Conversational Strategies that Support Idea Generation Productivity in Groups

Our recent work has shown that conversational tutors that use social interaction strategies can achieve significant learning improvement through better management of student attention. In continuation with this line of investigation, this paper presents an experiment that measures the separate and joint effects of the nature and style of automatically generated interventions. We investigate aspects of designing intervention prompts from two different angles: first in terms of the addition of socially oriented turns over and above the instructional turns required to keep the students on track, and second in terms of the style of presentation of the tutor’s instructional turns. We see positive effects of both manipulations, but on different outcome metrics. Thus, we conclude that the two directions can be productively combined into a single socially enhanced feedback approach, which contributes to the literature on dynamic forms of support for collaborative learning.