Steinert's 'Art: Brought to You by Creative Machines' (Philosophy & Technology, 2017) argued that creative machines (CMs) such as automated painters, improvising robotic musicians, and neural style-transfer or generative-adversarial-network systems can produce genuine works of art without being artistic agents: they are technical artifacts whose proper function is to bring about art, so the relevant intentions are those of the human makers and designers, transferred to the machine through its function. Steinert explicitly rejects what he calls the strong view — that technical artifacts can literally be agents — as resting on dubious ontological assumptions, and defends a moderate view on which machines are art-producing instruments, not artists. What the account deliberately leaves open is whether a defensible, non-ontologically-dubious account of machine artistic intentionality or of artistic agency distributed over humans and machines can nevertheless be developed, and what the architectural, training, and design limits of CMs imply for that attribution and for the artistic values their products can realize.

Background

'Creative machines' (CMs) are software or robotic systems designed to produce artifacts in established art forms: The Painting Fool (an automated painter), Shimon (a robotic marimba player that improvises with human musicians), neural style transfer (a technique that re-renders one image in the visual style of another), generative adversarial networks (GANs — pairs of neural networks trained adversarially so that a generator produces images a discriminator cannot distinguish from its training data), and, since 2022, large text-to-image diffusion models (generative networks that synthesize images from written prompts). Whether such systems can create art is a live question for the philosophy of art because many definitions of art tie an object's art status to the intentions of its maker, and a machine, on the usual view, has no intentions.

Steinert (2017) argued that machines can create works of art while granting them no agency. His argument turns on the notion of a proper function: what a technical artifact — a deliberately designed and produced object — is supposed to do in virtue of its design history, as a thermometer's proper function is to measure temperature. On his account, CMs are technical artifacts whose proper function is to bring about works of art; the intentions that confer art status are those of the human designers and makers, embedded in the artifact's function, so their products are artworks without any attribution of artistic intentionality to the machine itself. Steinert calls this a moderate view, adapting a distinction Peterson and Spahn (2011) introduced for the moral case: the strong view holds that technical artifacts can literally be agents (moral agents there, artistic agents here); the moderate view denies this while acknowledging that artifacts are causally and normatively significant. Steinert embraces the moderate view and dismisses the strong view as ontologically dubious — resting, critics would say, on anthropomorphic projection of the kind Romic (2022) documents in the public presentation of the humanoid robot 'artist' Ai-Da.

Independently, empirical aesthetics — the experimental study of how people experience and evaluate art — began measuring how machine provenance affects appreciation. The relevant framing is cognitive penetrability: the contested thesis that what one knows or believes can alter perceptual experience itself, not merely subsequent judgment — Nanay (2015) argued this via the 'gallery of indiscernibles', where an original and a perfect forgery look alike yet are said to be seen differently once their provenance is known. Chamberlain, Mullin, Scheerlinck and Wagemans (2018) found observers derogate computer-generated art, though the bias reversed when they watched robotic artists at work. Later and larger studies — Ragot, Martin and Cojean (2020, N=565); Gangadharbatla (2022); Bellaiche et al. (2023); Hitsuwari, Ueda, Yun and Nomura (2023); Seli et al. (2026) — converged on a robust 'AI-label' effect: physically identical images or poems are rated lower on liking, beauty, profundity, and worth when attributed to a machine, with the effect moderated by embodiment, framing, and perceived effort. Related work in experimental philosophy (using survey methods to study how people actually apply philosophical concepts) probes the agency concept directly: Mikalonytė and Kneer (2022, N=693) found people judge robot-made paintings to be art roughly as often as human-made ones, yet are much less willing to call the robots artists, mainly because they withhold artistic intention from them.

Against this backdrop, the philosophical debate has intensified rather than settled. Coeckelbergh (2017), in a companion piece to Steinert's paper, argues the question cannot be settled by object- or process-properties alone but depends on relational and social factors. McCormack, Gifford and Hutchings (2019) analyze autonomy, authenticity, authorship, and intention specifically for computer-generated art. Recent work divides sharply: Wojtkiewicz (2023) holds that generative AI programs are a new technique of automatic image-making that affords creative agency to their human users (a moderate-view verdict for the new model class); Chiodo (2024) argues that AI artifacts cannot be defined as art at all; Anscomb (2025) examines the three conditions the authorship literature uses — collective authorship, co-creatorship, and co-production — and concludes generative AI does not yet meet any of them, proposing the weaker framing 'AI-assisted production'; Berryman (2024) questions the originality claims of GAN systems that purport to invent new styles and stresses how training datasets perpetuate art-historical bias; Woodward (2025) explores 'hybrid artistic agency' between humans and creative AI. The upshot is that Steinert's deliberately avoided strong view remains an open, actively contested position rather than a settled one.

Problem Statement

Can a defensible, ontologically non-dubious account of machine artistic agency be developed — a strong view on which creative machines, or human–machine systems treated as distributed agents, count as genuine artistic agents (either by attributing artistic intentionality to the machine or by an account of artistic agency or authorship distributed over the human–machine system, which need not locate intentionality in the machine), rather than mere art-producing technical artifacts as on Steinert's moderate view? The question therefore asks, whether the verdict is positive or negative:

  1. Specify explicit criteria for artistic agency or distributed artistic authorship (adopting, refining, or rejecting the existing criteria from the literature on collective authorship, co-creatorship, and co-production) and argue that some actual or realistically constructible class of creative machines — from constrained systems with fixed generative repertoires and training-corpus-biased generative adversarial networks to contemporary text-to-image generators — either does or does not satisfy them.
  2. Address the ontological objections that motivated the moderate view: that attributing intentionality or agency to artifacts is a category mistake or rests on anthropomorphic projection — either discharging them or showing that they are decisive — rather than redefining 'agency' into a metaphor precisely in order to survive the objection.
  3. Be consistent with, and help explain, the empirical evidence on provenance effects and intention attribution: the convergent AI-label bias in aesthetic evaluation, its moderation by embodiment and framing, and the folk-intuition asymmetry in which robot-made paintings count as art while robots are denied artist status.
  4. State what follows for the limits of creative machines: whether architectural, training, and design constraints — limited generative novelty, fixed chord or generative sets, training-corpus bias — are principled limits on genuine novelty that block the attribution of artistic agency, or merely contingent constraints that permit it, and which artistic values (as opposed to merely aesthetic values) machine-made works can realize under the resulting verdict.

A negative answer — a principled argument that no defensible strong-view account is possible and the moderate view is the maximum — is within scope, and the sub-questions above apply to it as well; the question also covers the main candidate frameworks (functionalist accounts that ground agency in what a system does, relational accounts that ground it in social practices and responses, and distributed or collective accounts that ground it in the human–machine system as a whole) together with the question whether those frameworks can be defended against the empirical malleability findings rather than ignoring them.

References

  1. Steinert, S. (2017). Art: Brought to You by Creative Machines. Philosophy & Technology, 30(3), 267–284. doi:10.1007/s13347-016-0230-6, https://doi.org/10.1007/s13347-016-0230-6
  2. Peterson, M., & Spahn, A. (2011). Can Technological Artefacts Be Moral Agents? Science and Engineering Ethics, 17(3), 411–424. doi:10.1007/s11948-010-9241-3, https://doi.org/10.1007/s11948-010-9241-3
  3. Nanay, B. (2015). Cognitive penetration and the gallery of indiscernibles. Frontiers in Psychology, 5, Article 1527. doi:10.3389/fpsyg.2014.01527, https://doi.org/10.3389/fpsyg.2014.01527
  4. Coeckelbergh, M. (2017). Can Machines Create Art? Philosophy & Technology, 30(3), 285–303. doi:10.1007/s13347-016-0231-5, https://doi.org/10.1007/s13347-016-0231-5
  5. McCormack, J., Gifford, T., & Hutchings, P. (2019). Autonomy, Authenticity, Authorship and Intention in Computer Generated Art. In Computational Intelligence in Music, Sound, Art and Design (EvoMUSART 2019), Lecture Notes in Computer Science (pp. 35–50). Springer. doi:10.1007/978-3-030-16667-0_3, https://doi.org/10.1007/978-3-030-16667-0_3
  6. Chamberlain, R., Mullin, C., Scheerlinck, B., & Wagemans, J. (2018). Putting the art in artificial: Aesthetic responses to computer-generated art. Psychology of Aesthetics, Creativity, and the Arts, 12(2), 177–192. doi:10.1037/aca0000136, https://doi.org/10.1037/aca0000136
  7. Ragot, M., Martin, N., & Cojean, S. (2020). AI-Generated vs. Human Artworks. A Perception Bias Towards Artificial Intelligence? Extended Abstracts of the 2020 CHI Conference on Human Factors in Computing Systems. doi:10.1145/3334480.3382892, https://doi.org/10.1145/3334480.3382892
  8. Gangadharbatla, H. (2022). The Role of AI Attribution Knowledge in the Evaluation of Artwork. Empirical Studies of the Arts, 40(2), 125–142. doi:10.1177/0276237421994697, https://doi.org/10.1177/0276237421994697
  9. Bellaiche, L., Shahi, R., Turpin, M. H., Ragnhildstveit, A., Sprockett, S., Barr, N., Christensen, A., & Seli, P. (2023). Humans versus AI: whether and why we prefer human-created compared to AI-created artwork. Cognitive Research: Principles and Implications, 8(1), Article 42. doi:10.1186/s41235-023-00499-6, https://doi.org/10.1186/s41235-023-00499-6
  10. Hitsuwari, J., Ueda, Y., Yun, W., & Nomura, M. (2023). Does human–AI collaboration lead to more creative art? Aesthetic evaluation of human-made and AI-generated haiku poetry. Computers in Human Behavior, 139, 107502. doi:10.1016/j.chb.2022.107502, https://doi.org/10.1016/j.chb.2022.107502
  11. Seli, P., Ragnhildstveit, A., Orwig, W., Bellaiche, L., Spooner, S., & Barr, N. (2026). Beyond the brush: Human versus artificial intelligence creativity in the realm of generative art. Psychology of Aesthetics, Creativity, and the Arts, 20(4), 984–992. doi:10.1037/aca0000743, https://doi.org/10.1037/aca0000743
  12. Mikalonytė, E. S., & Kneer, M. (2022). Can Artificial Intelligence Make Art?: Folk Intuitions as to whether AI-driven Robots Can Be Viewed as Artists and Produce Art. ACM Transactions on Human-Robot Interaction, 11(4), 1–19. doi:10.1145/3530875, https://doi.org/10.1145/3530875
  13. Romic, B. (2022). Negotiating anthropomorphism in the Ai-Da robot. International Journal of Social Robotics, 14(10), 2083–2093 (published online 2021). doi:10.1007/s12369-021-00813-6, https://doi.org/10.1007/s12369-021-00813-6
  14. Wojtkiewicz, K. (2023). How Do You Solve a Problem like DALL-E 2? The Journal of Aesthetics and Art Criticism, 81(4), 454–467. doi:10.1093/jaac/kpad046, https://doi.org/10.1093/jaac/kpad046
  15. Chiodo, S. (2024). What AI 'art' can teach us about art. Journal of Aesthetics & Culture, 16(1), Article 2395511. doi:10.1080/20004214.2024.2395511, https://doi.org/10.1080/20004214.2024.2395511
  16. Anscomb, C. (2025). AI: artistic collaborator? AI & Society, 40(5), 3419–3429 (published online 2024). doi:10.1007/s00146-024-02083-y, https://doi.org/10.1007/s00146-024-02083-y
  17. Berryman, J. (2024). Creativity and Style in GAN and AI Art: Some Art-historical Reflections. Philosophy & Technology, 37(2), Article 61. doi:10.1007/s13347-024-00746-8, https://doi.org/10.1007/s13347-024-00746-8
  18. Woodward, A. (2025). Hybrid Artistic Agency: Experiment and Reflection with Creative AI. In Artistic Agency (pp. 251–270). Springer. doi:10.1007/978-3-031-97969-9_12, https://doi.org/10.1007/978-3-031-97969-9_12