Language Learning in the World of Fluent Machines: New Language Innovators Podcast Episode
We started the new season of the Language Innovators podcast with this episode featuring Dr. Soroush Sabbaghan from the University of Calgary. The episode centers on integrating Generative AI with Computer Assisted Language Learning (CALL) principles to enhance second language acquisition. The speakers agree that while AI can provide efficient information and targeted practice, it often lacks the paralinguistic features and communicative pressure found in human interaction. The conversation then shifts toward leveraging AI's unique capabilities such as “infinite patience” and feedback to foster metacognitive awareness and learner agency.
About Dr. Soroush Sabbaghan: Dr. Sabbaghan is an Associate Professor at the University of Calgary’s Werklund School of Education and the inaugural GenAI Educational Leader in Residence at the Taylor Institute for Teaching and Learning. He researches, teaches and speaks on generative AI in education.
Soroush also has a new book out on ‘Navigating Generative AI in Higher Education,’ available on his website.
CALL Evolution and Core Principles
Evolution of CALL: Progressed from isolated language labs to connected computer labs, then to mobile technology for accessibility, and finally to Generative AI.
Central Tension of AI: AI can either support or undermine learning; it may remove the cognitive friction necessary for actual learning if it performs the work for the student.
Essential Learning Principles: Meaningful language development requires meaningful exposure, active production, interaction and repair, and the development of confidence through social interaction.
AI's Current Limitations: AI can be effective at helping students at the syntax level (sentence formulation) but cannot develop competence at the discourse level, where paralinguistic cues (e.g. facial expressions) are critical for negotiating meaning.
Ideas for Using and Programming AI Chatbots for Learning
Targeted Practice: AI is highly effective for rehearsing specific scenarios, such as job interviews or practicing answering questions for exams like IELTS.
Graduated Support: The need for graduated hints, where support decreases as the learner's capability increases, to maintain a balance between assistance and challenge.
Schema-Based Feedback: Instead of providing direct translations (which is unhelpful), AI should provide schema support via related words or images to help learners retrieve the target word.
Metacognitive Reflection: Using recorded AI interactions to let learners analyze their own performance and hold themselves accountable to their own goals.
Learner Agency and Motivation
Accountability: Shifting agency to the learner by involving them in goal setting and using the system to remind them of their commitments.
Intrinsic Motivation: Capitalizing on motivation by asking "Why are you here?" at the start to align the AI's path with the learner's specific values (e.g., employment, travel, or culture).
Competence vs. Performance: Motivating learners by highlighting can-do statements and making the gradual decrease in AI support visible as a sign of growth.
Future Research and Development
Empirical Approach: The need for research to determine what learners actually pay attention to in AI environments to improve interventions.
Dynamic Feedback Triggers: Investigating how a system can deterministically decide when to focus on forms, form, and meaning based on the interaction's progress.
The podcast is hosted by Dr. Linh Phung and Nikolas Wolfe from Eduling. Subscribe to the podcast on YouTube to receive notifications of new episodes.