Study Links Perceived Non-Humanity and Structural Similarity to Deep Self-Disclosure Toward Generative AI

A recent study on arXiv examines deep self-disclosure towards generative AI, emphasizing psychological elements like perceived non-humanity and structural similarity, which extend beyond anthropomorphism. Researchers analyzed data from a cross-sectional survey involving 2,400 participants collected in 2025, exploring the links between self-disclosure occurrence and depth. The logistic regression indicated that participants in the high-perception group (Segment D) were significantly more likely to disclose than those in the baseline group (Segment A; OR = 11.35). Additionally, ANOVA results highlighted notable differences in the depth of disclosure among groups. These findings imply that factors influencing trust in deep self-disclosure may not solely rely on anthropomorphic perceptions. The study is exploratory and was conducted in 2025.