AI art fails to trigger the same empathy as human works

For centuries, philosophers and psychologists have argued that art does more than please the eye. It serves as a bridge between minds, allowing viewers to step into the experiences of others and develop a shared sense of humanity. A new series of experiments suggests that this bridge may be broken when the artist is a machine.

Researchers found that when people believe a work of art was created by artificial intelligence, they feel less awe. This reduced emotional response leads to a decrease in empathy for the subjects depicted in the work. The findings were published in the Journal of Experimental Social Psychology.

The study explores a psychological chain reaction that begins with the creator’s identity. Art is traditionally viewed as a deeply human act of expression. When we engage with a painting or a poem, we are not just processing visual or linguistic information. We are often attempting to understand the intent and perspective of another person. This process can trigger a sense of awe. Awe is an emotion we feel when we encounter something vast that challenges our current understanding of the world. Psychological theory suggests that awe diminishes our focus on the self and encourages us to feel connected to others.

Artificial intelligence has rapidly entered the creative sphere. Algorithms can now generate paintings, poetry, and music that mimic human styles with high fidelity. Michael W. White, a researcher at Columbia Business School, and his colleague Rebecca Ponce de Leon sought to understand if these AI-generated works function the same way human art does. They wanted to know if the knowledge of an artwork’s origin changes the emotional payoff for the viewer. They hypothesized that without a human mind behind the curtain, the sense of awe would evaporate. Without awe, the subsequent feelings of empathy might fail to materialize.

To test this, the researchers conducted five separate experiments involving over 1,500 participants. The first study took place in the real world rather than a laboratory. Research assistants recruited patrons at two major art museums in a large Northeastern city. These patrons viewed paintings depicting human suffering, such as miners, garment workers, or survivors of natural disasters.

The researchers used a deceptive experimental design to isolate the effect of the label. All the images shown were actually generated by AI. However, half the participants were told the art was created by a human artist named Jamie Kendricks. The other half were told the art was created by an artificial intelligence program. Participants then rated their empathy for the suffering people depicted in the images. The results showed a clear divide. Patrons who believed they were looking at AI art reported lower levels of empathy than those who thought they were viewing human art.

The second study aimed to ensure that the quality of the art was not the deciding factor. This time, the researchers used paintings actually created by human artists. They again manipulated the labels. Some participants were told the human-made art was the work of AI. The pattern held firm. Even when looking at human-created work, the mere belief that it came from a machine reduced the empathy participants felt for the subjects. This confirmed that the bias stems from the viewer’s beliefs about the creator, not the aesthetic properties of the image itself.

In the third study, the team expanded their scope to literary art. Participants read poems about love, nature, or family. The researchers also introduced a specific measure for awe. They asked participants how much wonder or amazement they felt. The data revealed that people experienced less awe when they attributed the poetry to a computer program. Statistical analysis showed that this lack of awe was responsible for the drop in empathy.

The fourth study moved back to a field setting to see if these feelings influenced behavior. The researchers set up a station in the lobby of a large office building. Passersby viewed a painting of disaster survivors. Afterward, they were given the opportunity to donate part of their compensation to charity. Participants who believed the painting was AI-generated reported less awe and empathy. Consequently, they were less likely to donate any money compared to those who believed a human painted the image.

The final study dug deeper into why AI art fails to elicit awe. The researchers measured two specific components of awe: perceived vastness and the need for accommodation. Vastness refers to the sense that something is larger than the self or ordinary experience. Need for accommodation is the feeling that a new experience challenges one’s existing mental structures. Participants viewed a painting of tsunami survivors. Those who thought it was AI-generated rated the work as less vast. They also felt less need to mentally accommodate the work. This lack of cognitive challenge stifled the experience of awe, which in turn suppressed empathy.

These findings align with a growing body of evidence regarding human reactions to AI creativity. A separate meta-analysis authored by Alwin de Rooij and published in Psychology of Aesthetics, Creativity, and the Arts examined nearly 200 effect sizes from various studies. De Rooij found that knowing an image is AI-generated negatively impacts how people process the work. This bias affects deep interpretation and even changes how viewers perceive basic visual features like color and brightness.

Similarly, a study authored by Kobe Millet and colleagues in Computers in Human Behavior found that people perceive AI art as less creative. Millet’s team identified “anthropocentric creativity beliefs” as a driving factor. This is the conviction that creativity is a uniquely human trait. People who hold this belief strongly are more likely to downgrade their appreciation of AI art. They experience less awe when viewing it. White and Ponce de Leon’s work builds on this by showing that the deficit in awe has social consequences. It stops the art from functioning as a tool for moral and emotional connection.

There are limitations to the current research. The studies primarily used art depicting suffering or serious subjects to measure empathy. It is unclear if the same blunting effect would apply to art meant to evoke joy or whimsy. Additionally, attitudes toward AI are shifting rapidly. As younger generations grow up with generative tools, they may not harbor the same biases against machine creation. Their capacity for awe in the face of algorithmic output might differ from the current norm.

Future research could investigate whether different types of art, such as music or film, suffer the same penalty. It could also examine if collaborative works, labeled as human-AI partnerships, manage to preserve the emotional impact. For now, the data suggests a hidden cost to the automation of creativity. We may gain efficiency in generating images, but we risk losing the profound connection that comes from witnessing another human’s expression.

The study, “Less “awe”-some art: How AI diminishes the empathic power of the arts,” was authored by Michael W. White and Rebecca Ponce de Leon.