Will AI Art Devalue Human Creativity?

doi:10.21203/rs.3.rs-2987022/v1

Download PDF

Article

Will AI Art Devalue Human Creativity?

https://doi.org/10.21203/rs.3.rs-2987022/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 03 Nov, 2023

Read the published version in Scientific Reports →

You are reading this latest preprint version

The contemporary art world is conservatively estimated to be a $65 billion USD market that employs millions of human artists, sellers, and collectors across the world. Yet recent attention paid to AI-made art in prestigious galleries, museums, and popular media has provoked debate around how these statistics will change. Unanswered questions fuel growing anxieties. Are AI-made and human-made art evaluated in the same ways? How will growing exposure to AI-made art impact evaluations of human creativity? Our research uses a psychological lens to explore these questions in the realm of visual art. We find that people devalue AI-made art across a variety of dimensions, even when they report it is indistinguishable from human-made art, and even when they believe it was produced collaboratively with a human. But we also find that comparing human and AI-made art increases perceptions of human creativity, an effect that can be leveraged to increase the value of human effort. Our results are robust across six experiments (N = 2,965) using a range of artistic stimuli and incorporating representative samples of the US population. Finally, we highlight conditions under which effects can be strengthened or weakened as well as dimensions where AI-devaluation effects are stronger.

Biological sciences/Psychology

Biological sciences/Psychology/Human behaviour

The contemporary art world is conservatively estimated to be a $65 billion USD market that employs millions of human artists, sellers, and collectors across the world [1]. Yet recent attention paid to AI-made art in prestigious galleries [2, 3], museums [4], and popular media [5] has provoked heated debate around how these statistics will change in the future [6, 7]. Anxiety around the changing value of human art is fueled by unanswered questions like: Will AI-made and human-made art be evaluated in the same ways? Should art markets even treat AI-made art as “art”? And how will growing exposure to AI-made art impact evaluations of solitary human creativity or evaluations of human artists using AI?

Historical examples from other industries provide ample evidence that, on average, automation decreases the value of human goods and labor [8]. But there is also reason to believe that the development of AI technologies capable of automating creativity (e.g., producing visual art largely indistinguishable from human art) should not similarly impact perceptions of human art and creativity. As the famous American painter James Whistler once said, "An artist is not paid for his labor but for his vision" [9]. Past research supports Whistler’s point. Whether made by experts or lay audiences, evaluations of art often depend upon both aesthetic and social dimensions that can be disentangled from the more tangible costs of production and labor [10, 11]. For example, an artist’s use of color and emotion, the complexity of their subject matter, and the artist’s brand all impact perceptions of value and creativity assigned to a given piece of art [12, 13]. Subjective factors like these make it difficult to predict whether the introduction of AI-made art will negatively impact the value of human creativity in artistic domains. Put another way, will the value of human art (monetary and aesthetic) increase or decrease when evaluated next to comparable works produced by AI programs? Is it true that, as some have suggested, creative jobs are “the last bastion of humanity” [14], or will the value of human labor decrease as has happened during so many other industrial revolutions of the past [8]?

The current research adopts a psychological perspective to begin answering these questions. We examine how artistic source attributions (e.g., human-made or AI-made) influence evaluative judgment of lay audiences across six experiments (total N = 2,965). To take a comprehensive view, we assess an array of different dimensions that include estimates of monetary value and labor as well as more artistic evaluations, such as perceptions of skill and creativity. Results reveal three key findings. First, even when people report an inability to differentiate between the two, they consistently devalue AI-made art relative to human-made art. This is true even when the art itself is held constant (i.e., labeling the same piece as “AI-made”) and regardless of participants’ overall feelings towards AI. Echoing historical examples of automation in other industries, this devaluation is more pronounced on evaluations of skill and monetary value and is less pronounced on artistic dimensions (e.g., evaluations of complexity or emotional intensity). Moreover, devaluation effects weaken substantively when participants are not directly comparing human and AI-made art. Our second key finding is that human-made art is seen as more creative as a function of exposure to AI-made art. That is, human art gains creative value when it is directly compared to AI-made art relative to when it is compared to other human-made art. This finding is surprising and important, because contrary to predictions expressed in popular media outlets, it suggests human artists may benefit from comparisons between their work and AI-made art. Our third and final key finding is that although collaborative art (humans working with AI) is perceived to be less valuable than work made only by humans, perceptions of the human artist’s status as the primary creative agent depends largely upon whether the collaboration is being compared to human or AI-made references. That is, an evaluative bias against AI-made artwork persists even in circumstances where the AI functions as a human aid but these detrimental effects on value can be moderated by anchoring the evaluator on the efforts of AI art produced without the help of humans.

Results and Discussion. In Experiment 1, participants (N = 119) were presented with 28 different images (observations = 3,332) that were pretested to capture a range of different artistic styles (see Fig. 1). Each image was randomly labeled such that every participant evaluated 14 images labeled as AI-made and 14 images labeled as human-made. These were evaluated on a battery of artistic dimensions that included how bright, colorful, complex, emotional, skillful, inspiring, expensive, and likable participants found each image to be (ɑ = .97).

Figure 1 displays our results. Overall, within-subjects comparisons revealed images labeled as AI-made were evaluated less favorably, with an aggregate effect of t(118) = 5.52, p < .001, m_diff = − .18, 95% CI[-.12, − .25], d = .51. The direction and significance of this effect remained unchanged when considering each dimension separately (Table 1) and when using multilevel regression models to control for variance attributable to specific image-styles or individual preferences (i.e., models controlling for image effects and nesting within participant).This effect was most pronounced on evaluations of expensiveness (m_diff = − .45, d = .57) and skill (m_diff = − .30, d = .48)—and occurred despite the majority of participants (> 70%) reporting they would not have been able to differentiate between human and AI-made images without the labels provided. That is, even though most participants reported art labeled as AI-made was largely indistinguishable from art labeled as human-made, they still evaluated art labeled as AI-made art less favorably. These results suggest a general bias against AI-made art.

Table 1

*Experiment 1 Evaluation Differences Between Human and AI-Labeled Images (Within-Subjects)*
Dimension	Mean Difference	Effect Size (Cohen’s d)
Expensive	− .45***	.57
Skillful	− .30***	.48
Complex	− .18***	.36
Colorful	− .12**	.30
Liking	− .12**	.25
Emotional	− .12***	.27
Bright	− .08*	.21
Inspiring	− .09*	.22
Overall (ɑ = .95)	− .18***	.51
Note. P-values reflect paired t-tests comparing average ratings by participants on images labeled as Human-made vs. images labeled AI-made. p < .05, p < .01, **p < .001

Results and Discussion. In Experiment 2, participants (n = 415) evaluated the same 28 images (observations = 11,620) using the same dimensions from Experiment 1 alongside an additional “willingness to pay” measure (ɑ = .88). Two key changes were made. First, we used a between-subjects design to provide a more conservative estimate of the bias observed in Experiment 1. Participants were randomly assigned to one of three conditions where (a) all images were either unlabeled (i.e., presumed to be human-made), (b) all images were labeled as AI-made, or (c) all images were unlabeled, but participants were told in advance that some images were made by humans and others by AI (i.e., ambiguous source information). Second, we expanded our research focus by asking participants to report whether they believed each image qualified as “art” or not (1 = no, 2 = maybe, 3 = yes).

OLS models dummy-coding conditions revealed images labeled as AI-made were evaluated less favorably overall, averaging across dimensions, t(11,618) = 2.29, p = .02, m_diff = − .06, 95% CI[-.01, − .12], d = .05). These effects were largely driven by AI-labeled images being rated as less expensive (m_diff = − .24, d = .15), skillful (m_diff = − .13, d = .08), and emotional (m_diff = − .12, d = .07). Participants were also less likely to say that AI-labeled images qualified as art t(11,618) = 7.19, p = .02, m_diff = − .10, 95% CI[-.07, − .13], d = .15), though it is notable that most of the participants in the AI-labels condition (87%) still considered the majority of images to qualify as art (see Table 1 in our Supplementary Information for detailed results of each separate evaluative dimensions and condition). These results replicate our first Experiment, indicating the bias against AI-made art is particularly pronounced on dimensions of artistic skill and monetary value, with images labeled as AI-made images being less likely to qualify as art.

Results and Discussion. In Experiment 3, we examine how evaluations of human creativity are impacted by comparisons made between human and AI-made art. Specifically, we randomize the order of image labels to assess how human artists are evaluated after exposure to AI-made art. All participants (n = 405) evaluated two images. These images were selected from the stimuli used in our previous Experiments because they were rated similarly across artistic dimensions (see Fig. 1 in our Supplementary Information). Participants were randomly assigned to either a control condition where both paintings were labeled as human-made or one of three experimental conditions where: (a) both paintings were labeled as AI-made, (b) the first painting was labeled human-made and the second painting as AI-made, or (c) the first painting was labeled as AI-made and the second painting as human-made. Holding the images constant in this way allowed us to determine the extent to which a single piece of art gains or loses value as a function of comparisons being made between human and AI-made art. In addition to evaluating aesthetic dimensions, participants estimated the monetary value of each painting, as well as the skill, talent, and execution shown by the labeled artist (human or AI). To increase external validity, we told each participant these images were paintings currently for sale at a private gallery and included information given by the gallery (e.g., “Untitled, Oil on Canvas…”).

Both paintings were evaluated as being worth less money when labeled as AI-made (see Tables 2 and 3 in our Supplementary Information); and the AI artist was evaluated as less capable on dimensions of skill, talent, and execution. However, rather than detracting from evaluations of human creativity, comparisons between human and AI-made art positively impacted evaluations of human effort. The second painting was rated higher on aesthetic dimensions when it was labeled human-made and followed a painting labeled as AI-made (see Table 3 in our Supplementary Information). That is, the exact same painting was more favorably when labeled as human-made and compared against an AI-made painting than when it was labeled as human-made and compared against another human-made painting.

Results and Discussion. Experiment 4 was preregistered and conducted to conceptually replicate our finding in Experiment 3. Are evaluations of human creativity positively impacted by exposure to AI-made artwork? Participants (n = 789) were shown the same two images used in Experiment 3 but in place of aesthetic dimensions they were asked to evaluate creativity directly. Specifically, participants rated how creative, novel, likable, and appropriate to be sold in a gallery each painting. In line with psychological research on creativity, these were averaged to construct overall measures of creativity (Image 1 ɑ = .78, Image 2 ɑ = .82). Participants were also asked to estimate monetary value and production-time (i.e., how much time in hours they thought it took to produce each painting). These were used as control variables in supplementary analysis to see if effects observed in previous studies could be explained by participants that might reasonably feel more threatened by the potential that AI represents. Again, labels were randomized such that participants were assigned to either (a) a control condition where both paintings were labeled as human-made, or one of two experimental conditions where (b) the first painting was labeled as human-made and the second labeled as AI-made, or (c) the first painting was labeled as AI-made and the second labeled as human-made.

Overall, both paintings in Experiment 4 were rated as less creative, worth less money, and estimated to have taken less time to produce when labeled as AI-made (see Table 2 below and Table 4 in our Supplementary Information Material). Replicating and extending the positive effect observed in Experiment 3, the second painting was evaluated as more creative when it was labeled as human-made and followed an AI-made painting (see Fig. 2). Again, this indicates comparisons of human and AI-made art increased the perceived creativity of humans. That is, the exact same painting was judged to be more creative, novel, likable, and appropriate to be sold in a gallery when it was labeled human-made and compared against an AI-made painting than when labeled as human-made and compared against another human-made painting. Results did not change when controlling for participants’ experience in artistic or technological industries in supplementary regression models. This suggests observed effects are not driven by anxieties specific to participants whose passions or livelihoods may be more by the introduction of AI.

Table 2

*Experiment 4 Evaluation Means for Image 2*
	Human-Anchor (Control Condition)	AI-Anchor (Experimental Conditions)
Dimension	Image 2 Human Label	Image 2 Human Label	Image 2 AI Label
Creative	4.62 (.08)	4.85 (.08)*	4.24 (.08)***
Monetary Value	148.35 (11.86)	152.99 (19.73)	111.74 (13.30)
Estimated Time to Produce	36.02 (1.43)	32.60 (1.46)	18.29 (1.35)***
Note. P-values reflect regression models dummy-coding conditions and using the “human-label” group as our control condition. p < .05, p < .01, **p < .001

Results and Discussion. In Experiment 5 (preregistered), we test the generalizability of effects observed in Experiments 3 and 4 by recruiting a representative sample of the US population (n = 709) to evaluate a larger set of images. Each participant was asked to evaluate two images randomly selected from those used in our first experiment. Labels were randomized such that participants were assigned to either (a) a control condition where both paintings were labeled as human-made, or (b) an experimental condition where the first painting was labeled as human-made and the second labeled as AI-made. Creativity, monetary value, and production time were all assessed using the same measures from Experiment 4. As a control variable, participants indicated their opinion towards AI using four items (ɑ = .77) from the General Attitudes Towards Artificial Intelligence Scale [15].

Replicating the bias against AI-made art documented in earlier studies, paintings were evaluated as less creative, worth less money, and estimated to have taken less time to produce when labeled as AI-made (see Table 5 in our Supplementary Information). Again, human art was seen as more creative when it followed AI-made art. Specifically, paintings labeled as human-made were seen as more creative when they followed paintings labeled as AI-made than when they followed paintings labeled human-made (see Table 3 and Fig. 3). Moreover, the direction and significance of these results was unchanged when using attitudes towards AI as a control variable. This suggests these effects are not driven by a general distaste toward AI. That is, regardless of how participants felt about AI, they generally (a) evaluated AI-made art less favorably and (b) saw human made art as more creative after evaluating AI-made art.

Table 3

*Experiment 5 Evaluation Means for Image 2*
Dimension	Image 2 (Post-Human)	Image 2 (Post-AI)	Mean Difference
Artistic Value	4.95 (.06)	5.14 (.06)	t(688.18) = 2.26, p = .024
Monetary Value	132.87 (6.44)	167.83 (19.33)	t(467.25) = 1.72, p = .087
Estimated Time to Produce	30.51 (1.23)	30.94 (1.18)	t(696.81) = .25, p = 0.81
Note. Means and standard error reported above. P-values reflect between-subjects T-tests comparing the left and right columns. The direction and significance of effects remained unchanged using OLS regression to additionally control for variation attributable to images and individuals (e.g., anxiety about AI technologies and artistic experience).

Results and Discussion. Because AI technologies might be adopted to supplement or facilitate the creativity of many human artists, Experiment 6 (preregistered, N = 527) used a representative sample of the US population (n = 527) to test whether the effects observed in our previous experiments extend to art collaboratively produced by humans and AI. Participants rated two randomly ordered paintings (see Fig. 2 in our Supplementary Information) that were pretested to make sure they were comparable in terms of creativity (p = .77), monetary value (p = .54), and estimated production time (p = .60). Participants were randomly assigned to one of two conditions where the first painting was either labelled as (a) human-made or (b) AI-made, and the second painting was always labeled as collaboratively made: “The following painting was created by the artist Avery Taylor, collaborating with an artificial intelligence program capable of imagining and painting painting entirely of its own accord, in January of 2019.” Creative value, monetary value, and estimated production-time were all measured using the same items from Experiment 5. Finally, to test whether humans are seen as more or less responsible for creative output when working with an AI, participants estimated the distribution of labor for the collaborative painting using a 100-point slider ranging from “All AI Effort” to “All Human Effort.”

Replicating Experiments 1–5, the first painting was evaluated less favorably on all dependent variables when labeled as AI-made (see Table 6 in our Supplementary Information). On average, the collaborative painting was rated less favorably than human-labeled painting and more favorably than AI-labeled painting (see Table 4). Consistent with our later experiments, participants estimated the collaborative human was responsible for a greater portion of creative labor (53.94%) when the collaboration was compared to an AI-made painting, but a smaller percentage of creative labor (36.76%) when the collaboration was compared against an human-made painting: t(514.1) = 9.09, p < .001, m_diff = 17.18, 95% CI[ 13.01, 21.35], d = .71 (see Fig. 4). In sum, these results indicate two important findings. First, the bias against AI-made art persists even when art is collaborative production of humans working with AI. Second, estimates regarding human labor in a collaboration with AI depended largely upon whether the collaborative piece was being compared to solitary human or AI efforts. That is, when compared against AI-made art, humans are seen as more responsible for the creative labor and when compared against human-made art, humans are seen as less responsible for the creative labor.

Table 4

*Experiment 6 Mean Difference Between “Collaboration” Painting and Anchor (Within-Subjects)*
Dimension	Collaboration vs. Human Anchor	Collaboration vs. AI Anchor
Creative Value	− .58 (.09)***	.46 (.07)***
Monetary Value	-126.60 (36.13)***	30.06 (5.25)***
Estimated Time to Produce	-16.45 (1.62)***	10.64 (1.03)***
Note. P-values and standard error are from paired t-tests. Significant effects reported above remained unchanged when additionally including evaluations of image 1 as control variable in an OLS regression model. p < .05, p < .01, **p < .001

Do we evaluate AI-made art and human-made art in the same way? Our findings suggest not. Across six experiments, we find that even when the content of the art is held constant, simply believing it was made by AI negatively impacted appraisal. This bias against AI-made art was particularly evident on dimensions of monetary value and skill (see Experiments 1 and 2). On the one hand, these findings are in line with anxieties expressed in popular media outlets that like other industries affected by automation in the past, AI-made art is likely to bring down the average monetary value in artistic markets. Indeed, the effects we observe are substantive. For example, AI-made labels in our last experiment led to a 62% decrease in monetary value and a 77% decrease in estimated time to produce relative to human effort. On the other hand, our findings reveal that AI-made images are less likely to be consider “art” at all and that devaluation effects can be mitigated by ensuring audiences do not directly compare human and AI efforts (see Experiments 2 and 3). This has clear practical implications. For example, the creative value humans currently command in markets may be less subject to change if efforts are made to segregate or separately categorize human art from AI-made goods. Indeed, one recent proposition that may help is to adopt the designation “synthography,” to create psychological distance between AI-art and comparable forms of human art like photography and digital art [16].

Will evaluations of human creativity be impacted by AI-made art? Our findings suggest yes. We find although participants consistently devalued art if they believed it was made by AI, human-made art was evaluated as more creative when compared against AI-made art than when compared against other human-made art. This refutes sensationalized sentiments “art is dead” [5] because of the introduction of AI. If anything, our results predict AI art will lead audiences to see human creativity in a new light.

How will collaborations between humans and AI (e.g., when human artists use AI tools) be received in creative industries? Here, our findings are nuanced but illustrative. Participants evaluated human-AI collaborations as more valuable than AI-made art but less valuable than human-made art. Importantly, however, we find that the human’s status as the primary creative agent (i.e., the target more responsible for the output) in a collaboration depends upon whether the collaboration is compared to solitary human or AI efforts. This finding has significant practical value, because maintaining status as the primary creative agent implies the ability to capture a greater portion of economic value. In short, is suggests human artist who use AI-tools would be smart to make comparisons of their own art to art made AI (without the assistance of humans).

Humans are only just beginning to understand the impact AI technologies will have on the value of human creativity, but it is worth noting that the heated debates it has already provoked are strikingly reminiscent of initial reactions to the invention of the camera. For example, in response to the camera, the famous French painter Paul Delaroche declared, “From today, painting is dead” [17]. Similarly, the famous art critic Charles Baudelaire said, “it will not be long before it [photography] has supplanted or corrupted art altogether” [18]. In some ways, their anxieties were justified. The livelihoods of 19th century portraiture artists were threatened by the invention of the camera [19]. And yet, the camera also gave us new ways to look at the world, defining a period of innovation that eventually led to invention and appreciation of forms of art that include impressionism, cubism, and digital photography [20]. Likewise, today respected media outlets and artists alike have flooded the internet with responses to AI like, “More than ever before…I’m concerned for the future of human creativity” [6, 21, 22]. Our findings allow us to imagine a different future than the currently prevailing forecasts of doom and gloom, one where the value of human creativity persists. Just as the invention of photography ultimately inspired innovations like impressionism [23], we suggest that a heightened appreciation for human creativity can provide a fertile bed for new forms of human creativity and expression.

We report all conditions, measures, data exclusions, and provide copies of all study materials on our Open Science Framework page (see Data and Code Availability). All research protocols were approved by the *name anonymized for review* University Human Research Protection Office and Institutional Review Board and performed in accordance with relevant guidelines/regulations (in accordance with the Declaration of Helsinki), including informed consent from all participants. Notably, our first two experiments were run in 2017 and 2018, before the more recent introduction of AI-art innovations like Midjourney and DALL-E2. In contrast, our last experiments were run in 2023 just as Midjourney Version 4 and DALL-E2 were beginning to reach a national audience. We share these dates because we believe it is important to note, historically, that our data was collected both before and after more recent (and sensationalized) coverage of AI art in prestigious media outlets like the New York Times [5] and Washington Post [6], documenting evaluations of art just as the implications of AI were beginning to be realized.

Sample size determination and randomization

All sample sizes were determined before collecting data, with data collection halting after analysis began. For Experiment 1, we hypothesized a small to medium effect size (Cohen’s d = .33), which we used to determine our sample size (i.e., roughly 95% power to detect an effect). We then used effect sizes from earlier experiments to make sample size determinations for later experiments. Data quality was ensured in several ways. For instance, we removed duplicate responses (e.g., repeated IP addresses or research IDs), removed participants who failed attention checks, and collected entirely new samples for every study.

Data analysis and reporting

All data analysis was conducted in R (v.4.2.2). Effect sizes were calculated as Cohen's d using the ‘effectsize’ package [24]. Whenever a T-test in our analysis did not demonstrate equal variance, a Welch's T-test with corrected degrees of freedom was used instead. Reported p values are all two-sided. Finally, all regression models include dummy-coded conditions that compare outcomes from against the control condition designated for that experimental design.

Experimental samples and procedures

Pilot Study. We pre-tested the 28 images used in Experiments 1, 2, and 5 in a pilot study (n = 105). This was done to ensure our stimuli captured a range of styles and quality. Half of these images were lesser-known paintings from respected artists (e.g., William Gear, Andy Warhol, and Paul Gauguin) while the other half were AI-generated images rendered in the same styles of those artists. These stimuli were chosen and tested to ensure that (a) participants could not tell the difference between human and AI-made art and (b) so that participants in Experiments 1 and 2 would be presented with style-matched pairs randomly labeled as “human-made” or “AI-made”. Results confirmed images represented a range of quality (m = 4.31, sd = 1.46) and that participants generally could not tell the difference between images that were or were not AI-made. For example, when asked to guess the origin of an image (1 = definitely human-made, 7 = definitely AI-made), responses were generally below the midpoint (regardless of each image’s actual origin), indicating participants thought images looked more “human-made” on average (human-made m = 2.79, sd = 1.58 vs. AI-made m = 2.89, sd = 1.56). To ensure our data accurately represented lay evaluations of stimuli participants were unfamiliar with (i.e., experimental fidelity) we used the question “Before taking this survey, had you seen any of these paintings before?” Participants who responded yes to this question in any experiment were removed before any analysis, though supplementary analysis including their responses did not change the direction or significance of effects reported in this article.

Experiment 1. We recruited 143 English speaking US residents from Mturk. Participants were excluded for failing to pass attention checks or reporting they recognized stimuli used in the study, yielding a final sample of n = 119 participants (men = 52%, ${m}_{age}=34$ ). Participants were paid $2 to complete the survey.

After rating three buffer images to acclimate participants to the task, all participants rated 14 images labeled as human-made and 14 images labeled as AI-made. Images were all presented in a random order. To make sure that differences in artistic style did not confound any results, labels were randomly assigned within style-matched pairs (see our pretest) such that one image in each style-matched pair was always labeled as AI-made and the other as human-made. This allowed us to make comparisons between images labeled as human or Ai-made while holding style constant. Participants then rated each painting on a battery of dimensions: how much they liked it, how skillfully it was painted, how colorful it was, whether they found it inspiring, how bright it was, how complex it was, how emotionally evocative, and whether they thought it was expensive (1 = Not at all, 7 = A great deal; ɑ = .88). For exploratory purposes we also asked participants about their general affinity for art (e.g., “Some people seem to need art in their lives more than others; I consider myself that kind of person.”) and their feelings about technological innovations (e.g., “I tend to dislike new technologies.”). Supplementary analysis revealed these had no impact on our main findings.

Experiment 2. We recruited 555 English speaking US residents from Mturk. Participants were excluded for failing to pass attention checks, comprehension checks, or reporting they recognized any stimuli used in the experiment, yielding a final sample of n = 415 participants (men = 51%, ${m}_{age}=36$ ). Participants were paid $2 to complete the survey.

After rating three buffer images to acclimate participants to the task, all participants rated 28 images in random order. Participants were randomly assigned to one of three conditions. In a control condition they were told we were simply interested in “how people perceive each painting on a number of dimensions.” In experimental conditions participants were either told “each painting was made by an artificial intelligence” or that “some of these paintings were made by a human and others were made by an artificial intelligence” but not which ones. Participants then rated each painting on a battery of dimensions: how much they liked it, how skillfully it was painted, how colorful it was, whether they found it inspiring, how bright it was, how complex it was, how emotionally evocative, whether they thought it was expensive, and how much they’d be willing to pay (1 = Not at all, 7 = A great deal; ɑ = .94). For exploratory purposes we also asked participants about their mood during the study (e.g., “Overall, my mood is:” = -10 = Very unpleasant, 10 = Very pleasant) and about their personal tastes in art (e.g., “I feel I have good taste in art.”). Supplementary analysis revealed that these did not differ by condition and had no impact on our main findings.

Experiment 3. We recruited 541 English speaking US residents from Mturk. Participants were excluded for failing to pass attention checks, comprehension checks, or reporting they recognized stimuli used in the study, yielding a final sample of n = 405 participants (male = 53%, ${m}_{age}=38$ ). Participants were paid $1 to complete the survey.

To increase the external validity of our findings, participants were given a cover story about these images representing real paintings for sale at a private gallery:

On the next page, you'll be shown two images of paintings currently for sale at the Lenham Private Gallery. We are curious about consumer impressions of these paintings and the blurbs attached to them. Please review the painting and information provided by the gallery and answer all questions honestly

Participants were then randomly assigned to one of four conditions where they rated two images. In a control condition, both images were labeled as human-made. In one experimental condition both images were labeled as AI-made. In another experimental condition, the first image was labeled human-made and the second AI-made. And in a final experimental condition, the first image was labeled AI-made and the second human-made. Image order was always held constant. Human and AI-made labels read as follows, “The following painting was created by Jamie Kendricks, in January of 2019.” or “The following painting was created by an artificial intelligence program, which imagines and paints images entirely of its own accord, in January of 2019.” Building upon our cover story, all paintings were presented with unique ID numbers and fabricated gallery information (e.g., “Lenham Private Gallery ID: #A2461; Untitled, 2019; Oil on canvas; 24 in x 36 in). Participants rated each image on a battery of dimensions: how they liked each painting, how skillfully it was painted, how colorful it was, whether they found it inspiring, how bright it was, how complex it was, how emotionally evocative, whether the creator was talented, and whether they were impressed by the execution (1 = Not at all, 7 = A great deal;${\alpha }_{image 1}=.86, {\alpha }_{image 1}=.90$).

In addition, direct estimates of monetary value were obtained on a separate page immediately after participants evaluated each painting on the dimensions listed above. On this page, participants were informed about pricing with the prompt: “The average painting in the Lenham Gallery sells for somewhere between $50 to $220, with most pieces retailing at $150.“ They were then asked, “How much do you personally think the Lenham gallery should sell this painting for?” and “Assuming that you wanted this painting and given the gallery's prices, how much would you pay to acquire it?” For exploratory purposes, we asked participants about their taste in art (e.g., “Compared to other people, I generally have a better eye for art.” and “I like artwork that depicts "real things" more than I like artwork that is abstract.”). Supplementary analysis revealed artistic taste had no impact on our main findings.

Experiment 4. We recruited 792 English speaking US residents from Prolific. Participants were excluded for failing to pass attention checks, comprehension checks, or reporting they recognized any stimuli used in the study, yielding a final sample of n = 789 participants (male = 49%, ${m}_{age}=38$ ). Participants were paid $1 to complete the survey. Our pre-registration can be found here: https://aspredicted.org/DJV_MN7 .

Participants were randomly assigned to one of three conditions where they were asked to evaluate two images. In a control condition, both images were labeled as human-made. In one experimental condition, the first image was labeled human-made and the second AI-made. In another experimental condition, the first image was labeled AI-made and the second human-made. We used the same labels and gallery information provided in Experiment 3. The perceived creativity of each image was measured by asking participants to rate how much they thought each painting was creative, novel, appropriate to be sold in a gallery, and likable (${\alpha }_{image 1}= .78, {\alpha }_{image 2}= .82$). As in Experiment 3, participants were then given information about pricing on a separate page and asked to estimate the monetary value of each painting. Additionally, participants were additionally asked to estimate labor with the item: “How many hours of active painting time do you think it took to create the painting above?” Finally, to make sure effects were not confounded by expertise in domains of art and technology, participants responded to five items about artistic experience (e.g., “I used to [or currently] work in a job that primarily deals with the visual arts [e.g. designer, gallery manager, art dealer].”; $\alpha = .79)$ and five items about technological experience (e.g., “I used to [or currently] work in a job that primarily deals with computer programming, data science, or engineering.”: $\alpha = .72)$. Expertise did not differ by condition and supplementary analysis revealed it had no impact on our main findings.

Experiment 5. We recruited 731 English speaking US residents, using Prolific to collect a representative sample of the U.S. population. Participants were excluded for failing to pass attention checks, comprehension checks, or reporting they recognized stimuli used in the study, yielding a final sample of n = 710 participants (male = 48%, ${m}_{age}=45$ ). Participants were paid $1 to complete the survey. Our pre-registration can be found here: https://aspredicted.org/GJ4_VS4 .

Participants responded to the same survey used in Experiment 4 with two differences. First, images were randomly selected and ordered from the larger pool of 28 pretested images used in Experiments 1 and 2. Second, participants were asked at the end of the survey to indicate their own specific attitudes toward AI (e.g., “I think Artificial Intelligence programs are a threat to human artists.” and “I think Artificial Intelligence programs are an exciting new tool for human artists.”; $\alpha = .77)$. Notably, though participants who felt anxious about AI technology rated AI-labeled artwork less favorably overall, attitudes towards AI did not differ by condition and supplementary analysis using attitudes towards AI as a control variable had no impact on our main findings.

Experiment 6. We recruited 698 English speaking US residents, using Prolific to collect a representative sample of the U.S. population. Participants were excluded for failing to pass attention and comprehension checks yielding a final sample of n = 527 participants (male = 45%, 49). Participants were paid $1 to complete the survey. Our pre-registration can be found here: https://aspredicted.org/DXP_K4M .

Participants were given the same cover story and prompt used at the beginning of Experiments 4 and 5 (i.e., rating “paintings currently for sale at the Lenham Private Gallery”) before being presented with two images in random order. These images were not drawn from our previous studies but were instead both generated by the AI program Midjourney and pretested on a separate sample (see our supplementary materials online) to ensure they were comparable in terms of participant evaluations of creativity, estimated monetary value, and estimated production time.

Participants were randomly assigned to one of two conditions. In a control condition, the first image was labeled with the tag: “The following painting was created by Jamie Kendricks, in January of 2019.” In our experimental condition, the first image was labeled with the tag: “The following painting was created by an artificial intelligence program, which imagines and paints images entirely of its own accord, in January of 2019.” For participants in both conditions, the second image was always labeled with the tag: “The following painting was created by the artist Avery Taylor, collaborating with an artificial intelligence program capable of imagining and painting images entirely of its own accord, in January of 2019.” Participants rated both images using the same items from Experiment 5 as well as an additional question, specific to the second image, that asked “how much work do you think was done by the AI vs the human?” using a sliding scale (0 = All AI Effort, 100 = All Human Effort).

Data and Code Availability

All data, survey materials, and code are available on the Open Science Framework at https://osf.io/xs8bv/?view_only=96a2b4c29a5b4adf8cca18025e18811c.

Acknowledgements

We provide thanks to ***, Chairman, Department of Physiology at *** Medical School for providing early review and feedback; ***, Professor of Innovation, Director of ***, and esteemed robotics expert for feedback and guidance during the preparation of this manuscript. We also thank ** for historical fact-finding that contributed to this project.

*** is used above to indicate names anonymized for submission.

McAndrew, C. (2022, March 22). The Art Basel and UBS Survey of Global Collecting 2022. Retrieved March 15, 2023, from https://www.ubs.com/global/en/our-firm/art/collecting/art-market-survey.html
Graham, T. (2018, December 12). Art made by AI is selling for thousands – is it any good? BBC. Retrieved May 9, 2023, from https://www.bbc.com/culture/article/20181210-art-made-by-ai-is-selling-for-thousands-is-it-any-good
Sutton, B. (2019, March 6). Artificial Intelligence Artwork by Mario Klingemann Sells for £40,000 at Sotheby’s. Artsy. Retrieved January 12, 2023, from https://www.artsy.net/article/artsy-editorial-artwork-created-ai-sold-40-000-sothebys-failing-generate-fervor-propelled-ai-work-sell-40-times-estimate-year
Anadol, R. (2022). Unsupervised — Machine Hallucinations — MoMA. The Museum of Modern Art, New York, November 19, 2022–April 15, 2023. Retrieved May 8, 2023, from https://www.moma.org/calendar/exhibitions/5535
Roose, K. (2022, August 3). An A.I.-Generated Picture Won an Art Prize. Artists Aren’t Happy. New York Times. Retrieved January 10, 2023, from https://www.nytimes.com/2022/09/02/technology/ai-artificial-intelligence-artists.html
Burnett, T. B., & Taplin, J. (2023, March 14). Opinion | to protect human artistry from AI, new safeguards might be essential. The Washington Post. Retrieved March 20, 2023, from https://www.washingtonpost.com/opinions/2023/03/14/artificial-intelligence-threatens-creative-artists/
Santos, R. (2022, November 8). Can AI-Generated Art Replace Creative Humans? VICE. Retrieved January 10, 2023, from https://www.vice.com/en/article/epzkwm/artificial-intelligence-art-creatives-ai
Xu, M., David, J. M., & Kim, S. H. (2018). The Fourth Industrial Revolution: Opportunities and Challenges. International Journal of Financial Research, 9(2), 90–95.
Garg, A. (2005, p.163). Another Word a Day: An All-new Romp Through Some of the Most Unusual and Intriguing Words in English. John Wiley & Sons.
Spee, B. T., Pelowski, M., Arato, J., Mikuni, J., Tran, U. S., Eisenegger, C., & Leder, H. (2022). Social Reputation Influences on Liking and Willingness-To-Pay for Artworks: A Multimethod Design Investigating Choice Behavior Along with Physiological Measures and Motivational Factors. PloS One, 17(4), e0266020.
Wohl, H. (2021). Bound by creativity: How contemporary art is created and judged. University of Chicago Press.
Marin, M. M., & Leder, H. (2013). Examining complexity across domains: relating subjective and objective measures of affective environmental scenes, paintings and music. PloS one, 8(8), e72412.
Moulard, J. G., Rice, D. H., Garrity, C. P., & Mangus, S. M. (2014). Artist authenticity: How artists’ passion and commitment shape consumers’ perceptions and behavioral intentions across genders. Psychology & Marketing, 31(8), 576–590.
Salter, P. (2022, April 11). David Glick On The Business Creativity, Alexander McQueen and Venture Capital. Forbes. https://www.forbes.com/sites/philipsalter/2022/04/11/david-glick-on-the-business-creativity-lee-mcqueen-and-venture-capital/?sh=7b524fe57362
Schepman, A., & Rodway, P. (2020). Initial validation of the general attitudes towards Artificial Intelligence Scale. Computers in Human Behavior Reports, 1, 100014.
Reinhuber, E. (2022, February). Synthography–An Invitation to Reconsider the Rapidly Changing Toolkit of Digital Image Creation as a New Genre Beyond Photography. In ArtsIT, Interactivity and Game Creation: Creative Heritage. New Perspectives from Media Arts and Artificial Intelligence. 10th EAI International Conference, ArtsIT 2021, Virtual Event, December 2–3, 2021, Proceedings (pp. 321–331). Cham: Springer International Publishing.
Ballinetti, C. (2019, April 8). "From Today Painting is Dead": Photography's Revolutionary Effect. Art & Object. Retrieved March 15, 2023, from https://www.artandobject.com/news/today-painting-dead-photographys-revolutionary-effect
Baudelaire, C. (1955). The Salon of 1859. The Mirror of Art, 230–231.
McCouat, P. (2018). Early Influences of Photography on Art. Journal of Art in Society. Retrieved March 3, 2023, from https://www.artinsociety.com/pt-1-initial-impacts.html
Martinique, E. (2019, October 12). How Did Photography Influence the Impressionists? Widewalls. Retrieved March 3, 2023, from https://www.widewalls.ch/magazine/impressionists-photography-museo-thyssen-bornemisza
Marcus, J. (2022, December 10). Artists Decry Use of AI Art: 'I'm concerned for the future of human creativity'. The Independent. Retrieved March 15, 2023, from https://www.independent.co.uk/news/world/americas/ai-art-lensa-magic-avatar-b2242891.html
Shaffi, S. (2023, January 23). ‘It’s the Opposite of Art’: Why Illustrators Are Furious About AI. The Guardian. Retrieved March 15, 2023, from https://www.theguardian.com/artanddesign/2023/jan/23/its-the-opposite-of-art-why-illustrators-are-furious-about-ai
Silva, E. (2022, June 4). How Photography Pioneered a New Understanding of Art. The Collector. https://www.thecollector.com/how-photography-transformed-art/
Ben-Shachar, M. S., Lüdecke, D., & Makowski, D. (2020). effectsize: Estimation of effect size indices and standardized parameters. Journal of Open Source Software, 5(56), 2815.

No competing interests reported.

SupplementaryInformation.docx

Download PDF

Journal Publication

published 03 Nov, 2023

Read the published version in Scientific Reports →

Editorial decision: Major revision
21 Aug, 2023
Reviews received at journal
06 Aug, 2023
Reviewers agreed at journal
26 Jul, 2023
Reviewers agreed at journal
24 Jul, 2023
Reviewers invited by journal
24 Jul, 2023
Editor assigned by journal
24 Jul, 2023
Editor invited by journal
01 Jun, 2023
Submission checks completed at journal
01 Jun, 2023
First submitted to journal
26 May, 2023

You are reading this latest preprint version

Will AI Art Devalue Human Creativity?

Status:

Journal Publication

Version 1

Abstract

Figures