I approach relational AI as the study and design of AI systems that shape the conditions under which people interpret, decide, learn, create, participate, and remain accountable for what follows.
This work extends beyond companion AI and emotionally intimate systems. It focuses on everyday human–AI systems from AI tutors and research assistants to workplace agents, governance simulations, and institutional tools.
This research program examines the roles AI systems take on across human activity: tutor, assistant, collaborator, agent, simulator, evaluator, or institutional support layer. The focus is not only what these systems produce, but how their participation shapes judgment, agency, and responsibility.
Relational AI is not limited to companion bots, emotional attachment, or systems that imitate intimacy. AI also becomes relational when it structures attention, frames choices, redistributes authority, or changes how people understand themselves, others, and their obligations.
The aim is to keep human agency, interpretive integrity, role-bounded participation, contestability, and traceable accountability intact as AI systems enter learning, knowledge work, governance, creativity, and decision-making environments.
Many AI systems reshape human agency without looking intimate, social, or companion-like. They become relational when they alter what people notice, what choices feel available, whose judgment carries weight, and where responsibility appears to sit.
This research program starts from a simple premise: AI does not need to feel like a companion to participate in human life. A tutor can shape how a learner understands their own ability. A research assistant can influence what evidence feels relevant. A writing system can shift authorship and voice. A workplace agent can redirect who appears responsible for action. A governance simulation can frame what futures seem plausible.
Companion systems make relational dynamics visible, but they do not define the full terrain. Relational AI also includes everyday systems that structure attention, mediate judgment, guide learning, scaffold decisions, or change how people understand their roles.
A system can produce acceptable outputs while still shaping a user’s assumptions, confidence, dependencies, or sense of responsibility over time. Relational evaluation therefore has to examine interaction patterns, not only isolated responses.
Tools become consequential when they help frame problems, rank options, mediate trust, or distribute accountability. The question is not whether AI is a person, but what kind of participation its design makes possible.
This program does not begin by asking whether AI is human-like, emotionally convincing, or deserving of moral standing. It asks whether AI participation preserves the conditions for human agency, interpretation, accountability, and trust.
The central concern is authority migration: the subtle movement of framing, judgment, authorship, responsibility, or legitimacy from people and institutions into AI-mediated systems. Relational AI, in this sense, is not about treating AI as a person. It is about understanding when AI begins to participate in the relationships, roles, and decisions that shape human life.
AI systems do more than provide information. They can frame what a situation means, which evidence appears relevant, what options feel reasonable, and how people understand their own judgment. This makes interpretation a central governance concern.
Systems often begin as assistants, tutors, co-writers, or advisors, then gradually take on functions closer to evaluator, strategist, gatekeeper, confidant, or decision-shaper. Relational integrity requires making those role shifts visible and contestable.
The question is not simply whether an AI system is useful or aligned in a given moment. The question is whether its participation remains bounded, transparent, accountable, reversible, and appropriate to the context in which it is being used.
These commitments guide how I assess whether AI participation remains bounded, accountable, and supportive of human agency across learning, judgment, creativity, governance, and institutional life.
AI systems should expand a person’s capacity to think, choose, question, and act, not quietly narrow their sense of what is possible, appropriate, or worth considering.
Systems should make their framing visible enough for users to inspect, revise, or reject it. Interpretation should remain a shared and contestable process, not something silently absorbed by the system.
AI should remain clear about what role it is playing: tutor, assistant, co-reader, collaborator, advisor, simulator, or support layer. When that role changes, the shift should be visible.
People need meaningful ways to question, redirect, override, or exit AI-shaped pathways. Relational integrity depends on preserving room for disagreement and course correction.
When AI contributes to decisions, recommendations, interpretations, or institutional actions, responsibility should not disappear into the system or scatter across workflows, models, and organizations. Human and organizational accountability must remain traceable.
AI design should account for the relationships it reshapes: between learners and teachers, workers and institutions, citizens and systems, creators and audiences, people and their own judgment.
This approach is designed for everyday systems that shape learning, work, creativity, governance, and institutional judgment, not only systems built for companionship.
Educational AI can shape how learners understand their abilities, questions, mistakes, and progress. Relational integrity asks whether support strengthens agency, reflection, and learner-led interpretation.
AI research support can influence what sources feel relevant, which claims appear central, and how uncertainty is handled. The concern is not only accuracy, but how inquiry itself is framed.
Co-writing systems can help people clarify ideas, but they can also reshape voice, authorship, confidence, and judgment. Relational AI asks how collaboration preserves the human author’s interpretive role.
AI agents in professional settings can route tasks, recommend actions, summarize people’s work, or influence what counts as performance. These systems require clear role boundaries and traceable responsibility.
Scenario-based AI systems can help people rehearse decisions under uncertainty, but they also frame which futures, stakeholders, and tradeoffs become visible. The design of the simulation shapes the imagination of governance.
AI systems used in education, civic life, hiring, care, or public administration can redistribute authority between people, policies, and platforms. Relational integrity asks whether accountability remains traceable and contestable.
The longer-term goal is to make relational integrity something people can recognize, design for, evaluate, and govern in the real AI systems they already use to learn, work, create, decide, and participate.
Build evaluation methods that assess how AI systems affect human agency, interpretive integrity, role-bounded participation, contestability, and traceable accountability as interaction deepens over time.
Translate relational AI concepts into prompts, interfaces, protocols, and institutional practices that keep AI participation visible, bounded, contestable, and accountable within real contexts of use.
Create indicators and review practices that help educators, researchers, organizations, and public-interest teams assess when AI systems reshape roles, judgment, responsibility, or trust.
These questions guide the central concern of this work: how AI systems participate in human judgment, agency, responsibility, and institutional life as their roles become more consequential.
At what point does an AI system move from supporting a task to shaping what people notice, trust, choose, or feel responsible for?
How can systems make their roles clear while keeping AI participation bounded, contestable, accountable, and appropriate to the context in which it is used?
What methods can assess whether AI systems preserve agency, interpretive integrity, accountability, and trust as interaction deepens over time?
Closing thesis
“Relational AI is not only about whether humans form bonds with machines. It is about whether AI systems reshape agency, interpretation, responsibility, and trust with integrity.”
This page names a developing research direction: a way of studying and designing AI systems that participate in human meaning-making, judgment, learning, governance, and institutional life.
The aim is not to claim that all AI systems are relational in the same way. It is to ask more carefully when AI begins to shape roles, relationships, responsibility, and participation, and what forms of accountability should follow.
I welcome conversations with researchers, educators, designers, institutions, and public-interest teams exploring how AI systems can support human judgment, participation, and responsibility without quietly replacing them.