Faces Merely Labelled as Artificial are Trusted Less

Abstract

Artificial intelligence plays a crucial role on our daily lives. At the same time, artificial intelligence is often met with reluctance and distrust. Previous research demonstrated that faces that are visibly artificial are considered to be less trustworthy and remembered less accurately compared to natural faces. Current technology, however, enables the generation of artificial faces that are indistinguishable from natural faces. Accordingly, we tested whether natural faces that are merely labelled to be artificial are also trusted less. In three experiments (N = 399), we observed that natural faces merely labeled as being artificial were judged to be less trustworthy. This bias was robust and did not depend on the degree of trustworthiness and attractiveness of the faces, nor could it be modulated by changing raters’ attitude towards artificial intelligence. At the same time, we did not observe differences in recall performance. We conclude that understanding and changing social evaluations towards artificial intelligence goes beyond eliminating physical differences between artificial and natural entities.

Too Beautiful to be Fake: Attractive Faces are Less Likely to be Judged as Artificially Generated

“…Our results found a positive linear trend between trustworthiness and simulation monitoring for females only. Given prior evidence that faces presented as computer-generated were rated less trustworthy (Balas & Pacella, 2017;Hoogers, 2021;Liefooghe et al, 2022), we expected such a linear association to be more clearly present for both genders. One of the underlying mechanisms that possibly contributed to this dimorphism could be the increased risk-taking aversion reported in females (explained evolutionarily as a compromise to their reproductive potential, Van Den Akker et al, 2020), to which perceived facial trustworthiness relates (Hou & Liu, 2019).…”

“…), which suggests that salient and emotional stimuli are perceived to be more real (up to a point of reversal after which beliefs of fiction becomes used an emotion regulation strategy), we hypothesize a quadratic relationship between ATTRACTIVENESS AND REALITY 8 perceived realness and attractiveness: faces rated as highly attractive or unattractive will more likely be believed to be real. We expect a similar relationship with trustworthiness ratings given its well-established link with attractiveness (Bartosik et al, 2021;Garrido & Prada, 2017;Liefooghe et al, 2022;Little et al, 2011), and a positive relationship with familiarity (as more familiar faces would appear as more salient, self-relevant and anchored in reality). Additionally, we will further explore the link shared by dispositional traits, such as personality and attitude towards AI, with simulation monitoring tendencies.…”

“…Interestingly, artificially created faces rated as more attractive (by an independent group of raters) were perceived as less real (Tucciarelli et al, 2020). Conversely, Liefooghe et al (2022) reports that attractiveness ratings were significantly lower when participants who were told that the faces were AI-generated were compared to those who had no prior knowledge. Whereas this line of evidence suggests that reality beliefs have an effect on face attractiveness ratings, the opposite question -whether attractiveness contributes to the formation of reality beliefshas received little attention to date.…”

The Value of AI Guidance in Human Examination of Synthetically-Generated Faces

“…However, mere awareness of the existence of synthetic (GAN-generated) in public domain may erode human trust (Tucciarelli et al 2022). If additionally we show authentic faces to humans along with a wrong classification of being synthetic, such samples may be judged to be less trustworthy than the same images without such wrong label (Liefooghe et al 2022). (Nakano and Yamamoto 2022) showed that humans trust more in self-resembling faces than those judged by a neural network-based face recognizer as more distant.…”

Real is the New Sexy: The Influence of Perceived Realness on Self-Reported Arousal to Sexual Visual Stimuli

“…What about the response to putatively AI-generated people more specifically? So far, the few studies that explicitly deal with the social perception of fake persons demonstrated that faces are trusted less when presented as fake (Liefooghe et al, 2023), or even only suspected to be fake (Tucciarelli et al, 2022, study 3). Interestingly, this is independent of the fact that faces are actually AI-generated (as in Tucciarelli et al, 2022) or genuine photos presented as fake (Liefooghe et al, 2023).…”