Relational AI: Understanding the Human Side of Artificial Intelligence
Artificial intelligence is often evaluated by capability. Can it generate text? Can it recognize images? Can it write software, summarize documents, answer questions, or produce something that looks enough like expertise to be useful?
Relational AI begins from a different observation.
Increasingly, people are not only using AI systems. They are talking to them, confiding in them, arguing with them, trusting them, correcting them, thanking them, and sometimes treating them as if someone is there.
That shift tells us something important.
People do not usually describe these interactions in neutral technical language. They say the AI “understood” the assignment, “digested” the data, “hallucinated” an answer, “refused” a request, or “isn’t listening” today. The language may be casual, but it reveals something important. As AI systems become more conversational, responsive, and persistent, people begin to interpret them through social categories that were not designed for tools.
Whether the system is a chatbot, an AI tutor, a workplace assistant, a customer service agent, or a social robot, people routinely attribute personality, intention, reliability, competence, mood, and even companionship to artificial systems. These responses are not strange in the way they are sometimes described. They are predictable features of human social cognition. Humans are pattern-making, meaning-making, relationship-forming creatures. We read responsiveness as social information. Relational AI is the study of those interactions.
Throughout this site, I use Relational AI to refer to the study of how humans form relationships characterized by trust, emotion, and social meaning with increasingly responsive artificial systems. My work is particularly concerned with the individual, organizational, and societal consequences of those relationships.
This is different from simply asking whether a system works. Traditional human-computer interaction has often focused on usability, interface design, task completion, and user experience. Those questions still matter. Relational AI asks what happens after the system becomes familiar. How does repeated interaction shape trust? When do people begin treating an artificial system as a social partner? How do design choices encourage reliance, intimacy, deference, or emotional attachment? What responsibilities follow when systems are intentionally designed to feel attentive, supportive, or companionable?
These questions extend far beyond AI companionship. Similar relational mechanisms appear in education, healthcare, organizational AI, customer service, digital assistants, and workplace tools. A student may rely on an AI tutor not only for answers, but for reassurance. An employee may begin to treat an AI assistant as a judgment partner. A patient may experience a health chatbot as a nonjudgmental presence. A customer may respond differently to a system that seems patient, apologetic, or concerned. In each case, the technology is not only completing a task. It is shaping expectations about attention, competence, care, and trust.
Much of my research examines these questions through three connected areas.
Loneliness, Attachment, and Emotional Substitution considers how artificial systems become meaningful companions and what happens when simulated responsiveness enters the territory of emotional need.
Affect in Human-AI Systems examines why emotion is not separate from technology use, but part of how many systems function. People do not make decisions about technology through cognition alone. Confidence, frustration, relief, dependence, curiosity, and anxiety all become part of the system in practice.
Trust, Reliance, and Quiet System Failure explores how relationships with technology shape adoption, decision-making, and organizational outcomes. Systems do not only fail when they break. They also fail when people trust them too much, too little, or in the wrong way.
Together, these questions form a broader research program about how humans adapt to emerging technologies. Relational AI is one part of that larger inquiry, but it is becoming harder to treat as a niche concern. As artificial systems become more conversational, personalized, persistent, and embedded in everyday life, the relationship between human and machine becomes part of the system itself. That is why relational AI matters.
The central question is not whether people should form relationships with AI. They already are. The more useful question is how those relationships are being designed, interpreted, governed, and normalized.
A Public Conversation
This site focuses on research, frameworks, and public scholarship. I also write for a broader audience at Synthetic Relations, where I explore the cultural, psychological, and ethical dimensions of AI companionship in a more essayistic form.
If you are interested in how these ideas appear in everyday experience, read:
→ "Relational AI: What's Actually Happening When We Befriend Machines" on Synthetic Relations
That essay explores the same terrain through stories, examples, and lived experience rather than research language.
Relational AI in Practice
Relational AI provides a lens for understanding a wide range of human-AI interactions, including:
- AI companions
- AI coworkers
- AI tutors
- Healthcare AI
- Customer service systems
- Digital assistants
- Social robots
These technologies differ in purpose, but they often engage the same underlying psychological and social mechanisms.