2686 - LONGITUDINAL ANALYSIS OF THE IMPACT OF ANTHROPOMORPHIC AI PERCEPTION ON USER ATTITUDES AND SELF-CONGRUENCE

Session: D03S003 - Development in Digital Contexts 3
AUTHORS:
Kolanska-Stronka Magdalena (SWPS University ~ Warsaw ~ Poland) , Mamcarz Piotr (The John Paul II Catholic University of Lublin ~ Lublin ~ Poland)
Abstract text:
The integration of AI in education underscores the need to understand the psychological mechanisms governing student-AI interactions. A central construct is self-congruence—the alignment between an individual's self-concept and their perception of a system—which promotes satisfaction and engagement. Research suggests that anthropomorphic AI systems enhance this alignment (Waytz et al., 2014). Drawing on the Technology Acceptance Model (TAM; Davis, 1989), this study investigates how attitudes toward technology, shaped by perceived usefulness and ease of use, influence self-congruence.
A longitudinal study was conducted among 185 university students (mean age = 20.73, SD = 3.09) with assessments in February 2025 (T1) and May 2025 (T2). During this period, students reported a significant increase in weekly AI usage, from 2.55 hours (SD = 3.54) to 4.24 hours (SD = 6.64). Participants completed the AISE Anthropomorphic Interaction scale, the AI Attitude Scale (AIAS-4), and the Self-Congruence Scale. All measures demonstrated high internal consistency (Cronbach's α between 0.81 and 0.90).
We hypothesized that: (H1) higher anthropomorphic perception at T1 predicts greater self-congruence at T1; (H2) AI attitudes at T1 predict self-congruence at T2; and (H3) AI attitudes mediate the longitudinal effect of anthropomorphic perception on self-congruence.
Path analysis revealed strong stability in AI attitudes (β = 0.499) and anthropomorphic perception (β = 0.440). Anthropomorphic perception at T1 had a significant direct effect on self-congruence at T1 (β = 0.289) and influenced T2 self-congruence indirectly via AI attitudes (total indirect effect β = 0.292). The model explained substantial variance in T2 self-congruence (R² = 0.362) with good fit indices (SRMR = 0.067, NFI = 0.935).
These findings highlight self-congruence as a vital component of student-AI interactions. In educational contexts, designing AI tools with anthropomorphic features may enhance students' motivation and learning outcomes by fostering positive attitudes and alignment with their self-concept.