The Metaphors We Use, the Possibilities We Lose: AI and the Redistribution of Agency

In this conceptual article, we argue that the metaphors humans use for generative AI are not neutral descriptions; rather, they redistribute creative agency, epistemic authority, and the sense of what is possible between humans and machines, shaping what people can imagine doing with these systems. This redistribution is often absorbed by people rather than chosen: a metaphor becomes part of how someone uses AI without being examined or recognized as a framing. We call this absorbed redistribution and distinguish it from deliberate or resisted redistribution. Drawing on possibility studies, we develop a framework for analyzing how metaphors picture AI, what possibilities they expand and contract, and the conditions under which these effects intensify. Applying it to a set of common AI metaphors, with closer analysis of several that are common in learning, we show that every framing tilts the human-AI relation somewhere, often in different directions. Education offers a clear case, because the metaphors people adopt for AI interact with the metaphors they already hold for learning. We argue that metaphor awareness, recognizing the framings one operates inside and choosing among them, is itself a form of creative and epistemic agency that supports possibility.