A Dopamine-Informed Framework for AI Voice Feedback and English Learning Motivation in Low-Achieving Grade 5 EFL Learners in Hsinchu, Taiwan

This study applies a dopamine-informed theoretical framework to examine how AI voice-detection feedback influences English-learning motivation in 10–11-year-old EFL learners. Drawing on reward prediction error theory (Schultz, 1998), incentive salience (Berridge & Robinson, 1998), and Self-Determination Theory (Deci & Ryan, 1985; Ryan & Deci, 2000), it proposes that automatic speech recognition (ASR) feedback, structured around a variable-ratio reward system, may generate reward expectation and motivational gains, particularly among learners with histories of academic failure.
Seventy Grade 5 students (N = 70; M age = 11.03) from three intact classes in a public elementary school in Hsinchu City, Taiwan, participated in a shared 12-week AI-supported English program. Students were classified as Low Achievers (LAG; n = 28) or Higher Achievers (HAG; n = 42). Motivation was measured using the Children’s English Academic Motivation Scale (CEAMS), a 17-item instrument integrating the Academic Motivation Scale for Children (AMS-C) and a child self-report adaptation of the Behavior Rating Inventory of Executive Function (BRIEF).
Results showed that the LAG demonstrated significant gains in intrinsic motivation (d = 0.86, p < .001), executive function self-regulation (d = 0.73, p < .001), and reductions in amotivation (d = 0.81, p < .001). No significant changes were found in the HAG. Within the LAG, socioeconomic resource disadvantage significantly moderated gains in intrinsic motivation (LAG-RD: d = 0.97). Findings suggest that AI-driven reward expectation mechanisms may be most effective where prior success expectations have been suppressed by failure history and resource disadvantage, supporting equity-focused AI deployment in EFL contexts.