One-on-one and small group conversations with an intelligent virtual science tutor

In this study we investigated students' conversations with a virtual science tutor (Marni), either individually or in small groups. These constituted two treatment conditions. Students were presented with narrated multimedia science problems and explanations followed by question-answer dialogs with the virtual tutor. Students who received either one-on-one or small group tutoring received the same set of multimedia presentations and questions posed by the virtual tutor. Students in the small group condition discussed their answer before one student from that group responded to the tutor. We asked if students receiving tutoring using the virtual tutor in groups would demonstrate learning gains equivalent to those of students receiving one-on-one tutoring. We also asked if both groups would demonstrate greater learning gains from pretest to posttest than students in business-as-usual (control) classrooms who did not receive supplemental tutoring. One hundred eighty-three (183) students (in 13 classrooms at 4 schools) participated in the study. Of the 183 students, 114 were randomly assigned to tutoring in small groups using Marni; and 69 students received one-on-one tutoring with Marni. When compared with the control group, effect sizes for were d = 0.048 for the group tutoring condition and d = 0.51 for the one-on-one tutoring condition. A two-way ANOVA suggested a main effect for tutoring group, F = 16.8, df (41,171), p < 0.001. In general, students reported benefiting from listening to one another, and from the small group interactions, even though they sometimes disagreed with the answer reported by the small group. We conclude our findings with a vision for a next generation of virtual science tutors that can facilitate discourse and argumentation among students in small groups, leading students to build on each other's ideas to construct accurate science explanations.