What AI Is Doing to Our Definition of Originality

The arrival of generative AI systems capable of producing plausible prose, music, and visual art has created a specific intellectual problem for psychology and aesthetics: the need to define originality with more precision than was previously required.

For most of the 20th century, discussions of creativity and originality could remain somewhat vague. Researchers knew human creativity was valuable and somewhat understood its mechanisms, but the contrast class, what creativity was being distinguished from, was limited to mechanical reproduction or formulaic output. Generative AI has changed that. When a system trained on billions of examples can produce outputs that pass as competent, the question of what makes human creative work distinctive becomes genuinely urgent.

Margaret Boden's Framework and Why It Matters Now

The most useful conceptual tool for this question comes from the work of Margaret Boden, Research Professor of Cognitive Science at the University of Sussex, whose taxonomy of creativity types was developed well before generative AI reached its current capabilities. Boden distinguishes three forms of creativity: exploratory creativity, which involves searching and extending within an existing conceptual space; combinational creativity, which involves unexpected connections between familiar ideas; and transformational creativity, which involves changing the underlying rules of the conceptual space itself.

Most generative AI systems, Boden has argued in interviews and written responses to the current AI moment, operate primarily in the combinational register. They identify statistical relationships between features of training data and generate outputs that recombine those features in plausible arrangements. The system has no access to the conceptual space it appears to be operating in because it has no understanding of what the concepts mean. It works with patterns, not with the referents of those patterns.

Transformational creativity is where the distinction sharpens. When Paul Cezanne began treating the painted surface as a series of geometric planes rather than a window onto a scene, he was changing the foundational assumptions of Western painting. When James Joyce dissolved the boundary between reported thought and narrative, he was restructuring what a novel could do. These moves were not optimizations within an existing system; they were modifications of the system's rules. Boden's framework suggests this capacity, the ability to identify and then violate the governing constraints of a domain, is what AI systems currently lack and may structurally be unable to develop without something resembling genuine understanding.

Aaron Hertzmann on the Question of Machine Art

Aaron Hertzmann, a computer scientist and artist at Adobe Research, addressed this problem directly in his influential 2018 paper "Can Computers Create Art?" published in the journal Arts. Hertzmann's argument is careful and worth examining in detail. He concludes that, in the present state of the technology, computers do not create art because art requires an author, and authorship requires intentionality and the capacity for genuine communication between a maker and an audience.

This is a position rooted in the philosophy of art rather than in technical limitations. Hertzmann is not claiming that AI systems will always produce worse outputs than humans. He is claiming that the category "art" depends on a relationship: the work carries meaning because a conscious agent made choices in producing it, and an audience receives those choices as communications from another mind. When a system generates an image without any stake in what the image says, the image may look like art, but the relationship that constitutes art is absent.

This framework has direct implications for how originality should be defined. If originality is purely about statistical novelty, AI systems can produce it at scale. If originality is about a specific kind of purposeful departure from convention, grounded in a maker's understanding of what the convention means and why departing from it matters, then the current generation of AI systems is excluded by definition.

Yann LeCun and the Limits of Generative Models

Yann LeCun, Chief AI Scientist at Meta and winner of the Turing Award, has been among the most vocal critics of the assumption that large language models and generative systems are on a path to human-like creativity. In public lectures and technical papers, LeCun has argued that current generative models lack the world-model that would allow them to understand causality, plan, and reason about consequences.

This matters for the question of originality because genuine creative departures require understanding what a convention is, why it exists, and what would be gained by violating it. A system that predicts likely next tokens cannot evaluate whether its output is surprising in any meaningful sense; it has no reference frame for what would constitute surprise. LeCun's position is that the current generation of AI, however impressive in surface-level output, is missing the architectural foundations that would make genuine creative understanding possible.

The Photography Parallel

Historical parallels are instructive here, though they require care. When photography became commercially viable in the mid-19th century, critics and painters initially treated it as a threat to visual art. Joshua Reynolds and the academic painters who followed him had established a tradition in which the skilled rendering of scenes and portraits was central to painting's value. Photography could render a scene with far more precision than even the most skilled painter, and it could do so cheaply and quickly.

The response of serious painters over the following decades was to clarify what painting could do that photography could not. Impressionism was, in part, a systematic exploration of the fact that painting could depict perception and sensation rather than objects. Post-Impressionism went further, using color and form as structural elements rather than descriptive ones. Abstract painting followed, eliminating representational claims entirely. These developments were produced by the pressure of a competing medium, and they were more genuinely original than what preceded them.

The neurological dimension of this process matters. Research by Rex Jung, a clinical neuropsychologist at the University of New Mexico, and colleagues on the neuroscience of creative insight suggests that genuinely original thinking involves the capacity to hold competing conceptual frames simultaneously and find unexpected relationships between them. This process involves specific patterns of activation in the prefrontal cortex and default mode network. It is grounded in the maker's accumulated experience, emotional history, and understanding of the domain. AI systems have none of these. They have training data, which is not the same thing.

What Originality Actually Means Neurologically

Research in cognitive neuroscience has clarified the brain mechanisms underlying creative insight in ways that are relevant to the AI comparison. Studies by Roger Beaty, a cognitive neuroscientist at Penn State University, using functional MRI have identified that highly creative individuals show stronger connectivity between the default mode network, which is associated with imagination and spontaneous thought, and the executive control network, which is associated with focused attention and evaluation.

This pattern suggests that original thinking is not simply divergent or associative thinking. It is divergent thinking that is simultaneously subjected to evaluative constraints. The creative individual generates unusual possibilities and also has a developed sense of which unusual possibilities are worth pursuing. This evaluative capacity depends on deep knowledge of a domain, accumulated aesthetic judgment, and an embodied relationship to the materials of the craft.

AI systems generate possibilities but have no mechanism for evaluation that is analogous to developed aesthetic judgment. They can be trained with feedback signals that approximate evaluation, but those signals are themselves derived from human judgments. The system is, in this sense, always downstream of human aesthetic standards rather than capable of generating them.

What the Observer's Perspective Adds

Nikita Datar's book The Observer addresses the psychological dimension of this question from a different angle. The capacity to observe one's own creative process, to notice what is actually happening when an idea forms rather than what one assumes is happening, is central to understanding originality from the inside. For creative practitioners, AI's arrival creates a practical pressure to develop this observational capacity more deliberately. If the value of human creative work lies in the specific choices made by a specific person with a specific history, then knowing what one's choices actually are, and why one is making them, becomes more important than it was when "good enough" was sufficient.

What This Means

For psychology and for people who make things, the arrival of AI generation does not diminish human creativity. It clarifies it. Several practical implications follow from the research reviewed here.

Understanding domain rules is necessary before departing from them. Boden's taxonomy makes clear that transformational creativity, the rarest and most valuable kind, requires genuine understanding of what the existing rules are. Developing that understanding through sustained engagement with a field is a prerequisite that AI bypasses by working with surface patterns rather than conceptual structures.

Intentionality is not a luxury. Hertzmann's framework establishes that the relationship between maker, work, and audience depends on the maker's genuine stake in the work. Developing clarity about why you are making a particular thing, what you are trying to communicate and to whom, is what separates human creative work from generative output.

AI pressure may be clarifying rather than threatening. If, like painting after photography, the arrival of AI forces creative practitioners to become more deliberate about what their work does that machine output cannot do, the result may be a more rigorous and more genuinely original creative culture rather than a less original one.

The process is not incidental to the product. Csikszentmihalyi's research on flow and creative experience establishes that the subjective experience of making something is part of the value of creativity for the person doing it. AI can produce outputs but it cannot have experiences. For human creators, the experience of making is inseparable from the value of having made.