How emotional and social AI are reshaping human–machine relationships

A new wave of research suggests that the future of human–computer interaction will not be defined by efficiency alone, but by trust, emotional connection, and even forms of perceived intimacy between humans and intelligent systems. A review maps this transformation, tracing how human–AI relationships are moving beyond functional interaction toward deeper, socially embedded partnerships.

Published in AI & Society, the study titled “Advancing human–AI teams: evolving from instrumental tools to trusted partners” examines over a century of technological evolution and compiles 134 sources to identify how human–computer interaction has transitioned through multiple paradigm shifts toward what researchers call a “coexistential AI era”.

From tools to partners: the four-stage evolution of human–AI interaction

The study identifies a clear historical trajectory in how humans engage with computing systems, structured across four major phases. Each phase reflects not only technological advancement but also a shift in how humans conceptualize their relationship with machines.

In the earliest phase, known as the Equipment Era, computers functioned strictly as tools. During this period, spanning the mid-twentieth century, systems were designed primarily for efficiency and accuracy. Human interaction was limited, highly technical, and often inaccessible to non-specialists. The relationship between humans and machines was purely instrumental, with no expectation of interaction beyond task execution.

The transition to the Interactive System Era in the late twentieth century marked a turning point. The rise of personal computing introduced graphical interfaces and real-time feedback, making systems more accessible and user-friendly. Humans began to interact with computers as dialog partners, engaging in two-way exchanges that emphasized usability and experience rather than raw performance.

The emergence of AI led to the Autonomous Agent Era, where machines gained the ability to act independently and make decisions. AI systems began to assist humans in complex tasks, from recommendation engines to early conversational agents. This phase introduced new concerns around trust, transparency, and explainability, as users increasingly relied on systems capable of autonomous reasoning.

The most recent phase, described as the Coexistential AI Era, represents a fundamental shift. In this stage, AI systems are evolving into co-creators and collaborative partners. Rather than simply executing tasks or offering recommendations, they actively participate in shared goals, adapting to human behavior and contributing to decision-making processes. This transformation is driven by advances in generative AI, natural language processing, and affective computing, which enable systems to engage in more natural, context-aware interactions .

This progression highlights a key insight: the role of AI is expanding from functional utility to relational engagement. As machines become more integrated into human workflows, the quality of interaction becomes as important as the outcome.

Emotional intelligence and anthropomorphism redefine human–AI relationships

One of the most significant developments is the growing role of anthropomorphism and emotional intelligence in shaping human–AI interaction. Modern AI systems are increasingly designed to mimic human-like behavior, including emotional responsiveness, conversational nuance, and social awareness.

Anthropomorphism, defined as the attribution of human traits to non-human entities, has become a central design feature in AI systems. Research shows that when users perceive AI as human-like, they are more likely to trust it and engage with it more deeply. This has direct implications for adoption across sectors such as healthcare, education, and customer service, where trust is a critical factor.

The integration of affective computing allows AI systems to detect, interpret, and respond to human emotions. These capabilities enable more personalized and empathetic interactions, transforming the user experience from transactional to relational. AI systems can now simulate emotional understanding, respond to user sentiment, and adapt their behavior accordingly.

However, the study highlights that these developments come with complex and sometimes contradictory outcomes. While emotional AI can enhance user satisfaction and engagement, it can also create psychological tensions. For instance, users may feel more understood when interacting with AI-generated responses, but their perception of value decreases when they are aware that the interaction is machine-generated .

This paradox underscores a broader challenge: the line between authentic and simulated interaction is becoming increasingly blurred. As AI systems become more sophisticated in mimicking human behavior, users may form emotional attachments that are not reciprocated in any meaningful sense. These pseudo-intimate relationships raise important questions about the long-term psychological and social effects of human–AI interaction.

The study also finds that not all forms of anthropomorphism produce positive outcomes. Different human-like traits can have varying effects on user behavior. While cognitive and rational traits may reduce anxiety and improve trust, emotional and moral traits can sometimes trigger discomfort or identity-related concerns. This suggests that the design of human-like AI must be carefully calibrated to balance engagement with ethical considerations.

Intimacy, collaboration, and the risks of human–AI partnership

The study argues that concepts such as efficiency, accuracy, and even trust no longer fully capture the complexity of human–AI interaction. Instead, new relational metrics are emerging, including intimacy, mutual adaptability, and social bonding. These measures aim to assess the depth and quality of interaction between humans and AI systems, reflecting a shift toward more holistic evaluation frameworks.

Intimacy, in particular, is identified as a critical but controversial dimension. In human–AI contexts, intimacy is not based on mutual emotional experience but on the human perception of connection. Users may disclose personal information, seek emotional support, and develop attachment-like behaviors toward AI systems, even though these systems lack genuine consciousness or भावना.

This asymmetry introduces significant ethical and methodological challenges. Measuring intimacy in human–AI relationships may effectively measure the strength of an illusion rather than a reciprocal bond. This raises concerns about manipulation, especially if systems are designed to maximize user engagement by simulating emotional closeness.

The study warns that optimizing AI for perceived intimacy could lead to unintended consequences, including dependency, social isolation, and reduced human-to-human interaction. Vulnerable populations, such as children and individuals with mental health challenges, may be particularly at risk of forming unhealthy attachments to AI systems.

Further, the research highlights potential benefits. AI systems can provide companionship, reduce loneliness, and offer emotional support in contexts where human interaction is limited. The challenge lies in balancing these benefits with safeguards that prevent harm.

Apart from individual interaction, the study examines the rise of multi-agent systems, where multiple AI entities collaborate with humans and with each other. These systems are already being deployed in sectors such as healthcare, transportation, and disaster response, where coordinated decision-making is essential.

In healthcare, AI agents assist with patient monitoring, diagnosis, and care coordination. In transportation, they optimize traffic flow and enable autonomous vehicle systems. In disaster response, they support search-and-rescue operations through coordinated robotic and sensor networks.

Despite their potential, these systems introduce new complexities. The study identifies challenges related to transparency, coordination, and user trust. In some cases, the introduction of AI teammates has been found to reduce human collaboration and trust rather than enhance it, contradicting assumptions about the benefits of automation .

This finding highlights a critical tension in the development of human–AI teams. While AI has the potential to augment human capabilities, poorly designed systems may disrupt existing social dynamics and reduce overall effectiveness.