LAUSR

Abstract Conversational agents are systems capable of processing and responding to human language. They have evolved over the years from a means to pass the Turing Test to chatbots that fulfill a utilitarian purpose. Closed-domain chatbots are specialized in a specific knowledge base and are often used in an attempt to assist users in an educational context. Existing open-source, educational assistant chatbots are narrow in immediate functionality, thus limiting the content and services that can be provided to students and educators. To address this limitation, a novel multiagent framework is proposed, providing diverse capabilities and component flexibility to better meet varied educational requirements. The version presented in this experiment can not only answer questions in different styles but is also able to provide content summaries and links. The solution is tested by presenting participants with a lesson containing information related to coronavirus disease 2019 (COVID-19), followed by engagement with the chatbot system, a subsequent evaluation of its responses, and a quiz to quantify its pedagogical efficacy. COVID-19 was chosen as the knowledge base due to its current relevance in society. The chatbot framework's knowledge base is comprised of two data sets containing facts related to the virus, one which is used to provide longer, frequently asked questions-type responses, and another used to provide short answers. The resulting participant evaluations indicate a preference for more informative responses in the experimental context and showcase the benefit of the framework's malleability in not only fulfilling but discovering varying needs in educational assistance.