A recent study published in the journal Cognition and Emotion reveals that people find images of individuals in underwear more sexually arousing when they believe the images are of real people rather than artificially generated. Conducted by researchers from multiple institutions in Italy and Finland, the study sheds light on how our perception of authenticity influences our response to evocative stimuli.
The rapid advancement of technology for generating photorealistic images, particularly through artificial intelligence (AI), has posed significant challenges for contemporary society. One of the primary motivations behind the research was to understand how these AI-generated images, which can now be indistinguishable from real photographs, impact human perception and emotional responses.
These technologies are increasingly used not only in media and entertainment but also in malicious activities such as spreading fake news and creating non-consensual intimate images. The researchers aimed to explore whether the perceived authenticity of an image influences its ability to elicit sexual arousal, thus providing insights into the broader psychological effects of AI-generated content.
“Technological improvement in AI allows the generation of procedural images that are not real, but look very much real,” said study authors Alessandro Demichelis and Alessandro Ansani in a joint statement to PsyPost. “Some of them can be sexually arousing. We wanted to understand how much thinking that an image is real or not real would impact self-reported arousal levels. In particular, we wanted to answer the question: are the images thought to be artificially generated capable of eliciting the same level of arousal as real ones, or do the latter still keep an edge in that regard?”
To investigate this, the research team, which hailed from IMT School for Advanced Studies Lucca, University of Jyväskylä, Institute of Cognitive Sciences and Technologies, and Roma Tre University, conducted two studies to examine the hypothesis that real images are perceived as more arousing than AI-generated ones.
In the first study, the researchers aimed to understand how the perception of authenticity influences sexual arousal when viewing images of individuals in underwear. To achieve this, they conducted an online experiment involving 57 participants, who were recruited through social media and academic mailing lists. Participants were presented with a set of 60 high-quality images of male and female models wearing lingerie or swimsuits. These images were carefully selected to ensure they met specific criteria, such as showing the entire face of the model and having a neutral background.
Participants were asked to complete two tasks in a counterbalanced order. In the first task, they rated their level of sexual arousal for each image on a six-point scale ranging from “not at all” to “very much.” In the second task, they assessed the perceived authenticity of each image, indicating whether they believed the image was a real photograph or generated by artificial intelligence. The aim was to determine if images perceived as real elicited higher arousal than those perceived as AI-generated.
The findings from Study 1 confirmed the researchers’ hypothesis: images judged as more likely to be real were rated as more sexually arousing. This effect was consistent regardless of the order in which the tasks were presented.
Additionally, the study found gender differences in arousal ratings, with male participants reporting higher levels of arousal than female participants. These results suggest that the perception of authenticity significantly influences sexual arousal, supporting the idea that real images are more emotionally impactful than those perceived as fake.
“Our findings support the view that photos believed to be artificially generated are less arousing than those considered real, but we found that allegedly fake images are still capable of generating arousal, especially in men, just in an inferior amount,” Demichelis and Ansani explained.
To further investigate the causal relationship between perceived authenticity and sexual arousal, the researchers conducted a second study with 108 participants. This study used the same set of images as Study 1 but introduced a manipulation to explicitly label half of the images as real photographs and the other half as AI-generated. The participants were again recruited through academic mailing lists, and the experimental procedure was conducted online.
Participants were divided into two groups and presented with two consecutive blocks of 30 images each. One block was labeled as real photographs, while the other was labeled as AI-generated images. The order of the blocks was counterbalanced across participants to control for order effects. For each image, participants rated their level of sexual arousal on a six-point scale, similar to the first study. Additionally, participants’ attitudes toward artificial intelligence were measured to see if these attitudes influenced their arousal ratings.
The results of Study 2 reinforced the findings from the first study. Images labeled as real photographs elicited higher arousal ratings compared to those labeled as AI-generated. This effect was significant and consistent across different presentation orders and genders. The study confirmed that the perception of authenticity directly influences sexual arousal, demonstrating that labeling an image as real or fake can significantly impact emotional responses.
Both studies combined provide strong evidence that perceived realness enhances the sexual arousal elicited by images. This research highlights the importance of authenticity in shaping emotional responses to visual stimuli, offering insights into how AI-generated images might affect human interactions with digital content.
“AI-generated images are here to stay, and as every technological advancement, offer both opportunities and danger,” Demichelis and Ansani said. “Within the domain of sexual arousal, our findings suggest that they are not going to replace the ‘real’ world, since the mere belief that an image is AI-generated (even when it is not) is enough to reduce arousal. To put it differently, it seems that we (still?) have a strong preference for humanness over artificiality, even when such artificiality is just purported.”
But the study, like all research, includes some caveats. Firstly, it focused exclusively on heterosexual participants and images of models in underwear, which may not generalize to other types of sexual stimuli or diverse sexual orientations. Future research should explore whether these findings hold true with more explicit sexual content and with participants of various sexual orientations and gender identities.
Additionally, the study relied on self-reported arousal, which might not fully capture the complexity of sexual responses. Incorporating physiological measures such as heart rate or skin conductance could provide a more comprehensive understanding of the arousal process.
“Due to the novelty of the topic, our research cannot stand on the ground as firmly as we would have liked to,” Demichelis and Ansani noted. “In particular, the copyright and ethical issues and the normative vacuum revolving around the use of AI-generated images prevented us from employing them. All of the images used in our study depict in fact real people; the difference is that, for some of them, the description given was describing them as AI-generated; something very akin to what we call ‘framing effect’ in cognitive psychology.”
“This is both a con and a pro: on the one hand, the stimuli are not truly AI-generated. On the other hand, if the mere fact of presenting an image as AI-generated is enough to produce a decrease in arousal, one could predict that an image which is blatantly AI-generated will produce an even stronger decrease. But this is not something we can take for granted.”
Regarding the long-term goals for this line of research, Demichelis and Ansani said they “would like to replicate the results with a direct comparison between AI-generated and real images. We hypothesize that the effect found in our study would even increase, solidifying the strength of our claims.”
“Second, our second author (Alessandro Ansani) replicated these findings in the domain of music performance (Ansani et al., forthcoming): compared to a human performance condition, people have a strong negative bias toward music performances if they are told that the music is performed by an AI, even when the audio is identical.”
The study, “Real is the new sexy: the influence of perceived realness on self-reported arousal to sexual visual stimuli,” was authored by Marco Marini, Alessandro Ansani, Alessandro Demichelis, Giovanna Mancini, Fabio Paglieri, and Marco Viola.
