Students are often unable to access instructors outside of class, due to work and family obligations, packed class schedules or shyness. Electronic channels (e-mail and discussion fora) can partly alleviate this problem, but these interactions are usually fragmented and can significantly delay learning. Consequently, students often resort to looking for answers on the internet. While such active learning efforts can positively impact student growth, online searches are also overtly sensitive to an overabundance of irrelevant, incomplete or misleading information.
This project explored the combination of existing artificial intelligence (AI) tools to create a “Teaching chatbot,” or T-bot, that can answer student questions effectively and within the appropriate context of a class.
Current chatbots effectively handle queries about dates and times, cost, availability, location and specific information about a product or service. In contrast, and quite understandably, when students are learning new material they will often come up with ill-formulated questions that include conceptual mistakes. The long term goal of this project was to combine natural language processing and understanding, deep learning and smart searches to allow the T-bot to engage the student in a dialogue designed to narrow down the true scope and intent of their question and then be able to provide them with a more useful answer within the context of the course. The funding through this grant allowed the project team to explore the feasibility, power and limitations of integrating existing AI APIs as a first step towards the long-term goal.
Principal Investigator
Mariano A. Loza-Coll, Cal State Northridge
Grant Cycle
Spring 2018
