A Study of the Human Perception of Synthetic Faces

Shen, Bingyu; RichardWebster, Brandon; O’Toole, Alice J.; Bowyer, Kevin W.; Scheirer, Walter J.

doi:10.1109/fg52635.2021.9667066

Cited by 23 publications

(8 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Tucciarelli et al (2022) found a partial negative contribution of attractiveness, which aligns with our predictions based on face-space, because faces at the core of face-space (more average faces) tend to be more attractive, all else being equal (Rhodes, 2006). Shen et al (2021) also found that removing background scenery made AI and human faces indistinguishable; however, background information was matched for our stimuli, rendering this latter explanation unlikely here.…”

Section: Methodssupporting

confidence: 85%

“…1) and therefore, potentially, especially realistic. Consistent with this theory, Shen et al (2021) and Tucciarelli et al (2022) found preliminary evidence of AI hyperrealism to the extent they tested White faces, although Tucciarelli et al's (2022) stimuli were also preselected to be particularly realistic, biasing them toward this finding. Intriguingly, Nightingale and Farid (2022) also reported more errors for White than non-White AI face detection.…”

mentioning

confidence: 73%

See 1 more Smart Citation

AI Hyperrealism: Why AI Faces Are Perceived as More Real Than Human Ones

Miller,

Steward,

Witkower

et al. 2023

Psychol Sci

View full text Add to dashboard Cite

Recent evidence shows that AI-generated faces are now indistinguishable from human faces. However, algorithms are trained disproportionately on White faces, and thus White AI faces may appear especially realistic. In Experiment 1 ( N = 124 adults), alongside our reanalysis of previously published data, we showed that White AI faces are judged as human more often than actual human faces—a phenomenon we term AI hyperrealism. Paradoxically, people who made the most errors in this task were the most confident (a Dunning-Kruger effect). In Experiment 2 ( N = 610 adults), we used face-space theory and participant qualitative reports to identify key facial attributes that distinguish AI from human faces but were misinterpreted by participants, leading to AI hyperrealism. However, the attributes permitted high accuracy using machine learning. These findings illustrate how psychological theory can inform understanding of AI outputs and provide direction for debiasing AI algorithms, thereby promoting the ethical use of AI.

show abstract

Section: Methodssupporting

confidence: 85%

mentioning

confidence: 73%

AI Hyperrealism: Why AI Faces Are Perceived as More Real Than Human Ones

Miller,

Steward,

Witkower

et al. 2023

Psychol Sci

View full text Add to dashboard Cite

show abstract

“…These synthetic faces are perceived as real human face images under many circumstances (B. Shen et al, 2021). The use of GANs and cycle-GANs has started to make its way into face perception research (e.g.…”

Section: Synthetic Facesmentioning

confidence: 99%

Reproducible Methods for Face Research

DeBruine¹,

Holzleitner²,

Tiddeman³

et al. 2022

Preprint

View full text Add to dashboard Cite

Face stimuli are commonly created in ways that are not explained well enough for others to reproduce them. In this paper, we document the irreproducibility of most face stimuli, explain the benefits of reproducible stimuli, and introduce the open-source R package webmorphR that facilitates scriptable face image processing. We explain the technical processes of morphing and transforming through a case study of creating face stimuli from an open-access image set. Finally, we discuss some ethical and methodological issues around the use of face images in research that may be ameliorated through the use of reproducible stimuli.

show abstract

“…This prompts the question—is AI‐generated content ecologically valid? Behavioural research confirms that it is difficult for humans to detect high‐quality AI‐generated images, suggesting they accurately portray real people, places and things (e.g., Bray et al, 2022; Korshunov & Marcel, 2020; Lu et al, 2023; Shen et al, 2021). The ultimate test might be whether real and AI‐generated faces elicit similar neural responses, given human expertise in face perception (Haxby et al, 2000).…”

Section: Introductionmentioning

confidence: 99%

Embracing deepfakes and AI‐generated images in neuroscience research

Becker

Laycock

2023

Eur J of Neuroscience

View full text Add to dashboard Cite

The rise of deepfakes and AI‐generated images has raised concerns regarding their potential misuse. However, this commentary highlights the valuable opportunities these technologies offer for neuroscience research. Deepfakes deliver accessible, realistic and customisable dynamic face stimuli, while generative adversarial networks (GANs) can generate and modify diverse and high‐quality static content. These advancements can enhance the variability and ecological validity of research methods and enable the creation of previously unattainable stimuli. When AI‐generated images are informed by brain responses, they provide unique insights into the structure and function of visual systems. The authors argue that experimental psychologists and cognitive neuroscientists stay informed about these emerging tools and embrace their potential to advance the field of visual neuroscience.

show abstract

A Study of the Human Perception of Synthetic Faces

Cited by 23 publications

References 28 publications

AI Hyperrealism: Why AI Faces Are Perceived as More Real Than Human Ones

AI Hyperrealism: Why AI Faces Are Perceived as More Real Than Human Ones

Reproducible Methods for Face Research

Embracing deepfakes and AI‐generated images in neuroscience research

Contact Info

Product

Resources

About