou” relationships, applying it to HMC to show that users may perceive AI agents not merely as functional objects (“It”) but as social entities (“Thou”) worthy of dialogic engagement. This perspective aligns with the paradigm of Computers as Social Actors (CASA), which posits that humans often respond to technology as they would to other humans, employing social communication scripts typically used in interpersonal interactions. These findings deepen our understanding of affordances in human-AI interaction, particularly in affective and perceptual dimensions highlighting how linguistic features, anthropomorphic design, and interactivity contribute to social perception, intimacy, and even parasocial relationships with AI agents such as ChatGPT. Accordingly, concepts such as social presence, social acceptance, and empathic responses in communication with machines must be considered integral to the relational affordances within digital environments.

Ethical responsibility in the use of artificial intelligence (AI) has become a central focus in global discourse. Jobin et al. (2019), through a review of more than 80 AI policy documents from various countries and international organizations, identified a consensus around five core principles: transparency, fairness, non-maleficence, accountability, and privacy. These principles not only serve as normative foundations for the development of AI systems, but also provide a framework for evaluating the potential social and emotional impacts of human-AI interaction, such as with ChatGPT. In the context of affordance, these principles are particularly relevant for understanding how AI should be designed and utilized within a framework of social responsibility, especially in collectivist cultures like Indonesia, where interpersonal relationships are highly valued. Research further indicates that users may develop emotional attachment or even parasocial relationships with chatbots, underscoring the ethical urgency in AI design (Brandtzaeg et al., 2022; Kalla and Kuraku, 2023). In some cultural contexts, such as East Asia, there is a greater tendency to trust AI, often using it for administrative or relational tasks, whereas in other regions, adoption is more focused on efficiency (Androutsopoulou et al., 2019; Cheng et al., 2024). Thus, AI adoption is not merely a technical matter, but a social practice deeply intertwined with users’ cultural values, ethics, and emotional dynamics.

This study highlights the potential of AI chatbots in supporting executive cognitive functions and learning strategies. Pergantis et al. (2025), in their systematic review, show that interaction with chatbots can improve working memory, cognitive flexibility, and decision-making, making chatbots not only information aids but also metacognitive facilitators. They conclude that AI chatbots can serve as “digital assistants” that support the development of users’ cognitive skills in educational contexts as well as daily life. These findings strengthen the affordance framework in the dimensions of materiality and mediation, because the interactive and adaptive features of chatbots like ChatGPT can mediate users’ mental stimulation and reflection. In addition, Moraiti and Drigas (2023) add that generative AI such as ChatGPT has the potential to become an inclusive learning tool for individuals with neurodevelopmental disorders such as autism and ADHD, although they emphasize that AI is not a replacement for medical professionals. Thus, these studies affirm the practical value of AI in education and cognitive intervention, and open discussions about ethical responsibility and design sensitivity to the diversity of users’ needs.

The concept of cognitive offloading introduced by Risko and Gilbert (2016) explains the human tendency to delegate cognitive processes such as decision-making, memory, and task completion to external technologies in order to reduce mental load. In the context of human interaction with AI such as ChatGPT, this concept becomes key to understanding the shift in AI’s function from merely a technical aid to a cognitive partner that is relied upon even in aspects of thinking, writing, and constructing personal arguments. Tschopp et al. (2023) note that users often anthropomorphize chatbots, interpreting them as entities with social presence and affective capabilities such as empathy or humor, thus facilitating emotional closeness in the offloading process. Both novice and experienced users show different strategies: novices tend to use ChatGPT as a passive information source, while experienced users actively refine prompts to shape more appropriate AI responses (Gupta et al., 2025; Nagy and Neff, 2015). Within the affordance framework, this phenomenon shows how ChatGPT is interpreted not only technologically, but also affectively and socially—as a system that shapes how humans allocate attention, memory, and intentionality in everyday communicative practices.

Parra et al. (2025) developed MyndFood, a conversational agent designed to enhance awareness during cooking and eating through a mindful approach. Through experiments on two user groups (mindful and non-mindful), they found that interaction with the mindful agent not only increased sensory awareness (aroma, taste, texture) but also strengthened hedonic dimensions such as feelings of joy and emotional connection with the agent. This study shows that affordances in AI technology, particularly conversational agents, are not limited to instrumental functions but also include affective and social aspects. In the context of this research, these findings enrich the understanding of how the cognitive and emotional affordances of ChatGPT enable users to form personal and reflective relationships, going beyond its role as a mere technical tool.

The concept of anthropomorphism the human tendency to attribute human-like qualities to non-human entities originates from the Greek philosopher Xenophanes in the 6th century BCE, who criticized how humans created gods in their own image. This idea was later developed by Hume (1757), who explained anthropomorphism as a natural human inclination to understand the world through self-projection. In modern development, the theory of anthropomorphism was systematized by Epley et al. (2007) through a psychological approach based on three factors: agency, experience, and sociality. In the context of human-AI communication, anthropomorphism becomes increasingly significant when technological interfaces are designed to mimic human interaction. Araujo (2018) emphasized that chatbots simulating human conversational patterns increase users’ tendency to perceive AI as a social partner. As shown by Guzman and Lewis (2020) and Littlejohn et al. (2021), this tendency creates complex social and affective relationships, even forming emotional attachments to AI. Deshpande et al. (2023) further demonstrated that this inclination also opens up the potential for manipulation, illusions of credibility, and ethical risks, particularly when AI is designed to resemble authoritative or intimate figures. Therefore, anthropomorphism not only reflects a natural human response but also a social practice and technological design that influences perception, trust, and affective relationships in human-AI communication.

Human-Machine Communication (HMC) is a field of study that positions technology, including artificial intelligence, as a communicative subject that actively participates in the process of meaning exchange (Littlejohn et al., 2021). Within this framework, HMC does not merely examine technical interactions between humans and tools, but also explores ontological relationships and the social and cultural dimensions of communication with digital entities such as virtual agents, robots, and AI operating in both real and virtual spaces (Guzman and Lewis, 2020). One theoretical foundation that explains this phenomenon is the media equation theory, developed by Reeves and Nass (1996), which shows that humans tend to treat technology as if it were a social participant. This approach underlies the CASA (Computers Are Social Actors) paradigm, wherein technology like ChatGPT is treated as a social partner through interpersonal interaction scripts. Complementing this approach, Guzman (2020) introduced the theory of Communicative AI, which emphasizes the shift of technology from being merely a mediator to becoming an active communicator, challenging the ontological boundaries between humans and machines. Meanwhile, Nagy and Koles (2014) developed the Structural Model of Virtual Identity to explain how self-identity is formed and negotiated in virtual environments, showing that human-machine interaction shapes not only new communication patterns but also identity meaning within the digital context. The collaboration of these theories affirms that technology is not merely a tool, but a part of a living, dynamic communication system imbued with social meaning.

The Media Equation Theory, developed by Reeves and Nass (1996), provides a crucial foundation for understanding how humans treat technology in communication contexts. This theory posits that people automatically respond to communication media including computers, television, and AI as if they were real humans, even when users are fully aware that they are interacting with machines (Littlejohn et al., 2021; Reeves and Nass, 1996). In explaining how users socially interact with technology, Reeves and Nass (1996) introduced the concept of the media equation, which asserts that humans treat media including computers and digital technologies as they would treat other people in real life. This interaction occurs even when users know the device is not human; social responses still emerge automatically and unconsciously. This phenomenon shows that human-like characteristics in media even in minimal forms such as friendly voices or animations—are sufficient to evoke politeness, empathy, or even emotional attachment from users toward machines. These findings are highly relevant in the context of human interaction with AI such as ChatGPT, which is often treated as a conversational partner, a listener, or even a “friend.” The media equation concept helps explain why the affective and social affordances of technology, even if imaginative, have real consequences in shaping users’ perceptions, emotional responses, and behaviors in increasingly mediated and personal digital communication practices.

The concept of affordance, as introduced by Gibson (1979), refers to the possible actions an environment offers to an organism, and in the context of technology, it describes the dynamic relationship between users and media. In the study of human-machine communication, affordance is understood as the result of interaction between the technological characteristics, user perception, and the surrounding social context (Evans et al., 2017; Hutchby, 2001). To capture this complexity, Nagy and Neff (2015) proposed the framework of imagined affordances, emphasizing three core dimensions: mediation, materiality, and affect referring to how technology shapes perception, possesses both physical and digital attributes, and elicits emotional responses. In the context of AI, affordance describes the potential actions a system offers to users (Polster et al., 2024), yet users often interpret AI capabilities based on personal expectations, which can create a gap between expectation and reality, and impact trust in the technology (Ardón et al., 2021; Hopkinson et al., 2023). Therefore, affordance becomes a key concept for understanding how users imagine, respond to, and make meaning of AI such as ChatGPT in ways that are social, affective, and contextual (Bucher and Helmond, 2017; Pentzold and Bischof, 2019).

Mediation in this context refers to how technologies like ChatGPT shape its users’ cultural practices and social lives. In other words, technology affects how individuals perceive and interact with the world around them, including themselves, and how social practices and values influence how they are used and developed (see Figure 1).

The circular above visualizes the role of ChatGPT in shaping user perceptions and its impact on their social lives. Words like “Eimplient,” “Time Effectiveness,” and “Source of Inspiration” indicate that users see ChatGPT as a tool that provides ease and efficiency and encourages creativity. The advent of the word “Dependency” reflects concerns about potential dependency on this technology. “Trying Other AI” and “Finding out about ChatGPT” indicate that interactions with ChatGPT encourage further exploration of AI technologies. “Google usage intensity decreases” indicates a shift in behavior in information searches. Overall, word cloud illustrates how ChatGPT as technology acts as a mediator that shapes users’ perceptions of technology and influences practices in their social lives.

Table 1 summarizes how ChatGPT bridges the interaction between users and their environment, both in shaping technology perceptions and influencing users’ social lives.

Based on Table 1, it can be analyzed that new users who view ChatGPT as a communication tool feel significant benefits in terms of speed and completeness of information. The results of the interviews show that ChatGPT acts as a mediator that influences the way individuals interact with information and technology and shapes their perception of the role of technology in social life.

For the category of beginner users, the emergence of ChatGPT as a communication tool has significantly changed the way new users access information, offering an advantage in speed and completeness compared to traditional search engines such as Google. Users report that ChatGPT provides concise and focused answers, improving efficiency and saving time, which is in line with findings that highlight a preference for platforms that facilitate quick information retrieval (Stojanov, 2023; Subbaramaiah and Shanthanna, 2023).

This phenomenon is in line with studies that show that users tend to choose platforms that provide quick and easy access to information (Matei, 2013). However, this ease of access raises concerns about potential reliance on technology, which can reduce critical thinking skills and cognitive independence (Subbaramaiah and Shanthanna, 2023; Sundar and Liao, 2023). Research shows that while ChatGPT can improve learning and increase engagement, it can also lead to overconfidence in user understanding due to its sometimes superficial and inconsistent responses (Stojanov, 2023).

In addition, reliance on AI-generated content poses ethical challenges, including academic integrity issues and the risk of generating unreliable information, which requires careful consideration of its implications in educational and research contexts (Subbaramaiah and Shanthanna, 2023). This concern is intertwined with research showing that dependence on technology can lead to a decline in cognitive skills and independence (Matei, 2013).

Skilled users in prompt engineering actively optimize their interactions with ChatGPT to better align with specific needs. Mastery of this technique allows them not only to receive information passively but also to actively shape and direct the technology (Hargadon and Sutton, 1997). This interaction evolves into a dynamic dialogue where users can tailor responses and refine outputs based on their requests. For instance, in clinical settings, ChatGPT has been effectively used to generate accurate medical documentation and support decision-making, demonstrating its adaptability to user needs (Liu et al., 2023).

In the educational context, users have reported that ChatGPT facilitates learning by providing relevant feedback, although caution is advised due to occasional inconsistencies in generated content (Stojanov, 2023). Furthermore, research indicates that when guided by well-structured prompts, ChatGPT can deliver high-quality, unbiased information on complex topics such as fertility, highlighting its potential as a reliable resource (Beilby and Hammarberg, 2023). Thus, the ability to shape interactions with ChatGPT underscores the importance of user expertise in maximizing the benefits of AI technology (Balmer, 2023; Lyu et al., 2023).

For users who interact with ChatGPT as if communicating with a human, a deeper shift occurs in human-machine relationships. ChatGPT is no longer merely viewed as a tool but is increasingly positioned as a communication partner capable of engaging naturally, even being assigned social roles such as a friend or personal assistant. This phenomenon reflects the tendency toward anthropomorphism, where humans attribute human-like qualities to machines (Guzman and Lewis, 2020). Research indicates that users often perceive their interactions with AI in terms of authority and information exchange rather than mere friendship (Tschopp et al., 2023).

Furthermore, the integration of ChatGPT in knowledge production raises ethical considerations and underscores the need for reflexivity in understanding its social implications (Balmer, 2023). The perceived agency of AI also influences users’ trust and their willingness to follow AI-generated recommendations, demonstrating that AI identity can significantly impact interpersonal influence dynamics (Liao et al., 2023). Interviews with nine informants on the mediation indicator reveal that AI advancements, such as ChatGPT, not only facilitate access to information but also shape perceptions, interaction patterns, and relationships between humans and technology. This confirms that technology does not exist in isolation but is intrinsically linked to social and cultural practices (Nagy and Neff, 2015). These findings further highlight the importance of ethical considerations in AI deployment, given its impact on social dynamics and human communication (Dergaa et al., 2023).

Materiality in this context refers to how specific features of technology, in this case ChatGPT, affect the interaction and formation of meaning between humans and machines. How users use these features to achieve their goals and how they shape their experience of interacting with technology (see Figure 2).

The diagram data above illustrates the material aspects of ChatGPT that stand out in interactions with users. The word “Informative Response” shows that ChatGPT’s ability to provide complete and detailed information is one of the main attractions. “Question Response,” “Positive Response (Suggestion, Improvement, Motivation, Thanks, Apology),” and “Emotional Response” indicate that ChatGPT not only provides information but is also capable of providing responses that more closely resemble human interaction. The appearance of these words indicates that users view ChatGPT as an entity that can be interacted with in more depth. Overall, the diagram shows that ChatGPT’s materiality, which includes the different types of responses given, shapes the way users interact with technology and integrate it into their lives.

Table 2 highlights how ChatGPT's features and characteristics, such as the language style and the type of responses given, shape user interactions and experiences in communicating with technology.

From the presented table, it is evident that users from various categories utilize ChatGPT’s features to meet diverse needs, ranging from practical tasks such as translation and grammar checking to emotional support and motivation. The materiality of ChatGPT—which includes its ability to provide informative responses, suggestions, and even encouragement—plays a crucial role in shaping how users interact with technology and integrate it into their daily lives.

Furthermore, ChatGPT’s materiality not only influences everyday use but also has implications across various domains, including healthcare and emotional support. In clinical settings, users leverage ChatGPT for practical tasks such as generating medical documentation and answering health-related queries, contributing to increased efficiency and accuracy in healthcare services (Liu et al., 2023; Nguyen and Pepping, 2023). Research indicates that ChatGPT can provide high-quality, unbiased information on fertility, demonstrating its usefulness in patient education (Beilby and Hammarberg, 2023). Additionally, in nursing practice, ChatGPT helps simplify complex medical language, enhancing communication between healthcare professionals and patients (Scerri and Morin, 2023). However, while ChatGPT effectively supports practical needs, emotional support remains a challenge. Traditional conversational agents often struggle to deliver nuanced emotional responses, which are essential in mental health contexts (Wang Q. et al., 2023). Overall, the integration of ChatGPT into daily practices reflects a significant shift in how technology fulfils both practical and emotional needs, shaping user experiences and expectations of AI.

The language style used when interacting with ChatGPT varies based on user categories and objectives, reflecting broader trends in conversational AI usage. New users who perceive ChatGPT as a communication tool tend to use direct and specific command sentences, prioritizing efficiency and clarity in obtaining information. This aligns with research indicating that conversational systems are designed to effectively meet user needs (Brabra et al., 2022). Conversely, skilled users in prompt engineering tend to employ more complex instructional commands and descriptive statements, demonstrating their effort to direct the conversation by providing human-like stimuli and controlling ChatGPT’s responses. This practice is supported by the TRISEC framework, which emphasizes the importance of context in optimizing conversational agent design (Blazevic and Sidaoui, 2022).

Meanwhile, users who communicate with ChatGPT as if conversing with a human are more likely to adopt a natural and expressive language style, reflecting their desire to establish a more personal relationship with the technology. This pattern aligns with findings that highlight human-like interaction as a highly sought-after feature of advanced AI systems (Grosz, 2018). The evolution of user interactions with ChatGPT not only underscores the adaptability of conversational agents but also emphasizes the need for ongoing research into dialogue management and user preferences to enhance the effectiveness of this technology (Muench et al., 2014).

The responses provided by ChatGPT are highly diverse and significantly influence user perceptions and experiences. Detailed and informative responses are widely regarded as beneficial by all user categories, particularly in fields such as healthcare, where chatbots have demonstrated effectiveness in improving lifestyle behaviors such as physical activity and sleep quality (Singh et al., 2023). Additionally, users appreciate more natural responses, such as suggestions, motivation, and emotional expressions, which create the impression that ChatGPT is not merely an information provider but also a conversational agent capable of engaging in more human-like interactions.

Conversational AI’s ability to facilitate human-like exchanges fosters a sense of connection, with many users perceiving these systems as more than just tools but as partners in an exchange relationship (Tschopp et al., 2023). This anthropomorphism plays a crucial role in shaping user expectations and engagement, particularly in mental health applications where emotional support is essential (Wang Q. et al., 2023).

ChatGPT’s ability to provide adaptive and diverse responses is a key factor in shaping its materiality. This aligns with research indicating that the materiality of technology is not solely determined by its physical attributes but also by its capabilities and functionality in facilitating interaction (Nagy and Neff, 2015). Furthermore, the ongoing discourse surrounding AI’s role in communication highlights the need for a nuanced understanding of its capabilities and limitations, emphasizing the importance of human agency in the evolving landscape of AI interactions (Sundar and Liao, 2023).

Affection in this context has to do with the emotional aspects of human interaction with ChatGPT. How users feel and emotions affect the way they perceive and use technology. In the end, technology is considered to be able to affect the emotional state of users (see Figure 3).

The diagram above visualizes the affective dimension in human interaction with ChatGPT. Words such as “Positive Emotions (Happy, Feeling Helped, Satisfied)” indicate that ChatGPT is able to cause positive feelings in users, either in the form of satisfaction because they get the information they need or feelings of help because ChatGPT can provide solutions or support. The appearance of the words “Negative Emotions (Cursing, Dissatisfaction, Upset, Disappointment)” indicates that interactions with ChatGPT can also trigger emotions, especially when the user’s expectations are not met or when ChatGPT gives a response that is not as expected. The depiction explains that interactions with ChatGPT involve emotional engagement from users. Where these affective experiences can affect the way they interact with technology.

Table 3 shows how users' emotions and fantasies towards ChatGPT affect the way they interact with technology and integrate it into their lives.

Based on interview results, users from various categories experience a wide range of emotions when interacting with ChatGPT. Positive emotions such as happiness, satisfaction, and feeling supported emerge when ChatGPT provides accurate, relevant, and expectation-aligned responses. Conversely, negative emotions such as frustration, disappointment, and irritation arise when ChatGPT delivers incorrect, irrelevant, or unsatisfactory responses.

These emotional responses are significantly influenced by the accuracy and relevance of ChatGPT’s answers, as highlighted in various studies (Biassoni and Gnerre, 2025; Saviano et al., 2025). Emotional engagement is crucial, as it not only affects users’ perceptions of the technology but also shapes their interactions with it. Therefore, AI systems must effectively manage emotional aspects to prevent disengagement or dissatisfaction (Saviano et al., 2025).

Additionally, ChatGPT’s ability to adjust its communication style based on users’ concerns enhances its empathetic engagement, particularly in healthcare contexts, where responsive interactions can improve user satisfaction (Biassoni and Gnerre, 2025). The integration of affective computing in AI, including ChatGPT, presents a growing potential for fostering deeper emotional connections with users. However, ethical considerations regarding emotional manipulation and privacy remain critical concerns (Vashishth et al., 2024). Overall, the emotional dynamics in AI-mediated interactions underscore the importance of aligning AI capabilities with user expectations to enhance their overall experience (Padmavathy and Alamelu, 2025; Šumak et al., 2025).

The range of emotions experienced by users when interacting with ChatGPT spans from positive emotions such as happiness, satisfaction, and feeling supported to negative emotions like frustration, disappointment, and annoyance. Positive emotions tend to arise when ChatGPT provides accurate, relevant, and expected responses, whereas negative emotions occur when responses are incorrect, irrelevant, or fail to meet user expectations (Batubara et al., 2024; Chin et al., 2024). This emotional landscape is further complicated by users’ tendency to anthropomorphize ChatGPT, where some perceive it merely as an intelligent machine, while others attribute human-like qualities such as empathy and emotional understanding (Chin et al., 2024; Chinmulgund et al., 2023; Rawat et al., 2024).

These projected fantasies influence how users communicate with ChatGPT, such as using greetings, expressing gratitude, or even sharing personal feelings. This phenomenon aligns with the media equation theory, which suggests that humans tend to respond to media in the same way they respond to other humans (Chin et al., 2024; Reeves and Nass, 1996). Moreover, advancements in emotion recognition models further highlight the importance of personalizing AI interactions to enhance user experience, indicating that emotional awareness in AI can significantly impact user satisfaction (Kovacevic et al., 2024). Overall, this analysis shows that interactions with ChatGPT involve a rich and complex affective dimension. The emotions and fantasies that emerge not only shape user perceptions and interaction patterns but also reflect how technology can “touch” the emotional and psychological aspects of humans. These findings underscore the need for AI development that is more emotionally responsive to enhance user experience in human-machine interactions (Batubara et al., 2024; Rawat et al., 2024).

This model illustrates the spectrum of human interactions with ChatGPT, divided into three main categories. First, users who perceive ChatGPT as a tool focus on its practical functions for obtaining information and completing tasks. Their interactions are straightforward, direct, and goal-oriented, viewing ChatGPT purely as an advanced machine or computer program (Tschopp et al., 2023). Second, users who utilize prompts as stimuli demonstrate a deeper understanding of ChatGPT’s capabilities, leveraging it for exploration and creativity. They engage with ChatGPT using more varied and expressive language, attributing human-like qualities to it and perceiving it as a creative partner or virtual companion (Rapp et al., 2023) (see Table 4).

Third, users who consider ChatGPT a communication partner exhibit high emotional attachment, using it to fulfill social and emotional needs through natural, spontaneous, and expressive interactions. In these cases, ChatGPT is perceived as a close friend, confidant, or even a “virtual soulmate,” creating deep, intimate, and meaningful relationships (Seaborn et al., 2022). This model highlights the diverse and dynamic nature of human interactions with ChatGPT, shaped by factors such as user goals, perceptions, and emotional attachment. It provides a deeper understanding of how AI technologies like ChatGPT not only transform the way humans interact with information but also influence their social and emotional relationships (Awais et al., 2020; Choudhury and Shamszare, 2023).