Shuichiro Ogawa
日本語

Notes · updated 2026-07-19

Research Currents in AI and the Study of Art and Culture: Aesthetics, Media Studies, Computational Creativity (2024–2026)

When generative AI began to draw pictures, compose music, and write stories, what wavered first was not the technology but the concepts. Who is the author? What becomes of the aura that the age of mechanical reproduction had supposedly stripped away, once we enter the age of generation? The peer-reviewed literature of 2024 through 2026 takes up these questions in the vocabulary of each field.

This note organizes that scholarly response through sixteen works. It does not enter the industrial ecosystem or market dynamics (that belongs to generative-art-ecosystem-industry). What it follows is how aesthetics, computational creativity, media and cultural studies, musicology, and critical cultural theory began to conceptualize generative AI. For the intersection of AI and digital humanities, see ai-humanities-digital-humanities-trends.

Authorship and aura were reframed as a problem of generation, not reproduction

The aura Benjamin theorized in the age of technical reproduction rested on a specific structure: reproduction erodes the singularity of the original. Generative AI shifts that structure, because a generated work has no prior original to reproduce.

In 2024, Park argued that generative AI redefines the traditional aura while producing a dynamic between technological liberation and technological dependence (AI & Society). In 2025, Salas Espasa and Camacho gathered this discussion into a systematic literature review, arguing that AI-generated works blur the very boundary between original and copy. They place a middle concept there: the “semi-aura.” It situates AI-generated works at a position that is neither the fully singular aura of an original nor a purely mechanical copy.

From the side of authorship, analytic aesthetics posed a normative question. In 2026, Lopes diagnosed existing generative AI (DALL·E and the like) as lacking the genuine intentionality and executive agency required to be an “artist,” making it a “tool” rather than an artist (Journal of Aesthetics and Art Criticism). Yet he does not stop there. He argues that if AI is designed according to the principle of “aesthetic alignment,” it could respect aesthetic values across cultural practices that humans will not personally appreciate, and thereby expand aesthetic diversity. Using the robotic AI artist Ai-Da as a proof of concept, he concludes that existential anxiety about AI artists is not justified.

Some work connects to history instead. In 2025, Peters overlaid the experimental spirit of early-twentieth-century avant-gardes (Constructivism, Dada) onto contemporary AI art (AI & Society). The figure of the artist that emerges is not the solitary creator. It is a shift toward the “engineer/curator” who critically recomposes technical and cultural systems.

Computational creativity shifted its weight toward testing received wisdom

While aesthetics reworked its concepts, research in computational creativity reworked its methods of evaluation.

In 2024, Franceschelli and Musolesi assembled a survey of computational-creativity theory, key machine-learning methods, and automatic evaluation methods in ACM Computing Surveys. It positions generative deep learning within the lineage of creativity research as a reference axis for the field. In the same year, Ismayilzada and colleagues redefined creativity as “the ability to produce novel, useful, and surprising ideas” and surveyed AI’s creative capabilities across four domains: problem-solving, language, art, and science (arXiv). State-of-the-art models excel at linguistic and artistic output, yet retain serious limits in abstract reasoning, originality, and long-range coherence, accompanied by hallucination. They advocate evaluation that views the process in multiple dimensions.

The emblematic move in this field put a piece of received wisdom itself to experiment. The temperature of an LLM has widely been called its “creativity parameter.” In 2024, Peeperkorn and colleagues tested this in narrative generation and showed that temperature correlates only weakly with novelty and is unrelated to cohesion or typicality (ICCC’24). The “creativity” produced by higher temperature was far weaker and more nuanced than the received view assumes. A shared footing is forming here: whether creativity is measured as a product or as a process shapes the conclusion.

The politics of representation and labour moved to the centre of cultural studies

Media and cultural studies treat generative AI not as an aesthetic object but as an apparatus of cultural production.

In 2024, Gillespie empirically tested three generative-AI tools and showed that they enact a “politics of visibility,” reproducing normative identities and narratives while suppressing the representation of minorities (Big Data & Society). It is a theoretical extension of platform content-moderation research into the moment of generation. In the same year, Laba strikes at this problem of representation in a more concrete scene. Analyzing images generated with Midjourney using the war in Ukraine as a case, she dismantles the Silicon Valley narrative of the “engine for the imagination” (Media, Culture & Society). The generated images of war converged toward something homogeneous, stripped of context and diversity. This paper is cited as a turning point in the field (Scopus citation count [requires primary verification]).

The response from the side of labour is substantial. In 2024, Lee argued that generative AI breaks the human monopoly on creativity and produces “creative precarity,” critically recomposing the existing three-part typology of labour in cultural policy (Media, Culture & Society). In the same year, Bender analyzed the discourse of the 2023 Writers Guild and actors’ union strikes and put the very premise of “the primacy of human creativity” under examination. Erickson offers a somewhat different picture. From six case studies of AI products, he showed that AI products are in fact more labour-intensive than conventional media products, combining traditional production skills with new computational expertise (Creative Industries Journal). Two diagnoses, the “disappearance” and the “continuity” of labour, stand side by side in the same year, 2024.

Academic disciplines themselves began to make AI an object of study

AI became at once a tool of cultural research and a mirror in which each discipline reflected its own premises. Musicology showed that turn most sharply.

In 2025, White argued, in the mainstream journal of the field, the Journal of the American Musicological Society, that music has an AI problem and AI has a music problem, treated bidirectionally. The structure places the problems AI poses for musicology and the problems musicology poses for AI in symmetry. That a central learned journal took AI up as a formal topic is itself one index of objectification. He extends this argument into the book The AI Music Problem (Routledge, 2025).

Critical theory placed AI art within political economy and coloniality

Critical cultural theory reads the skew of representation not as a surface malfunction but as a symptom of structure.

In 2025, Jääskeläinen and colleagues analyzed 180 images generated with Stable Diffusion through an intersectional lens (AI & Society). Images of law enforcement skewed statistically toward white men, poverty toward the Global South, and feminist imagery toward conventional images of women. They see there a need for “algorithmic reparation.” It is a design orientation that does not merely flatten the skew but actively redresses historical injustice.

The decolonial view is more structural. In 2025, Correa Lucero and Martens mapped the literature through a Latin American decolonial framework and showed that AI development reproduces the historical structure of the “coloniality of power, knowledge, and being” (AI & Society). They name its contemporary reconfigurations: data colonialism, the coloniality of labour, and digital feudalism.

An empirical study from computational social science joins the same critical lineage. In 2025, Roland, So, and Long, drawing on Bourdieu’s theory of the “cultural field,” simulated 101 AI “authors” modeled on real-world counterparts (AI & Society). Language models reduced literary distinction to a simple binary across race, gender, and publication context, and flattened within-group diversity. The skew of representation appears as a structural tendency reproduced with each act of generation.

Cross-cutting themes (candidate axes for review)

  1. Redefining authorship: A normative position that separates AI off as a “tool” (Lopes 2026) runs alongside a relational position that locates authorship in the interaction of human intention, algorithmic generation, and social evaluation. The aura debate (Park 2024; Salas Espasa & Camacho 2025) supplies the vocabulary of aesthetic history to this redefinition.
  2. The methodology of evaluation: In computational creativity, conclusions change depending on whether creativity is measured as a process or a product (Peeperkorn et al. 2024; Ismayilzada et al. 2024). Falsifying the received view (temperature = creativity) advanced the field’s self-examination.
  3. Disappearance vs. continuity: On creative labour, the diagnosis that AI erases human labour (Bender 2024; Lee 2024) stands at the same time as the diagnosis that it preserves labour-intensity (Erickson 2024).
  4. Is skewed representation a symptom of structure?: Whether the skew is read as a technical defect or as a reproduction of coloniality and gender order changes the prescription, from “debiasing” to “algorithmic reparation” and “decolonial evaluation” (Gillespie 2024; Jääskeläinen et al. 2025; Correa Lucero & Martens 2025; Roland et al. 2025).
  • Redefining concepts: Lopes 2026 (the aesthetic-alignment principle); Salas Espasa & Camacho 2025 (the semi-aura SLR).
  • Falsifying method: Peeperkorn et al. 2024 (the temperature experiment).
  • Turning point in representation: Laba 2024 (Midjourney and war imagery); Roland et al. 2025 (simulating the cultural field).
  • Disciplinary objectification: White 2025 (musicology’s AI problem).

References

Sixteen works in total. DOIs/URLs verified as of the retrieval date 2026-07-19. Items marked [requires primary verification] are those whose primary page was not reached (publisher authentication wall), but whose DOI and bibliographic details were cross-checked against multiple independent sources (residual bibliographic confirmation that does not affect the main facts). The internal working ledger is source/review/ai-arts-culture-studies-trends/papers.md.

Aesthetics and philosophy of art

  • Park, S. (2024). The work of art in the age of generative AI: aura, liberation, and democratization. AI & Society 40, 1807–1816. https://doi.org/10.1007/s00146-024-01948-6 [requires primary verification]
  • Salas Espasa, D., & Camacho, M. (2025). From aura to semi-aura: reframing authenticity in AI-generated art—a systematic literature review. AI & Society 40, 6727–6759. https://doi.org/10.1007/s00146-025-02361-3 [requires primary verification]
  • Lopes, D. M. (2026). AI Art and Artists: What They Are, What They Could Be, What They Should Be. Journal of Aesthetics and Art Criticism (advance article). https://doi.org/10.1093/jaac/kpag001
  • Peters, J. (2025). Generative AI and the avant-garde: bridging historical innovation with contemporary art. AI & Society 40, 6407–6424. https://doi.org/10.1007/s00146-025-02410-x [requires primary verification]

Computational creativity

  • Franceschelli, G., & Musolesi, M. (2024). Creativity and Machine Learning: A Survey. ACM Computing Surveys 56(11), Article 283. https://doi.org/10.1145/3664595
  • Ismayilzada, M., Paul, D., Bosselut, A., & van der Plas, L. (2024). Creativity in AI: Progresses and Challenges. arXiv:2410.17218 [cs.AI]. https://arxiv.org/abs/2410.17218
  • Peeperkorn, M., Kouwenhoven, T., Brown, D., & Jordanous, A. (2024). Is Temperature the Creativity Parameter of Large Language Models? Proc. 15th International Conference on Computational Creativity (ICCC’24). https://arxiv.org/abs/2405.00492

Media and cultural studies

AI in musicology

Critical cultural theory

  • Jääskeläinen, P., Sharma, N. K., Pallett, H., & Åsberg, C. (2025). Intersectional analysis of visual generative AI: the case of stable diffusion. AI & Society 40(6), 4341–4362. https://doi.org/10.1007/s00146-025-02207-y
  • Correa Lucero, H., & Martens, C. (2025). Colonial structures in AI: a Latin American decolonial literature review of structural implications for marginalised communities in the Global South. AI & Society 41(3), 2511–2527. https://doi.org/10.1007/s00146-025-02547-9 [requires primary verification]
  • Roland, E., So, R. J., & Long, H. (2025). The social AI author: modeling creativity and distinction in simulated cultural fields. AI & Society (online first). https://doi.org/10.1007/s00146-025-02790-0 [requires primary verification]

Unverified items

The following were not reached at the primary page (publisher authentication wall), but their DOI and bibliographic details were cross-checked against multiple independent sources; they will be resolved in a future revision once the primary page is reached. Park 2024, Salas Espasa & Camacho 2025, Peters 2025, Gillespie 2024, Bender 2024, White 2025, Correa Lucero & Martens 2025, Roland et al. 2025 (eight items). In addition, the Scopus citation count for Laba 2024 (the basis for the turning-point claim) requires primary verification.

← All Notes · Home