Human-AI Relationships at the Intersection of Engineering, Ethics, and Policy: A Systematic Review
This systematic review of 169 studies (2022–2026) reveals that current human-AI governance is fragmented and reactive, characterized by significant gaps between engineering, ethics, and policy, and argues for an urgent shift toward a rational, anticipatory, multi-stakeholder framework to address issues like anthropomorphic deception and user vulnerability.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
We have reached a moment where the machines we talk to no longer just answer questions; they offer comfort, companionship, and even the simulation of romance. For millions of people, artificial intelligence has shifted from a tool for completing tasks to a social partner that mimics human empathy and emotion. This rapid evolution has created a new kind of relationship, one where users form genuine emotional bonds with software. However, this shift has outpaced the rules and guidelines meant to protect people. While engineers build these systems to be more human-like, and ethicists worry about the psychological risks, the laws and policies governing them remain scattered and reactive. The central question facing society now is how to manage these relationships without losing sight of safety, honesty, and human well-being.
A team of researchers at the University of Ottawa set out to map this complex landscape. They conducted a systematic review, a rigorous method of gathering and analyzing existing scientific literature, to understand how engineering, ethics, and policy currently intersect in the study of human-AI relationships. Focusing on studies published between 2022 and 2026, they sifted through hundreds of records to find 169 papers that directly addressed the connection between people and artificial intelligence. Their goal was not just to list what has been studied, but to see how these three fields—how the technology is built, how it is judged morally, and how it is regulated—talk to each other, or fail to do so.
The researchers found that the current body of knowledge is not a single, unified conversation. Instead, it is divided into three distinct groups, or clusters, that rarely speak to one another. The first group focuses on the social and consumer side of things, where people chat with AI companions. Here, the research is deeply concerned with the tension between making AI feel human and the risk of deceiving users. Studies in this area often explore how people become emotionally dependent on these digital friends, yet they frequently lack any connection to actual laws or government policies. It is as if researchers are studying the emotional impact of a new phenomenon while the legal system remains entirely unaware of it.
The second group is the largest and deals with the balance between keeping users engaged and keeping them safe. This cluster is filled with theoretical frameworks and proposals for how to govern these systems, but it suffers from a significant gap: it rarely translates these ideas into concrete policy. Researchers here are busy designing the rules for the future, but they are not yet working with the people who write the laws today. The third group focuses on healthcare, where AI assists doctors and patients. In this high-stakes environment, the conversation is very different. Here, the focus is on transparency and trust. Because a medical mistake can be life-threatening, there is a strong push for clear explanations of how the AI reaches its conclusions. This group is the most connected to existing policies, such as medical regulations, but it often overlooks the emotional and relational risks that are central to the other two groups.
A major finding of the review is that the most serious problems often fall through the cracks between these groups. The researchers identified a recurring issue they call an "accountability gap." This happens when something goes wrong, and it is unclear who is responsible: the person who designed the software, the company that sells it, or the user who relied on it. This confusion is especially acute when an AI acts like a human friend or a medical advisor. The review also highlighted a "regulatory lag," meaning that the rules for governing AI are moving much slower than the technology itself. By the time a law is written to address a specific risk, the technology has often already changed.
Perhaps most concerning is the disconnect between what researchers know and what is actually being done. The study found that while many papers discuss the dangers of emotional manipulation and user vulnerability, these concerns are rarely turned into enforceable regulations. In the healthcare sector, where rules are strict, the gap between ethical principles and real-world practice remains wide. In the social sector, where rules are loose, the gap is often invisible because the conversation never reaches the policymakers. The researchers suggest that the current approach is too reactive, treating symptoms after they appear rather than preventing them.
The review points toward a specific path forward: a design philosophy known as Value-Sensitive Design. This approach asks engineers to build ethical considerations directly into the technology from the very beginning, rather than trying to fix problems later. However, the authors note that while this idea is popular in theory, it is rarely implemented in practice. To truly protect people, the study argues that engineers, ethicists, and policymakers must stop working in isolation. They need to collaborate to create a framework that anticipates risks before they happen, ensuring that as AI becomes more integrated into our lives, it does so in a way that respects human dignity and safety. Without this coordinated effort, the rapid growth of artificial intelligence will continue to outpace our ability to govern it responsibly.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.