How Domain Differences Impact the Mode Structure of Expert Tutoring Dialogue

Abstract

While human-to-human dialogue in tutoring sessions has received considerable attention in the last 25 years, there exists a paucity of work examining the pedagogical and motivational strategies of expert human tutors. An established trend in the tutorial dialogue community is to study tutorial dialogues in a very fine-grained manner, at the level of the speech act or dialogue move. The present work offers a coding scheme that examines larger, pedagogically distinct phases as the unit of analysis, referred to as “modes”, which exist in expert tutoring and provide the context needed to understand patterns of dialogue moves. The eight modes identified by this coding scheme are the Introduction, Lecture, Modeling, Scaffolding, Fading, Highlighting, Off Topic, and Conclusion mode, and each mode was reliably identified at or above the .8 kappa level. After determining how often modes occur and the amount of dialogue devoted to them in expert tutoring sessions, differences between the domains of math and science were investigated. Significant variance between the domains was revealed using this largergrained coding scheme, particularly in how Lecture and Scaffolding are used in expert tutoring. While these two modes tend to dominate most tutorial dialogue in this sample regardless of domain, the differences in their frequency and the amount of dialogue devoted to each mode suggest diverse tutoring goals associated with each domain. Other subtle differences in mode distributions draw attention both to the complexities of expert tutoring and the danger of generalizing tutorial structures across domains.