• Alaviyya Bahruz Nuri

  Author

The study is driven by the observation that many computer-assisted language learning (CALL) resources lack consideration of the affective elements of language learning and that many learners are regarded as rational information processors. This paper explores the theoretical principles, design principles, and pedagogic promise of emotion-aware artificial intelligence (AI) systems to recognize, understand, and react to the emotions of learners in real-time to maintain their engagement in English as a Second/Foreign language (ESL/EFL) learning. The paper draws from the literature on affective computing, control-value theory and second language acquisition studies to integrate the latest findings from the field with existing review studies on multimodal emotion recognition, AI-based emotional intelligence, and language learning with chatbots. A conceptual emotion-aware artificial intelligence framework is presented, consisting of the multimodal sensing layer (facial expression, speech prosody, text or keystroke behavior), the emotion recognition module, the learner emotion and learner engagement model, and the adaptive pedagogical response engine that dynamically adapts the difficulty, pacing, and feedback tone of a task based on the inferred affective state. The paper also addresses how foreign language anxiety can be addressed in such systems, how they can keep learners in the optimal zone of enjoyment and flow, and how they can scaffold for learners with varying emotional profiles. The reviewed literature suggests that emotion-informed AI interventions can lead to a decrease in speaking anxiety and an increase in self-reported engagement and learning outcomes compared to emotion-uninformed interventions, with effect sizes differing depending on the specific type and mix of emotional and cognitive support provided. Finally, the paper outlines several challenges for the implementation, such as issues of data privacy and differences in emotion expression across cultures, algorithmic bias, and the potential for relying too heavily on automated emotion detection, and provides directions for future empirical research on emotion-aware AI in real English language learning settings.

• PDF