Humanoid interfaces in artificial intelligence-based language learning devices: Possible 'Uncanny Valley' effects?

The integration of artificial intelligence (AI) into educational tools is transforming learning environments by enabling personalized experiences. This study explores the effectiveness of AI-based educational interfaces, comparing humanoid avatar tutors with audio-based tutors in language learning contexts. Using a sample of 200 participants (55 % male; mean age = 25.01, SD = 4.20), the research employed a 2 × 2 between-subjects design to examine the impact of tutor type (avatar vs. audio) and tutor gender (male vs. female) on user perceptions and acceptance. Findings revealed that audio-based tutors were rated more positively in dimensions such as anthropomorphism, animation, and intention to use. Gender congruence between participants and tutors also influenced perceptions, with female participants rating female audio tutors as safer and more anthropomorphic. Regression analyses highlighted that openness to experience and personal innovativeness in technology were significant predictors of positive evaluations, particularly regarding tutor sympathy and perceived usefulness. These results underscore the challenges associated with humanoid avatars, including the "Uncanny Valley" effect, which may reduce user acceptance and engagement. The study suggests that audio-based AI interfaces may be more effective in promoting positive user experiences, emphasizing the need for tailored design strategies in AI-driven educational tools. Future research should explore long-term impacts and cross-cultural differences.