Perceived Algorithmic Bias, Perceived AI Surveillance, Linguistic Identity and Willingness to Communicate in AI ‐Mediated Education

ABSTRACT

The rapid integration of artificial intelligence (AI) into language learning environments has raised important questions about equity, identity and learner psychology in digital contexts. This study examined the relationships among perceived algorithmic bias, linguistic identity, perceived AI surveillance and willingness to communicate (WTC) among 679 Chinese EFL learners in AI‐mediated language learning environments. The data were collected through a self‐report questionnaire and analyzed using structural equation modelling (SEM) with SPSS version 27 and AMOS version 24. The findings revealed significant correlations among all four constructs. Perceived algorithmic bias was negatively associated with WTC confidence (β = −0.48, p  < 0.001) and positively associated with WTC avoidance (β = 0.55, p  < 0.001). Linguistic identity pressure emerged as the strongest predictor of both WTC dimensions (confidence: β = −0.31, p  < 0.001; avoidance: β = 0.35, p  < 0.001), while linguistic identity pride showed the opposite pattern. Perceived AI surveillance, including both monitoring and stress components, also significantly predicted WTC. The predictor variables collectively explained 48% of the variance in WTC confidence and 52% of the variance in WTC avoidance. The results highlight the important roles that algorithmic bias, linguistic identity and surveillance play in shaping learners' communicative behaviour in AI‐mediated environments. The findings suggest that educators and technology developers should work to reduce algorithmic bias, support learners' linguistic identities, and create learning environments that minimize surveillance‐related stress to enhance WTC.