ABSTRACT
In this chapter we focus on conversation-based learning and assessment environments via the AutoTutor system, in which one or more computerized agents interact with human learners across multiple turns for a given task. The development, implementation, and maintenance of the AutoTutor system require multiple forms of expertise that are often expensive and difficult to acquire but can be facilitated through the use of digital technologies that can help coordinate the development process. Specifically, we describe the architecture of the AutoTutor system with a focus on the conversational structures that it utilizes to guide conversations for effective learning and assessment as well as the computational challenges that require solving along the way. We then discuss various methodological challenges we had to face over the years and discuss some effective solutions as well as outstanding development horizons for the future.
