Conversational AIS as the Cornerstone of Hybrid Tutors

Abstract

This paper describes the benefits of artificially intelligent conversational exchanges as they apply to multi-level adaptivity in learning technology. Adaptive instructional systems (AISs) encompass a great breadth of pedagogical techniques and approaches, often targeting the same domain. This suggests the utility of combining individual systems that share concepts and content but not form or presentation. Integration of multiple approaches within a unified system presents unique opportunities and accompanying challenges, notably, the need for a new level of adaptivity. Conventional AISs may adapt to learners within problems or between problems, but the hybrid system requires recommendations at the level of constituent systems as well. I describe the creation of a hybrid tutor, called ElectronixTutor, with a conversational AIS as its cornerstone learning resource. Conversational exchanges, when properly constructed and delivered, offer substantial diagnostic power by probing depth, breadth, and fluency of learner understanding, while mapping explicitly onto knowledge components that standardize learner modeling across resources. Open-ended interactions can also reveal psychological characteristics that have bearing on learning, such as verbal fluency and grit.