A new study published in the Journal of Science Communication (JCOM) reports that highly realistic AI-generated avatars may be more effective in fostering trust during science communication than stylized, cartoon-like avatars. Conducted by Jasmin Baake and colleagues at the Center for Advanced Internet Studies (CAIS) in Bochum, Germany, the research challenges the assumption that hyper-realistic digital avatars risk triggering the “uncanny valley” effect—a phenomenon in which near-human likenesses evoke discomfort or distrust among viewers.
The study involved nearly 500 participants from Germany, selected to represent a cross-section of the population in terms of age, gender, and education. Participants viewed short science communication videos delivered by AI-generated avatars, which varied along two main dimensions: realism (highly realistic versus cartoonish) and gender (male or female). The researchers evaluated participants’ perceptions of the avatars’ trustworthiness using measures of competence, integrity, and benevolence.
Contrary to the researchers’ initial hypothesis, the realistic avatars were rated slightly higher in trustworthiness than their cartoonish counterparts across all three measured dimensions. The anticipated “uncanny valley” response did not appear under the tested conditions. Regarding gender, male avatars were perceived as more competent, but no significant differences were observed between male and female avatars in terms of perceived integrity or benevolence.
The study also found that individual factors, such as participants’ existing knowledge of AI and their level of trust in science, influenced how trustworthy they perceived the avatars to be. Baake notes that while the findings suggest realistic avatars may be suitable for science communication, future research should explore a wider spectrum of avatar realism to determine whether an uncanny valley effect might occur at intermediate levels of visual fidelity.
Photo credit: Jasmin Baake et al
