Creativity Reconsidered: Generative AI and the Problem of Intentional Agency

A new analysis argues that conscious intentional agency should not be treated as a necessary condition for creativity, pointing to the outputs of generative artificial intelligence as a key challenge to this long-held view [1]. In a paper posted to the preprint repository arXiv, researcher James Pearson contends that what the authors term the Intentional Agency Condition (IAC) must be abandoned [1]. The work highlights a growing predicament: generative AI systems are increasingly described as creative, yet they lack any capacity for conscious intent [2]. Two corpus analyses presented in the paper illustrate a rapidly increasing tendency for people to attribute creativity to these systems [2]. The paper critically evaluates a range of conflicting solutions proposed by other theorists and finds none of them satisfyingly resolves the tension [2]. The authors instead propose that ascriptions of creativity depend on what they call creative ability, a framework that explains why perceived intention matters for such judgments without making it a strict requirement [2]. The debate unfolds on arXiv, an open-access repository that, since its founding in 1991, has grown to host over two million papers and now receives about 24,000 submissions per month [8]. The platform itself is not peer-reviewed but provides rapid dissemination, a model that has made it central to fields like physics and computer science [8]. The tension the paper describes mirrors a broader psychological phenomenon. When new information, such as a machine producing a painting or a poem, conflicts with an existing belief that creativity requires a conscious mind, it can create cognitive dissonance [3]. According to Leon Festinger's theory, individuals experiencing this discomfort are motivated to reduce it, often by reframing a belief or explaining away the inconsistency [3]. The paper's proposed shift from an intent-based to an ability-based definition of creativity can be seen as one such conceptual reframing at a theoretical level. The practical implications of this shift are already visible in how creative work is organized. Crowdsourcing platforms, for instance, break down large creative projects into micro-tasks distributed across a dispersed group of participants, a process that values the cumulative result over any single participant's unified intention [4]. The transformer architecture, introduced in the 2017 paper "Attention Is All You Need," provided the technical foundation for the large language models now at the center of this debate [10]. That paper has since been cited more than 250,000 times [10].