Researchers from Singapore Management University and Duke-NUS Medical School have proposed a framework for understanding when AI companions help or harm. Rather than asking whether the technology is broadly beneficial or harmful, they argue the more useful question is who benefits, who is most at risk, and how society should respond.
The framework suggests that isolated users are a key risk group. People with weak real-world social networks may rely on AI companions in ways that gradually erode the social skills needed for human interaction. The authors frame this as a shift in focus: the technology itself is not uniformly good or bad, but its effects depend heavily on the user's circumstances.
Because this is a single source, there is no independent or conflicting view to compare. The paper is presented as a conceptual framework rather than an empirical study, so its claims are intended to guide future research and policy discussions rather than to settle the debate.