We are excited to announce that our extended abstract on the integration of an AI-based Teaching Assistant BOT in academic courses has been accepted by the International Conference on Multidisciplinary Research-2024 of the Eastern University of Sri Lanka. This is a great achievement for our team and we would like to share some highlights of our work with you.
The motivation behind our project was to address the challenges of maintaining educational quality in the face of expanding student populations. As more and more students enroll in online courses, it becomes difficult to provide personalized support and feedback to each learner. Traditional human resources, such as teaching assistants, are often insufficient or unavailable to meet the demand. This can affect student motivation, engagement and learning outcomes.
To overcome this problem, we developed an AI-based Teaching Assistant BOT that can offer real-time, contextually relevant assistance to large student groups, minimizing the reliance on human resources. Unlike existing AI tools that provide generic or scripted responses, our BOT leverages a Large Language Model (LLM) that is tailored to the specific course content, ensuring accurate and personalized responses. The BOT can handle a variety of queries related to the course material, such as definitions, examples, explanations, clarifications, references and more. The BOT can also provide feedback on assignments, quizzes and exams, as well as suggestions for improvement.
To achieve this level of functionality, we used advanced AI techniques such as 'gpt-3.5-turbo' and the Retrieval Augmented Generation (RAG) methodology. 'gpt-3.5-turbo' is a state-of-the-art LLM that can generate natural language texts based on a given prompt or context. RAG is a technique that allows the LLM to retrieve relevant information from a large corpus of documents, such as textbooks, lecture notes, research papers and more. By combining these two techniques, we were able to create a BOT that can generate precise and relevant responses based on the course content.
We conducted preliminary tests with a sample of students who enrolled in one of our online courses. The results were very encouraging. The students reported that the BOT was helpful, informative and easy to use. They also appreciated the BOT's ability to provide personalized and timely feedback. The students rated the BOT's performance as high or very high on various criteria such as accuracy, relevance, clarity, completeness and friendliness. The students also expressed their interest in using the BOT for other courses in the future.
We believe that our AI-based Teaching Assistant BOT offers a promising solution to sustain educational quality amid resource constraints. It can enhance the educational experience for both students and instructors by providing effective and efficient assistance. It can also reduce the workload and stress for human teaching assistants and allow them to focus on more creative and complex tasks.
Our BOT is accessible for experimentation at http://chat.ucsc.cmb.ac.lk and we welcome feedback from other educators and researchers who are interested in using or improving it.
We would like to thank our collaborators, sponsors and supporters for their valuable contributions to this project.
You can access the full publication trough following link
Comments