Analyzing the Impact of AI-Driven Chatbots as Virtual English Tutors on English Language Learning and Engagement

The following study aims to assess the effect of deploying LSTM-based chatbots in learning English and learners' engagement level. Thus, knowing how useful conversational AI is as a virtual tutor is useful during the advancement of education. The Embedded Self-Regulated Learning Framework was based on the LSTM structure of an AI-based chatbot that was used to engage with the student in natural language and assist the student in language exercises in real-time while helping the student navigate learning paths that had been constructed to specifically address the student's needs. A total of 176 junior college students from the University of Alicante Spain, and Silesian University of Technology, Poland participated in the study with B2-C1 language proficiency level of the CEFR and both native and non-native English users were included in the study. Data was collected from February to May, during the Spring term of the 2022 academic year and using two, two hour sessions per week whereby the participants devoted one month of their curriculum to this study. To determine the improvement on the language of the students, tests on the language Proficiency were given both before and after the intervention focusing on the changes on their use of vocabulary, grammar and fluency of the conversation. Quantitative feedback was obtained by use of quantitative questionnaires and tracking of the amount of time users invested in the chatbot and perception of learning. Altogether the study reveals that use of LSTM-based chatbots improves language learning outcomes especially in terms of vocabulary ability and grammatical skills. The participants also expressed higher involvement in terms of participating more in the interactions involving the chatbot and compliments for responsiveness and individuality of the chatbot. Nevertheless, it should be mentioned that some obstacles including the provision of the complex language and emotional context were mentioned. Based on the findings, this study suggests that LSTM-based chatbots are very useful in language learning and still, more enhancements could still be made to mitigate these drawbacks and harness the chatbots' functionality for learning enhancement to the maximum.