Creativity has long been considered the last refuge of human uniqueness. AI can compute, pattern-match, and optimize. But can it truly create? The answer from cognitive science is more unsettling than either side usually admits.
When DALL-E generated its first striking images in 2021, a predictable debate erupted. Artists said it wasn't creative — it was just recombining training data. Technologists said the same argument applied to human artists, who are also recombining influences. Both sides were partly right. Which is to say: both sides were missing the more interesting question.
The more interesting question is: what do we actually mean by creativity? And when we look closely at what cognitive science says, the AI case becomes far less clear-cut than the headlines suggest.
Margaret Boden's Three Types
The philosopher Margaret Boden, who has studied creativity for decades, distinguishes three types: combinational (making unfamiliar combinations of familiar ideas), exploratory (exploring the limits of an existing conceptual space), and transformational (altering the conceptual space itself). Most human creativity, she argues, is combinational or exploratory. True transformational creativity — the kind that invents a new genre, a new mathematical framework, a new scientific paradigm — is rare.
By Boden's taxonomy, current AI systems are clearly capable of combinational and exploratory creativity. They produce novel, surprising, aesthetically interesting outputs by combining and interpolating across vast conceptual spaces. Key point The question is whether this is “real” creativity — and that question turns out to depend almost entirely on what you think creativity is for.
Your product name here
One or two sentences on what this product does and why it fits a reader of this journal. Keep the tone of the surrounding writing — it reads better and converts better.
Learn more →The Intentionality Problem
The most philosophically serious objection to AI creativity is not about novelty or surprise. It is about intention. Human creative acts are intentional — the artist is trying to say something, express something, communicate something about their experience of being alive. The work means something because the maker meant it.
Current AI systems have no interiority — no experience, no intentions in the philosophically robust sense. When a model writes a poem about loss, it is not drawing on experience of loss. It is statistically interpolating across millions of human expressions of loss. The output may be beautiful. The process is fundamentally different.
The question isn't whether the painting is beautiful. The question is whether anyone suffered to make it — and whether that suffering is part of what makes it mean something.
— Adapted from a 2025 panel on AI and aesthetics, MIT Media LabWhy the Answer Matters
This isn't purely philosophical. It has concrete implications for copyright law, for creative industries, and for how we value human artistic work in a world where machines can produce competent output at scale.
If AI creativity is fundamentally different from human creativity — even when the outputs are indistinguishable — then there may be grounds for treating human creative work as categorically distinct in legal and economic terms. If AI creativity is continuous with human creativity — just a different implementation of the same underlying cognitive operations — then the arguments for special protection look weaker.
The cognitive science answer, honestly stated: we don't know enough about what human creativity is to know whether AI has it. Our folk theory of creativity doesn't survive close scrutiny even when applied to humans. We too are pattern-matchers, combiners, explorers of learned conceptual spaces. The difference may be one of degree. Or it may be categorical. We are, genuinely, still figuring that out.