The Creative Spark: Can AI Be Truly Original?

by

Author: Denis Avetisyan

A new perspective challenges traditional notions of creativity, arguing that consistent output, not intentional agency, may be the key to unlocking AI’s artistic potential.

This review proposes replacing intentional agency as a prerequisite for creativity with a focus on the consistent generation of novel and valuable outputs by AI systems.

The longstanding assumption that creativity necessitates intentional agency is increasingly challenged by rapidly evolving technologies. In ‘Creativity in the Age of AI: Rethinking the Role of Intentional Agency’, we argue against this ‘Intentional Agency Condition’ as a general prerequisite for creativity, demonstrating its problematic status in light of generative AI. We propose shifting the focus from who creates to what is created, advocating for a consistency requirement centered on the reliable generation of novel and valuable outputs. As AI systems increasingly contribute to creative endeavors, can our understanding of creativity adapt to prioritize product-first assessments over questions of authorship and intent?

The Illusion of Creativity: A Long-Held Belief

For decades, the prevailing understanding of creativity has centered on the production of outputs that are demonstrably both novel and valuable. This definition isn’t merely about producing something different; the generated output must also possess a discernible worth, whether aesthetic, functional, or otherwise. Historically, this framework has served as a cornerstone for evaluating creative endeavors across diverse fields, from artistic expression and scientific innovation to problem-solving and technological advancement. The emphasis on both novelty and value distinguishes creative acts from random occurrences or purely accidental discoveries, suggesting a purposeful generation of meaningful content. While seemingly intuitive, this long-held view is now being re-examined in light of increasingly complex systems capable of generating outputs that challenge its core assumptions, prompting a deeper consideration of what truly constitutes a creative process.

The longstanding definition of creativity fundamentally rests on the premise of intentionality, formalized as the ‘Intentional Agency Condition’ (IAC). This condition posits that a genuinely creative act necessitates conscious intent – a deliberate purpose behind the generation of a novel and valuable output. Historically, this makes intuitive sense, aligning with human experiences where creative endeavors are typically driven by a desire to express, solve a problem, or achieve a specific aesthetic. However, the IAC isn’t merely a descriptive observation; it’s a defining characteristic. Without demonstrable intent, an output, however novel or useful, is often categorized as accidental or a product of chance, rather than a true manifestation of creativity. This emphasis on conscious agency has shaped research and philosophical discourse on creativity for decades, influencing how creative processes are identified, analyzed, and even valued.

The longstanding definition of creativity, rooted in the generation of novel and valuable outputs stemming from conscious intent, encounters increasing difficulty when applied to the realm of artificial intelligence. As AI systems evolve beyond simple automation and begin to produce outputs that demonstrably meet criteria for novelty and value – composing music, generating art, even formulating scientific hypotheses – the requirement for ‘intentional agency’ becomes problematic. These systems often operate through complex algorithms and statistical probabilities, lacking the subjective experience and deliberate volition traditionally associated with human creativity. Consequently, strictly adhering to the intentionality condition risks either dismissing genuine creative potential in AI or necessitating a re-evaluation of what constitutes ‘intent’ itself, potentially broadening the definition in ways that challenge conventional understanding.

The longstanding emphasis on intentionality within definitions of creativity presents a significant challenge to fully grasping the phenomenon, particularly as artificial intelligence advances. By requiring conscious intent as a prerequisite for creative output, this perspective inadvertently limits the scope of inquiry, potentially overlooking innovative processes occurring in non-human entities – from complex algorithms generating novel art to natural phenomena exhibiting emergent patterns. This narrow focus not only risks dismissing genuine creativity outside the human realm, but also impedes a more comprehensive understanding of the underlying mechanisms driving creative acts themselves; it suggests creativity is inherently tied to subjective experience, rather than an objective capacity for generating valuable novelty, regardless of origin or awareness.

Beyond Intent: A Pragmatic Re-Evaluation

The traditionally accepted definition of creativity often necessitates intentionality – a conscious agent deliberately producing novel and valuable outcomes. However, this presents limitations when evaluating creativity in systems lacking conscious intent. Our proposed ‘New Standard Definition’ circumvents this requirement by focusing solely on the demonstrable production of novelty and value. This reframes creativity as an observable characteristic of a system’s output, independent of any internal states or motivations of the source. Consequently, a process or system can be deemed creative if it consistently generates outputs that are both original and considered beneficial or useful within a defined context, regardless of whether that output was purposefully designed or emerged as an unintended consequence.

The proposed definition of creativity centers on demonstrable output, specifically the consistent generation of novelty and value. This assessment is strictly external; the internal state, motivations, or conscious intent of the originating system are considered irrelevant to the determination of creative output. Evaluation focuses solely on whether the produced result is both new – differing from previously existing artifacts – and valuable, assessed through established criteria appropriate to the domain. This means that a system exhibiting creativity need not ‘intend’ to be creative, or even possess subjective experience; the objective characteristics of its output are the sole basis for classification.

Conceptual Engineering, as applied to the definition of creativity, involves a purposeful re-evaluation and modification of an established concept – in this case, creativity – to enhance its practical utility and alignment with observed phenomena. This process isn’t about discovering a ‘true’ definition, but rather constructing a definition that better functions within a specific framework, allowing for broader applicability and avoiding limitations imposed by historically contingent features. Specifically, by intentionally decoupling intentionality from the core criteria of creativity, the revised definition becomes more effective in identifying creative output across diverse systems, including those, like Generative AI, lacking conscious agency. This pragmatic approach prioritizes the observable characteristics of novelty and value production over internal states or motivations, thereby improving the concept’s analytical power.

The proposed revision to the definition of creativity explicitly allows for novelty and value to emerge from systems devoid of conscious intent. This is particularly relevant given the capabilities of Generative AI models, which demonstrably produce outputs – text, images, code – that can be considered novel and valuable according to established metrics, despite lacking subjective experience or intentionality. This does not require attributing consciousness to these systems, but rather recognizes that the source of the novel and valuable output is irrelevant to its classification as creative under the new standard. The focus shifts from the internal state of the creator to the external characteristics of the creation itself.

Measuring the Machine: Consistent Output as Evidence

The implementation of the ‘Consistency Requirement’ represents a shift from the previously utilized IAC (Innovation Achievement Criterion) for evaluating creative output. Unlike the IAC, which focused on singular instances of innovation, the Consistency Requirement prioritizes a demonstrated pattern of consistently producing both novel and valuable results. This necessitates evaluating creative sources not on the basis of isolated achievements, but on their ability to reliably generate outputs meeting defined criteria for originality and worth over a sustained period. The change reflects a need to assess the ongoing creative capacity of a source, rather than simply identifying occasional instances of impactful work.

Creative consistency, as a measurable attribute, extends beyond simply demonstrating a statistically significant frequency of novel outputs. It requires the identification of predictable patterns in the generation of those outputs, indicating a reliable process capable of consistently producing meaningful results. This means evaluating not just if something new is created, but how it is created – are there recurring stylistic elements, thematic preferences, or structural characteristics that define the creative source’s output? Establishing these patterns differentiates genuine consistency from random novelty, providing a basis for assessing the dependability of a creative entity or system over time.

Analysis of language use patterns within extensive digital corpora, such as the ‘News on the Web’ dataset, offers a quantifiable approach to assessing creative consistency. By examining the frequency and co-occurrence of terms related to creative acts and their agents – including the recent surge in phrases like ‘AI creates’ – researchers can identify shifts in the attribution of creativity. This method moves beyond simple output counts to reveal how language reflects evolving perceptions of who or what is generating novel content, providing empirical data relevant to understanding the reliability and patterned nature of creative production as attributed in publicly available news sources.

Ngram analysis of a large corpus, such as ‘News on the Web’, demonstrates a statistically significant increase in the frequency of the phrase “AI creates” beginning around 2017. This upward trend in linguistic co-occurrence indicates a growing association, within published media, between artificial intelligence and the act of creation. The observed increase in “AI creates” frequencies aligns with and corroborates the absolute increases in reported instances of AI-generated content documented within the ‘News on the Web’ dataset, providing quantitative support for the perception of increased AI involvement in creative endeavors.

The Implications of Letting Go: Re-thinking Creativity Itself

Historically, creativity has been tightly linked to intentionality – the idea that a conscious agent must intend to create something novel and valuable. However, recent conceptual work challenges this assumption, arguing that valuable novelty can arise without deliberate intent. This broadened perspective dramatically expands the scope of what qualifies as creative, encompassing phenomena previously excluded by the intentionality requirement. Consequently, outputs from artificial intelligence – algorithms generating art, composing music, or designing solutions – can now be legitimately considered creative, even though these systems lack conscious awareness or purposeful design. Furthermore, this redefinition opens the door to recognizing creative processes in other non-human entities, from naturally occurring patterns in physics to emergent behaviors in complex systems, fundamentally shifting the understanding of creativity’s origins and prevalence.

Historically, creativity has been tightly linked to the notion of expressive authenticity – the idea that a truly creative act stems from an individual’s genuine emotions and personal experiences. However, a revised definition of creativity, one that removes the requirement of intentionality, fundamentally alters this relationship. This decoupling suggests that creative outcomes are not necessarily dependent on who creates them, or why, but rather on the novelty and functional value of the creation itself. A computational algorithm generating a unique and useful design, for instance, can now be considered creative even without possessing subjective experience or authentic emotional expression. This shift doesn’t diminish the importance of human creativity, but broadens the conceptual landscape, allowing for the recognition of creative processes in systems previously excluded due to a presumed lack of inner life or genuine intent. The focus moves from the source of the idea to the idea’s inherent qualities and impact, reshaping how creativity is understood and evaluated.

A core tenet of this conceptual engineering of creativity lies in its emphasis on ‘Functional Value’ – the demonstrable usefulness of a novel output, irrespective of its origin. This shifts the focus from the often-intangible qualities of intention or emotional expression to the concrete benefits a creation provides. Consequently, assessing creativity becomes less about why something was made and more about what it accomplishes; a solution generated by an algorithm, a structurally innovative bird’s nest, or a uniquely effective tool all qualify, provided they demonstrably fulfill a purpose. This pragmatic approach not only broadens the scope of creative assessment but also directly facilitates the application of creative principles to problem-solving across diverse fields, from engineering and design to artificial intelligence and beyond, fostering innovation based on tangible outcomes rather than subjective interpretations.

Decoupling creativity from conscious intent fundamentally alters how researchers approach innovation across various domains. This shift allows for the recognition of creative outputs originating from systems lacking subjective experience, such as artificial intelligence, evolutionary algorithms, and even certain natural phenomena. Consequently, investigations can move beyond attempts to model the ‘creative process’ within a human mind and instead focus on identifying the computational mechanisms – the generation of novelty that possesses functional value – regardless of origin. This broadened perspective unlocks opportunities to harness creative potential in unexpected places, potentially leading to breakthroughs in fields ranging from automated design and problem-solving to the development of novel materials and artistic expressions, all without requiring an understanding of why a system produced a given result, only that it did.

The pursuit of ‘intentional agency’ as the cornerstone of creativity feels… quaint. This paper rightfully questions whether it’s even necessary anymore, proposing ‘consistency’ as a more practical metric. It’s a pragmatic shift – because, predictably, production will demonstrate the limitations of any elegant theory hinging on something as nebulous as ‘intent’. As Andrey Kolmogorov observed, “The most important problems are those which we can solve.” This focus on demonstrable output-reliable novelty and value-isn’t about diminishing creativity; it’s acknowledging that, ultimately, if a system consistently delivers, the philosophical niceties become secondary. Everything new is old again, just renamed and still broken, and sometimes, consistent functionality trumps conceptual purity.

What’s Next?

The substitution of ‘consistency’ for ‘intentional agency’ feels less like a resolution and more like a managed migration of technical debt. The paper correctly identifies the inadequacy of anthropocentric metrics when evaluating generative systems, but consistency, while measurable, merely shifts the goalposts. It offers a target for optimization, not an explanation of the phenomena. The bug tracker will, predictably, fill with edge cases demonstrating precisely how consistency fails-how reliably novel outputs can still be profoundly useless, or worse.

The focus on a ‘product-first’ approach risks conflating output with meaning. Value, even in a computationally defined space, remains stubbornly subjective. One anticipates a surge in metrics attempting to quantify ‘expressive authenticity’-a phrase that already feels primed for ironic detachment. The pursuit will be relentless, and the resulting dashboards will offer the illusion of control where none exists.

The real challenge lies not in defining creativity, but in accepting its essential messiness. The field will continue to refine the algorithms, and the systems will continue to surprise-not because they’ve achieved some analog of consciousness, but because complex systems always find new ways to break the models. It isn’t deployment; it’s letting go.

Original article: https://arxiv.org/pdf/2601.15797.pdf

Contact the author: https://www.linkedin.com/in/avetisyan/

See also:

2026-01-23 12:45