Do Realistic AI Avatars Invoke the Uncanny Valley?

Seeing human-like robots, dolls, and AI-generated faces can trigger eerie, put-off feelings towards the figure, a phenomenon known as the uncanny valley. Though these figures look almost human, there is just something slightly off about them. It could be the unblinking eyes, unnatural stillness, or — for AI-generated figures — poorly synced facial and lip movements.

Recently, researchers from the Center for Advanced Internet Studies (CAIS) in Bochum, Germany, have noticed AI-generated avatars of notable scientists, like Marie Curie and Nikola Tesla, circulating on social media platforms like TikTok. In these videos, the AI replica scientists would explain who they were and what sort of experiments they worked on. This had researchers from CAIS wondering if these AI avatars triggered the uncanny valley in viewers.

AI and the Uncanny Valley

For this study, published in the Journal of Science Communication, Jasmin Baake, one of the study authors, and colleagues wondered if receiving scientific information from these AI avatars would not only invoke the uncanny valley but also make people less likely to trust the science information that was being communicated.

“We wanted to do research on the perception of these avatars and especially on how their degree of realism and their gender might impact the trustworthiness perception of the recipient,” said Baake in a press release.

Surveying the Uncanny Valley

The study, which took place in Germany, surveyed about 500 people of different ages, genders, and educational backgrounds. The study participants were shown four different videos of AI-generated science communicators. Two of the videos depicted either a realistic male or a female speaking, while the other two videos depicted a cartoonish male and female speaking.

“For the degree of realism, we assumed that with the uncanny valley hypothesis in mind, the more stylistic avatars, so the ones that looked a bit more cartoonish, would be perceived as more trustworthy," said Baake in a press release.

“And, based on existing literature on gendered perceptions of science communicators — which show that male scientists are often associated with greater competence — combined with concerns that AI-generated avatars may reflect and reinforce such stereotypes due to biased training data, we hypothesized that male avatars would be perceived as more competent, and therefore more trustworthy overall, than female avatars,” Baake added in the release.

However, the study results shocked the researchers.

Perceiving AI Avatars

After watching the videos, the research team gave the study participants a questionnaire. The participants rated the avatars based on integrity, perceived competence, and benevolence of the avatars — which, according to the study, can reflect perceived trustworthiness.

The results of the questionnaire favored realistic avatars over cartoonish ones and saw the male avatars as more trustworthy versus the female avatars.

More surprising still is that none of the study participants seemed to cross into the uncanny valley.

“With our findings, we could not find a descent into the uncanny valley with a higher degree of realism, at least in our conditions,” said Baake in a press release.

According to the study, while realistic AI avatars may be suitable for communicating scientific information, factors such as a participant's prior knowledge of AI or trust in science may have impacted results. Overall, the study authors agree that more research is needed to determine how people will continue to perceive AI avatars.

Article Sources

Our writers at Discovermagazine.com use peer-reviewed studies and high-quality sources for our articles, and our editors review for scientific accuracy and editorial standards. Review the sources used below for this article: