A new online experimental study investigated how individuals perceive AI- and human-created art created to evoke emotions. They found that people attribute intentions and emotions to the author of a work, regardless of whether it was created by artificial intelligence or a human. However, they felt stronger emotions when they saw art created by humans. This research Computers in human behavior.
In recent years, the practical application of AI has made great progress. From AI chatbots to generative trans AI to AI that can perform tasks once believed only humans could do, AI technologies are reshaping our society. One of these areas is art.
Back in 2018, auction house Christie’s sold the painting for $432,500. This wouldn’t be unusual if it weren’t for the fact that the painting, titled “Portrait of Edmund Bellamy,” was created by an AI algorithm. Many people were shocked by this development. However, in the five years since this event, AI algorithms that generate artwork have become fairly common.
Study author Teresa Rahel Demmer and her colleagues wanted to investigate people’s emotional responses to computer-generated (or AI-generated) and human-generated art. Previous research suggests that humans may be good at conveying emotional experiences through works of art. The study authors wanted to know whether individuals routinely report experiencing specific emotions or sensing emotional intent with AI-generated art. They were also interested in whether individuals could distinguish between AI and human-made art if they were similar.
The study involved 48 participants who joined through social media and the online platform SurveyCircle. Of these, 25 of his were male, with educational backgrounds ranging from secondary education to Ph.D.
The artwork shown to participants consisted of a 32×32 black and white grid. These grids were randomly placed by computers using online noise generators or by human artists, including online visual artists, art students, and degree holders in the field. The human artist was tasked with selecting one or more emotions he wanted to convey through his artwork, and then reporting it.
Participants were shown these works and were informed in advance whether each work was created by a computer or by a human. It was also explained that people will be viewing approximately the same number of works of art created by humans and AI. In the end, they viewed 14 of his works created by humans and 10 of his works created by AI.
Participants expressed their evaluation of all works, rated their aesthetic qualities, and described specific emotions they experienced while viewing them. They also provided insight into the emotions the creator of the artwork was likely trying to convey.
In the first two blocks of the experiment, participants viewed two sets of 12 pieces. Before viewing, I received a message indicating whether the upcoming work was created by a human or a computer. In her third segment of the experiment, he viewed all 24 works without giving any information about the creator.
The results showed that participants correctly identified the creator of the artwork (human or computer) 63.8% of the time. Although not extremely high, this percentage exceeded random guessing. However, this accuracy varied widely across participants. He had one participant who guessed the author correctly 100% of the time, but he had two participants who guessed the author correctly only 25% of the time (much less than chance). did.
There were no noticeable differences in ratings for works labeled human-generated and computer-generated. However, works of art created purely by humans were rated slightly higher on a variety of experiential dimensions, including beauty, interest, quality, comfort, meaningfulness, and clarity. The most notable differences were in ratings of beauty and fun.
Participants reported feeling at least some emotion in 77% of the works they viewed. 78% of participants reported feeling no emotion towards at least one piece of art. They felt emotions most often for human-created artworks identified as such and least for computer-generated works labeled as such. Subsequent analysis revealed that these differences were primarily influenced by the label rather than the actual creator. Although there were no significant differences in the intensity of emotions reported for AI-created art and human-created art, slight differences emerged when rating individual emotions.
In 43% of cases, participants believed that the author did not intend to convey emotion. They most often attributed this lack of intended emotion to works of art created by such identified AIs. Similarly, results revealed a pattern in the number of emotions participants believed creators wanted to convey. However, participants were more likely to attribute emotional intent to AI-created artwork.
“In conclusion, our study shows that participants reported feeling emotions and perceived intentions in works of art, regardless of whether they were created by an artist, a computer, or a human artist. “We found fairly convincing and consistent evidence suggesting that “Interestingly, such reporting using computer-generated images suggests that AI art lacks the ability to evoke human emotional and intentional elements, at least when considered from a real-life perspective and viewer self-report. This contradicts the idea that
“At the same time, this study found that human-made art evoked stronger reactions, that participants often demonstrated an ability to recognize the emotions intended by humans, and that viewers’ evaluations and specific nature of their emotional experiences were Artists, it turns out, are influenced by their provenance. ”
This study makes an important contribution to scientific understanding of how humans perceive artwork generated by AI. However, there are also limitations that need to be considered. Of note, the number of study participants was very small and the complexity of the artwork used was very limited and restricted to similar ones. If the sample were larger and works created with full artistic freedom were compared, the results might not be the same.
the study, “Does an emotional connection to art really require a human artist? Emotional and intentional responses to AI versus human-created art, and their impact on aesthetic experience” was written by Teresa Rahel Demmer, Corinna Künapfel, Jorg Fingerhut, and Matthew Perowski.