Principal Investigator: Barry Gholson
Title: An Implementation of Vicarious Learning with Deep-Level Reasoning Questions in Middle School and High School Classrooms
Purpose: Research indicates that one-on-one human tutoring improves student learning. Indeed, the argument can be made that human tutoring is the gold standard for facilitating learning, compared with standard classroom activities and conventional computer-based training. However, the number of accomplished human tutors is small, and the time required often exceeds the number of hours available for one-on-one tutoring. Given these constraints, there have been attempts to develop computerized, intelligent tutoring systems. This research project will compare different versions of an intelligent tutoring system developed by the research team in prior work, AutoTutor, to examine how best to support learning of course content. For example, students will compare the use of AutoTutor in its typical interactive version to a vicarious condition where students listen to and observe the AutoTutor agent presenting the same course content, but they do not physically interact with the source of the materials. This will enable the research team to determine whether students are equally able to learn course content during interactive and vicarious learning. They are also examining whether mastery of course content can be improved by asking students to generate reasoning questions during study (e.g., “Why is x important?”). The researchers are working with school-aged students in the areas of computer literacy and Newtonian physics.
Setting: The research is taking place in the greater Memphis metropolitan area.
Population: Middle and high school students attending Memphis city schools are participating in this research project.
Intervention: Deep-level reasoning questions (such as Why and How questions) are embedded in an intelligent tutoring system called AutoTutor, which currently tutors in computer literacy and Newtonian physics. AutoTutor serves as conversational partner with the learner and encourages students to provide answers to questions until particular concepts are mastered. The researchers are developing guidelines for teachers to support the use of AutoTutor during classroom instruction.
Research Design and Methods: A series of studies are being completed over the 3 years of the grant. In the first year, the researchers are conducting laboratory experiments to determine if increases in learning observed when deep-level reasoning questions were used to facilitate learning by college students holds for when the strategy is used by middle and high school students. These experiments are being conducted using a between-subjects pretest/posttest design. In the first two experiments, students will be randomly assigned to one of three experimental conditions in order to test the effects of presenting content in an interactive or vicarious format. At the same time, the researchers are working with teacher consultants to help prepare course content to be presented in the context of deep-level reasoning questions. Teachers are developing guidelines for the use of small in-class discussion groups in the experimental condition that will complement the presentation of course content via the computer. During the second and third years of the projects, the impact of the deep-level reasoning questions on student learning is being compared to students who receive the same course content as standard classroom activities. Classes of students are being randomly assigned to condition. In addition, the number of course units instructed via this method is being increased each year, and two new courses at each grade level are being modified to use this technique.
Control Condition: In the laboratory experiments, the three different conditions allow explicit comparison between different versions of the AutoTutor. In the classroom experiments, control students receive the same course content presented as standard classroom instruction (e.g., delivered on the computer, but without the deep-level reasoning questions). Students who served as controls during the first half of the semester become experimental students during the second half of the semester, when they have two units of course content presented in the context of deep-level reasoning questions.
Key Measures: Both experimenter-developed and teacher-developed exams covering course content and asking for responses to deep-level questions are the primary measures used to measure learning. In the experimental evaluation of the use of deep-level questions in Newtonian physics, the Force Concepts Inventory is being used to assess pretest to posttest learning gains. Audiotapes are being collected of the in-class discussions before, during, and after course content has been presented in the context of deep-level reasoning questions.
Data Analytic Strategy: Analysis of variance and analysis of covariance are being used to determine the impact of embedding vicarious deep-level reasoning questions into AutoTutor on student learning. Audiotaped data is being used to evaluate the degree to which hearing deep-level questioning in the context of AutoTutor transfers to student generation of deep-level questions during in-class discussions.