The (in)congruence of speaking self-efficacy and anxiety in shaping L2 learners’ willingness to communicate within GenAI-enhanced learning environments: a response surface analysis

Purpose This study examined how speaking self-efficacy and speaking anxiety relate to second or foreign language (L2) learners’ willingness to communicate (WTC) in a GenAI-enhanced learning environment. More specifically, the study focused on the joint associations of these psycho-emotional variables with WTC and examined whether their alignment or discrepancy was related to learners’ communication willingness.Methodology A sample of 280 L2 learners completed measures of speaking self-efficacy, speaking anxiety, and WTC within a generative artificial intelligence (GenAI)-enhanced learning environment. Bivariate correlations were first examined to establish the associations among the variables. Response surface analysis (RSA) was then employed to examine the joint associations of speaking self-efficacy and speaking anxiety with WTC.Findings Speaking self-efficacy was positively associated with WTC, whereas speaking anxiety was negatively associated with WTC. The significant interaction between speaking self-efficacy and speaking anxiety indicated that their associations with WTC varied as a function of each other. The RSA showed that the direction of incongruence between speaking self-efficacy and speaking anxiety was systematically associated with WTC, with WTC higher when the discrepancy favored greater self-efficacy and lower anxiety. At the configuration level, the incongruent combination of high speaking self-efficacy and low speaking anxiety was associated with the highest WTC.Originality/value The study extends research on L2 learners’ WTC by moving beyond separate examinations of speaking self-efficacy and speaking anxiety and considering how these experiences operate in combination within a GenAI-enhanced instructional context.