Generative-AI Powered TA BOT for Enhanced Personalized Support in Software Engineering Education

EasyChair Preprint 15472

Abstract

The rising demand for higher education has strained resources, particularly in providing instructional support through Teaching Assistants (TAs). Traditional TAs are vital for personalized learning, but their scarcity in high-enrollment courses hampers the 'one-to-one' interaction essential for student success. This research develops and evaluates a Generative AI-based Teaching Assistant BOT (TA BOT) tailored to specific course content, leveraging large language models (LLMs) to enhance personalized interactions within Learning Management Systems (LMSs). By fine-tuning OpenAI's GPT-3.5-Turbo model using domain-specific data through a simplified Retrieval-Augmented Generation approach, we created a TA BOT for a software engineering course. The TA BOT was assessed by comparing it with generic ChatGPT and through student feedback. Results show that the TA BOT delivers more focused and contextually relevant responses, aligning closely with course objectives and enhancing students' understanding of core concepts. User feedback revealed high satisfaction, with 93.1% recognizing the TA BOT as a valuable educational tool. This study demonstrates the effectiveness of fine-tuning LLMs to create domain-specific educational chatbots that improve personalized learning experiences. Future work includes enhancing the user interface, integrating speech-to-text and text-to-speech, improving backend efficiency, developing a Moodle LMS forum connector for new interaction models, and expanding the TA BOT to other subjects.

Keyphrases: Generative AI, Learning Management Systems, Software Engineering, TA BOT, personalized learning

@booklet{EasyChair:15472,
  author    = {Sudam Rohanadeera and K Priyantha Hewagamage},
  title     = {Generative-AI Powered TA BOT for Enhanced Personalized Support in Software Engineering Education},
  howpublished = {EasyChair Preprint 15472},
  year      = {EasyChair, 2024}}