Summary: On platforms like TikTok, AI-generated characters are increasingly being used to give knowledge content, which raises questions about how viewer trust is impacted by their appearance. A new study examined whether nearly 500 individuals ‘ perceptions of honesty were influenced by avatar authenticity and sex.
Amazingly, more realistic characters were rated as more trustworthy, kind, and capable than cartoon-like types, defying the expected “uncanny river” effect. Total trustworthiness was not significantly impacted by gender, despite the perception that male avatars were somewhat more competent.
Important Information
- Authenticity triumphs: Practical Artificial avatars were viewed as more reliable than cartoon-like ones.
- Gender Variations: Man avatars were perceived as more competent, but no variations were observed in their sense of goodwill or dignity.
- No Uncanny Valley: The review found no evidence that people felt uncomfortable when they saw very realistic images.
Origin: SISSA
If you’re one of the 1.5 billion TikTok users in the world, you might have encountered excellent “testimonials” like Marie Curie and Nikola Tesla who have delivered brief, science-related messages that have received millions of views.
One of the many instances of AI-generated avatars being used to connect knowledge is used, which may also have some drawbacks.
The production of images and animations using artificial intelligence is a rapidly expanding field with quality improvements regularly.
However many avatars, while realistic, still exhibit minor flaws like glitches, delays, inconsistent facial expressions, or lip-syncing that are occasionally scarcely discernible but still easily noticed by a human observer.
The additional authors of the study, including Jasmin Baake, a researcher at the Bochum, Germany-based Center for Advanced Internet Studies, realized that these characters may induce a phenomenon known as the “uncanny valley.”
The amazing valley describes how people react to humanoid avatars, whether robotic or electronic, can cause severe discomfort, while more stylized or comical humanoid figures can’t.
Baake and colleagues wondered how much the amazing valley’s ability to elicit outright rejection from viewers stems from how much the admiring traits of AI-avatars representing technology communicators affect the viewer’s perception of their worth.
We wanted to study how these images ‘ degree of realism and sex may affect the recipient’s perception of their worth, says Baake.  ,
A series of movies featuring AI-generated avatars that depicted science speakers, both male and female, were the subjects of the study ( which was conducted in Germany, in German ). The experimental conditions were four, with variations in realism and avatar ( very high vs.  , cartoonish  , style ), gender  , ( male or female ).
The almost 500 individuals were chosen based on a cross-section of a select group of people in Germany in terms of age, sex, and training.
We assumed that the more artistic avatars, and the ones that looked a little more cartoonish, may be perceived as more trustworthy, according to Baake,” for the degree of realism.”
We hypothesized that female avatars had become perceived as more capable, and therefore more reliable nevertheless, than female avatars in light of existing literature on sexist perceptions of science communicators, which show that female scientists are frequently associated with greater competence.
But, Baake and his coworkers were surprised to find that practical characters were more popular than cartoon-style ones in their tests. In particular, the questionnaires that were distributed to members after watching the videos assessed the characters ‘ perceived ability, integrity, and benevolence, which all combined reflect their perception of trustworthiness.
The more accurate characters received substantially higher scores across all three aspects. In terms of gender, there was a partial impact: female avatars were perceived as more capable, but no discernible differences were observed between their moral standards or benevolence.
With our observations, at least in our circumstances, we were unable to find a origin into the amazing river with a higher degree of realism.
Additionally, it was discovered that individual components like people ‘ prior knowledge of AI and trust in science moderated trustworthiness opinions. More reasonable, human-like avatars, in the study’s opinion, appear to be appropriate for communicating medical content.
Baake emphasizes that while there is no amazing valley effect, future studies should look into whether there may be an intermediate amazing valley effect between the two experimental conditions, and whether or not people react to the observer’s perceptions of the realism of avatars different.
About this information about AI research
Federica Sgorbissa as the artist
Source: SISSA
Contact: Federica Sgorbissa – SISSA
Image: The image is credited to Neuroscience News
Original Research: The results may be published in the Journal of Science Communications.
