he empirical papers behind the pages on machine-made art, in date order, with two or three sentences each on what was done and found. Every DOI was checked against Crossref before it went on the shelf; the one preprint links to arXiv. The notes are our summaries, not quotations. The pictures in these studies are not reproduced here: most are in copyright, and the findings do not need them. Many of the images the models learned from were made by people who were never asked, and who are not in any of these reference lists; the shelf has no way to count them.

in the tour: a wing planned for the tour, called "Who made this?", will have a room called “The Hall of Hands” ···

1. Kruger et al. (2004). The effort heuristic. Journal of Experimental Social Psychology. doi:10.1016/S0022-1031(03)00065-9
   Holding a poem or painting constant, the researchers varied how long participants were told it had taken to make. More reported effort raised judged quality, liking and value, most of all when quality was hard to judge. The baseline for every "but a machine did it in seconds" reaction. Used in: reception
2. Newman & Bloom (2012). Art and authenticity: The importance of originals in judgments of value. Journal of Experimental Psychology: General. doi:10.1037/a0026035
   Why is an original worth more than a perfect duplicate? Two accounts survive their tests: the original is the record of a creative performance, and it carries physical contact with the artist, treated almost as contagion. Neither is a property of the surface. Used in: reception
3. Mori, MacDorman & Kageki (2012). The uncanny valley. IEEE Robotics & Automation Magazine. doi:10.1109/MRA.2012.2192811
   The English translation, by MacDorman and Kageki, of Mori's 1970 essay proposing the uncanny valley. A design caution, drawn as a curve without data; the original was published in Japanese. Used in: archive  timeline
4. Chamberlain et al. (2018). Putting the art in artificial: Aesthetic responses to computer-generated art. Psychology of Aesthetics, Creativity, and the Arts. doi:10.1037/aca0000136
   Viewers showed a bias against images they believed computer-generated, but the bias reversed when they saw robotic artists making work. Watching the making supplied what the label took away. Used in: reception  timeline
5. Hong & Curran (2019). Artificial intelligence, artists, and art: attitudes toward artwork produced by humans vs. artificial intelligence. ACM Transactions on Multimedia Computing, Communications, and Applications. doi:10.1145/3326337
   Knowing a work was made by a machine did not, overall, lower its ratings. What did was the viewer's prior belief that machines cannot make art. The effect lives in the belief, not the information. Used in: reception
6. Ragot, Martin & Cojean (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
   With 565 participants, paintings labelled as human-made were rated higher than those labelled as machine-made on liking, beauty, novelty and meaning. An early, clean demonstration of the label penalty. Used in: reception  timeline
7. Gangadharbatla (2022). The Role of AI Attribution Knowledge in the Evaluation of Artwork. Empirical Studies of the Arts. doi:10.1177/0276237421994697
   Participants could not reliably tell machine-made from human-made images, and guessed by genre: representational pictures were credited to people, abstract ones to machines. Used in: reception  timeline
8. Shan et al. (2023). Glaze: protecting artists from style mimicry by text-to-image models. USENIX Security 2023; arXiv:2302.04222. arxiv.org/abs/2302.04222
   A tool built for artists: it adds small changes to an image so that a model fine-tuned on it learns the wrong style. The paper documents artists' own descriptions of style mimicry; a technical answer to a consent problem. Used in: reception  timeline
9. Bellaiche et al. (2023). Humans versus AI: whether and why we prefer human-created compared to AI-created artwork. Cognitive Research: Principles and Implications. doi:10.1186/s41235-023-00499-6
   Labels assigned at random to the same images, so that only the label differed. "Human-made" raised liking, beauty, profundity and judged worth, and the effect ran partly through how much story and effort viewers perceived. Used in: reception  timeline
10. Millet et al. (2023). Defending humankind: Anthropocentric bias in the appreciation of AI art. Computers in Human Behavior. doi:10.1016/j.chb.2023.107707
    Four studies, 1,708 participants. Work believed to be machine-made was judged less creative and evoked less awe, and the bias was stronger in people with anthropocentric beliefs about creativity. Used in: reception  timeline
11. Shan et al. (2024). Nightshade: Prompt-Specific Poisoning Attacks on Text-to-Image Generative Models. 2024 IEEE Symposium on Security and Privacy (SP). doi:10.1109/SP54263.2024.00207
    Nightshade, from the same group as Glaze: it alters training images so that a model learns the wrong concept for a prompt. A deterrent rather than a defence, and argued over on both sides. Used in: reception  timeline
12. Zhou & Lee (2024). Generative artificial intelligence, human creativity, and art. PNAS Nexus. doi:10.1093/pnasnexus/pgae052
    Over four million artworks from more than 50,000 users of an art-sharing site. Adopting text-to-image tools raised output by about a quarter and favourites per view by about half; peak novelty rose while average novelty fell, and visual novelty fell. Productivity, taste and sameness in one dataset. Used in: reception  timeline
13. Porter & Machery (2024). AI-generated poetry is indistinguishable from human-written poetry and is rated more favorably. Sci. Rep.. doi:10.1038/s41598-024-76900-1
    Non-specialist readers judged poems written by a language model to be human more often than real poems by well-known poets, and rated them higher; told a poem was machine-made, they rated it lower. The authors argue readers took their own liking as evidence of a human author. Used in: reception  timeline  essay

see also

	How people respond to machine-made art — the argument these papers make
	timeline: machine-made art — with the court dates
	annotated: human–AI interaction — the neighbouring shelf