Why the most important questions about artificial intelligence are people-centered.
Have you noticed how quickly using AI begins to feel ordinary?
The first few times, the interaction may still feel novel. You ask for help drafting an email, summarizing a report, explaining a difficult concept, or thinking through a decision that has resisted your own attempts to organize it. The task gets easier, which is useful enough. But after repeated use, the more interesting change may not be the quality of the output. It may be the way the task itself begins to feel different.
People often describe that difference in similar terms. They say they brainstorm differently when AI is available. They check AI before searching the web. They remember fewer details because retrieval is always close at hand. Some feel more confident because they have a responsive system that can help them test an idea before showing it to another person. Others become less certain about where their own judgment ends and the machine’s suggestion begins. Any one of these shifts can be explained as an individual habit, but the pattern becomes harder to dismiss when similar accounts appear across classrooms, workplaces, families, and professions.
That is usually the point at which my attention sharpens. When many people independently report similar changes, I become less interested in treating those changes as personal quirks. The more useful question is what conditions are producing them. AI does not simply add a new tool to an unchanged environment. It changes what is available, what is easy, what is rewarded, what can be avoided, and what begins to feel normal. People adapt to those conditions, often before they have language for what has changed.
This is where my interest in AI ethics begins. Most discussions of AI ethics examine the systems themselves. They ask whether models are biased, whether data were collected responsibly, whether regulations are sufficient, or whether developers have built appropriate safeguards. Those are necessary questions because they help us evaluate the technologies we are creating and the institutions that govern them. They do not, by themselves, explain how people change once those systems become ordinary parts of everyday thinking, learning, work, and decision-making.
The questions that interest me begin with that adaptation. What happens when intelligent systems become ordinary companions in cognitive work? How does living alongside AI reshape judgment, trust, memory, relationships, and responsibility? What changes when technologies that once supported human thinking begin participating in the process of thinking itself? These questions point toward a broader understanding of AI ethics, one concerned not only with whether systems fail but with what happens when they become useful, familiar, and difficult to remove from daily life.
The ethical significance of AI does not begin when a system breaks. It begins when people begin reorganizing their habits around its presence. If we want to understand the long-term consequences of artificial intelligence, we need to study not only the systems we are building but also the conditions those systems create and the patterns of human adaptation that follow.
AI Ethics Is More Than Governance
Ask someone what "AI ethics" means and the conversation usually moves in a familiar direction. Before long someone mentions bias, privacy, regulation, copyright, government oversight, or the need for "responsible AI." That response makes sense. These are the ethical questions that receive the most public attention because they produce legislation, corporate policies, lawsuits, and headlines. They are also questions with reasonably clear objects of analysis. We can examine a model for bias. We can audit how data were collected. We can ask whether an organization complied with a particular regulation or whether appropriate safeguards were put in place.
There is nothing wrong with approaching AI ethics this way. These questions matter because they establish the conditions under which intelligent systems are designed, deployed, and governed. If those conditions are neglected, the consequences can be significant. It would be difficult to argue that governance is anything other than an essential part of AI ethics.
What interests me is what happens after those questions have been asked.
Imagine an AI system that has been carefully designed, responsibly deployed, and legally compliant. The governance questions have largely been answered. The system enters a workplace, a classroom, or a home and gradually becomes part of everyday life. At that point something begins happening that governance alone cannot fully explain. People start reorganizing their behavior around the new possibilities the system creates. Students study differently. Professionals approach writing differently. Teams begin solving problems differently. Over time these adjustments become ordinary enough that people often stop noticing they are adapting at all.
This strikes me as a different kind of ethical question. The system may be functioning exactly as intended, yet its presence is still changing the conditions under which people think, learn, trust, and make decisions. The explanation for those changes cannot be found entirely in the model or entirely in the individual. It emerges through the relationship between the two.
Once the question becomes one of adaptation rather than deployment, the scope of AI ethics begins to expand. We are no longer asking only whether a system complies with policy or satisfies a regulatory framework. We also begin asking how trust develops, what habits become easier to form, how judgment changes when thinking becomes collaborative, and under what conditions appropriate reliance gradually becomes dependence. These questions are not prompted by technological failure. They arise because the technology has become useful enough to become ordinary.
From that perspective, governance remains an indispensable part of AI ethics, but it is no longer the whole field. Understanding AI ethically requires understanding not only how intelligent systems should behave, but also how people and institutions change once those systems become woven into everyday life. The object of study is no longer the technology alone. It is the changing relationship between people, intelligent systems, and the environments they now inhabit together.
AI Changes More Than Work
Much of the conversation about artificial intelligence revolves around work. Will it make us more productive? Which jobs will change? Which professions will disappear? These are reasonable questions because new technologies have always altered the way work is organized. Looking only at work, however, has a way of making every technological change appear primarily economic. The more interesting changes often happen somewhere else.
Consider the technologies that gradually became ordinary over the past few decades. GPS made navigation easier, but it also changed what people needed to remember. Search engines made information easier to retrieve, but they also changed what it meant to know something. Spell check did more than catch typographical errors. It subtly changed the relationship between writers and the act of writing itself. Recommendation systems made discovery more efficient, yet they also began selecting what deserved our attention before we had decided to look for it. None of these technologies simply reduced effort. Each altered the conditions under which a particular cognitive task took place.
Taken individually, these changes seem modest. Together they suggest a different pattern. The technologies people tend to remember for their efficiency often leave a more lasting mark on habits of thought. They redistribute small pieces of cognitive work from the individual to the surrounding environment. Navigation becomes something we consult rather than remember. Information becomes something we retrieve rather than retain. Discovery becomes something we receive rather than pursue. The technology does not eliminate the human activity. It changes the conditions under which that activity occurs.
Generative AI appears to belong to this same history, but it also seems to extend it in an important way. Earlier technologies generally supported thinking from the outside. They helped people calculate, search, navigate, or correct. Increasingly, AI participates in the thinking itself. It suggests interpretations, develops arguments, identifies patterns, critiques ideas, proposes alternatives, and helps organize reasoning before a conclusion has been reached. The distinction is subtle, but it changes the relationship between person and technology in ways that earlier information systems did not.
That observation helps explain why conversations about AI ethics feel different from conversations about earlier technologies. The central question is no longer whether a tool performs its intended function. It is whether repeated interaction with that tool gradually reshapes judgment, memory, learning, trust, and decision-making. Those changes do not announce themselves all at once. They emerge slowly through ordinary use, becoming visible only after enough individual experiences accumulate into a recognizable pattern.
Seen this way, the ethical significance of AI lies not only in what the technology produces but also in what it reorganizes. If intelligent systems increasingly participate in activities that have long been understood as distinctly human forms of cognition, then AI ethics must concern itself with more than technical performance or workplace efficiency. It must also ask how those new conditions influence the ways people come to think, remember, learn, and understand the world.
Ethical Responsibility Is Distributed
Public conversations about AI failures tend to follow a familiar pattern. A story appears about an inaccurate recommendation, a biased hiring tool, a student using AI inappropriately, or a company deploying a system that produces unexpected consequences. The discussion quickly shifts away from the event itself and toward responsibility. Someone argues that the user should have exercised better judgment. Someone else blames the company that built the system. Others insist that regulators should have acted sooner. Although the conclusions differ, the conversation usually shares one assumption: somewhere there is a single place where responsibility properly belongs.
That assumption becomes more difficult to sustain the longer AI remains part of everyday life. Most interactions with intelligent systems are not isolated decisions. They are shaped by habits that develop over time, by systems designed long before any individual begins using them, and by institutions that establish what counts as acceptable, expected, or even possible. Looking for a single point of responsibility often tells us less about the problem than about our preference for simple explanations.
Take the individual using AI. A student decides whether to submit AI-generated work. A physician decides how much confidence to place in a clinical recommendation. A manager decides whether an AI-generated performance review deserves further scrutiny. These are clearly ethical decisions because they involve judgment, responsibility, and trust. At this level the question is not simply whether people use AI. It is how they decide when its recommendations deserve confidence and when independent judgment should prevail.
Those decisions, however, never begin with the individual. They begin much earlier, in the systems people encounter. Every model reflects thousands of design decisions concerning training data, optimization goals, interfaces, safety mechanisms, and acceptable trade-offs. Bias, explainability, privacy, and transparency do not emerge from users. They are properties of the systems users inherit. If similar patterns of confusion, overreliance, or misuse appear across many different people, it becomes difficult to explain those patterns solely through individual choice. The conditions created by the technology itself begin to matter.
Those conditions are, in turn, shaped by institutions. Universities establish policies governing acceptable AI use. Employers decide when AI should assist or replace human judgment. Governments determine regulatory frameworks. Technology companies decide which products are developed, what data are collected, and which values become embedded in commercial systems. These decisions create the environment within which both people and technologies interact. Institutions rarely determine individual behavior directly, but they shape the conditions under which certain behaviors become more likely than others.
What begins to emerge is not a hierarchy but a system. Institutions shape the systems people encounter. Those systems influence habits, expectations, and patterns of judgment. As those patterns become widespread, they reshape the institutions that produced them through new policies, cultural expectations, market pressures, and political demands. Responsibility does not sit permanently at any one point within that system because each part continually influences the others.
This is one reason I think AI ethics is often reduced too quickly to questions of governance or individual responsibility. Both matter. Neither is sufficient on its own. The ethical consequences of AI emerge through the interaction among people, systems, and institutions, making the relationship, not any single actor, the more appropriate object of study.
The Questions Matter More Than the Answers
People sometimes ask what my research has discovered. It is a reasonable question, but I have found myself answering it differently over the years. Rather than beginning with conclusions, I usually begin with the questions. That is not because answers are unimportant. It is because answers belong to a particular moment. Technologies change. Institutions adapt. New practices emerge. A satisfying answer today may become incomplete tomorrow. Well-constructed questions tend to survive those changes because they continue directing our attention toward the underlying mechanisms rather than the temporary forms they happen to take.
Those questions rarely emerge from theory alone. They usually begin with observations that seem too common to dismiss and too consistent to ignore. A teacher notices students approaching writing differently. A manager realizes that AI has quietly become part of everyday decision-making. Someone describes trusting an AI assistant more than they expected. Someone else realizes they no longer remember information because retrieving it has become easier than retaining it. None of these experiences proves anything on its own. The interesting question is why similar observations continue appearing among people who have never met, work in different professions, and encounter AI under very different circumstances.
Once those experiences begin repeating across classrooms, workplaces, healthcare, and public life, they become difficult to explain as isolated decisions. They begin suggesting that the surrounding conditions are changing. My research questions usually emerge at that point. Rather than asking whether AI is good or bad, I find myself asking what these recurring observations reveal about the relationship people are developing with increasingly intelligent systems.
One pattern, for example, appears whenever people describe AI as helping them think rather than simply helping them find information. Earlier technologies generally supported decisions by making information easier to access or calculations easier to perform. Increasingly, AI contributes ideas, organizes reasoning, proposes alternatives, and participates in the process that leads to a decision. That observation raises a different question. What happens when technologies that once supported judgment begin participating in it?
A similar pattern appears whenever people describe changes in how they understand the world around them. Recommendation systems influence what is encountered. AI assistants influence how problems are framed, how evidence is evaluated, and which possibilities seem worth pursuing. The individual experiences vary, but together they suggest that AI is mediating more than information. It is gradually becoming part of the environment through which people construct knowledge, trust expertise, and interpret reality.
The same shift appears when attention moves from systems to experience. Technical measures tell us whether a model performs accurately, efficiently, or reliably. They tell us much less about what it feels like to live with that model over months or years. Trust, uncertainty, confidence, frustration, dependence, and habit often appear in conversations about AI long before they appear in technical evaluations. That difference suggests another question. How do people experience intelligent systems, not simply use them?
The pattern appears again when institutions enter the picture. Organizations typically evaluate AI through governance, compliance, risk, and measurable performance. Individuals encounter the same systems through fairness, understanding, trust, and agency. Neither perspective is mistaken. They are simply observing the system from different locations. That difference, too, becomes a question worth following rather than resolving too quickly.
I do not expect these questions to produce permanent answers. Their value lies elsewhere. They help identify changes while those changes are still becoming ordinary. They provide language for experiences that many people recognize but have not yet connected to one another. Once those connections become visible, the object of inquiry begins to change. The question is no longer simply what AI is doing. It becomes what kinds of social conditions are emerging as people, institutions, and intelligent systems learn to live together.
Following the Pattern
Looking back, I don't think my work began with artificial intelligence. It began much earlier with a recurring observation that appeared in very different settings. Organizations often assumed that better technology would naturally produce better outcomes. Sometimes it did. Just as often, it did not. Systems that appeared technically successful failed to gain adoption. Other technologies became indispensable for reasons their designers had never anticipated. The technical explanation was often sound, yet it seemed to leave something important unexplained.
Over time, those experiences began to resemble one another. They suggested that technical performance and human response were related but not identical. Two systems with similar capabilities could produce very different patterns of use. People trusted one but ignored another. They incorporated one into everyday routines while resisting the other. If those differences appeared repeatedly across organizations, technologies, and contexts, then the explanation seemed unlikely to reside in the technology alone. Something about the relationship between people and systems deserved closer attention.
That line of inquiry eventually led me to artificial intelligence, although AI itself was never the destination. It became another environment in which the same questions appeared in new forms. People described AI as a coworker, a tutor, a brainstorming partner, or, in some cases, as a companion. Others rejected the technology despite acknowledging its capabilities. Neither response seemed especially surprising on its own. What drew my attention was that these patterns emerged repeatedly among people with different backgrounds, different professions, and different reasons for using AI.
Following those patterns gradually led my research in several directions. Trust became one area of study because confidence in a system often proved to be a better predictor of adoption than technical capability alone. Relational AI became another because people consistently formed social responses to systems they knew were artificial. Affect emerged as a third because feelings such as confidence, uncertainty, familiarity, or frustration were not simply reactions to technology. They became part of the conditions under which people decided whether to rely on a system, resist it, or incorporate it into everyday life.
The same process eventually widened again. Individuals never encounter AI in isolation. Their experiences are shaped by schools, workplaces, healthcare systems, governments, and technology companies that establish the conditions under which intelligent systems are introduced, encouraged, restricted, or normalized. More recently, similar questions have appeared in work on AI coworkers and AI companions. These may seem like distinct topics, yet they continue asking variations of the same underlying question: how do people adapt when increasingly intelligent systems become ordinary participants in everyday life?
Seen from a distance, these research areas may appear unrelated. To me, they have always felt like different observations of the same phenomenon. Each begins with an experience that many people recognize. Each becomes more interesting when that experience appears repeatedly across different settings. Each eventually leads away from individual behavior and toward the systems, institutions, and conditions that make those patterns possible. Artificial intelligence happens to be the environment I study today. The larger question has remained remarkably consistent: how do people adapt when the conditions of everyday life begin to change?
Conclusion
Following AI ethics through these questions gradually changes the object of inquiry. The conversation often begins with models, algorithms, governance, or regulation because those are the most visible expressions of the technology. They remain essential subjects of study. Systems should be safe. Data should be collected responsibly. Institutions should establish meaningful forms of accountability. None of those concerns disappear. They simply become part of a larger picture.
That larger picture comes into view once intelligent systems become ordinary enough that people begin reorganizing their lives around them. The ethical questions no longer concern only the technology itself. They concern the conditions the technology creates. People learn differently when AI becomes part of learning. They remember differently when retrieval becomes easier than recall. They make decisions differently when reasoning becomes collaborative rather than solitary. These changes rarely occur because anyone intended them. They emerge gradually as people adapt to a changing environment.
Seen from that perspective, AI ethics begins to look less like the study of intelligent machines than the study of increasingly AI-mediated societies. The technology remains important, but it is no longer the only subject. The relationships among people, intelligent systems, and institutions become equally important because those relationships shape how the technology is understood, adopted, resisted, and ultimately incorporated into everyday life. The ethical questions therefore move beyond what a system is capable of doing and toward what kinds of habits, expectations, and forms of social life that capability gradually makes possible.
I suspect this is why the field continues to expand. Earlier generations often asked how technology should be used. Increasingly, we find ourselves asking what happens when technology participates in activities that have traditionally belonged to human judgment, communication, creativity, and decision-making. Those questions are unlikely to disappear as AI becomes more capable because they are no longer only questions about technology. They are questions about the societies that emerge as people adapt to living alongside increasingly intelligent systems.
Perhaps that is where AI ethics has been heading all along. Not away from governance or technical design, but beyond them. Understanding intelligent systems will continue to matter. Understanding the changing conditions under which people live with those systems may matter just as much.
