Case Studies in Learning by Teaching Behavioral Differences in Directed versus Guided Learning

Case Studies in Learning by Teaching Behavioral Differences in Directed versus Guided Learning Gautam Biswas, Krittaya Leelawong, Kadira Belynne, and Bilikiss Adebiyi (gautam.biswas@vanderbilt.edu) Department of EECS/ISIS, Box 1824, Sta B Vanderbilt University Nashville, TN 37235. USA. Details of the Betty’s Brain system, and experiments that we have conducted with this system are presented elsewhere (Biswas et al., 2004a, 2004b; Davis et al., 2003; Leelawong et al., 2002). Our experiments in fifth grade science classrooms showed that learning by teaching with self-regulated learning (SRL) (Zimmerman, 1989) led to better transfer than situa- tions where the students were taught by a pedagogical agent (Johnson et al., 2000). The encouraging results from the Betty’s Brain experi- ments prompted us to take a more detailed look at how stu- dents used the different versions of the system, especially the feedback they got from Betty and the mentor. This called for more in-depth analysis of student activities and behaviors that can be attributed to the self-regulation strategies. In this pa- per, we perform a case analysis of the log files of student ac- tivities during the main and transfer studies. The next section briefly outlines the teachable agent system and the experi- mental study we conducted in the fifth grade science class- rooms. Abstract Our studies with Betty’s Brain, a learning by teaching envi- ronment, have shown that the system is effective in helping fifth grade students gain a good understanding of river ecosys- tem concepts. The use of self-regulation strategies demon- strated that the learning gains transferred to new domains where students worked without the self-regulation system. This paper analyzes the log files of the student activities to determine which activities in the learning environment contribute to the students developing metacognitive strategies that contribute to their preparation for future learning. Keywords: learning by teaching; self-regulation strategies; teachable agents; behavior log analysis. Introduction An important challenge for computer-based learning envi- ronments is to demonstrate that they develop students’ abili- ties to learn, even after they leave the computer environment (Anderson et al., 1995). The cognitive science and education literature have established that understanding and transfer are greatly aided by constructivist, exploratory, and anchored learning environments (Bransford et al., 2000). Relevant so- cial interactions can add motivation and also enhance effec- tive learning (Soller, 2001). We have adopted the learning by teaching paradigm (Palincsar et al., 1984) as the basis for designing learning en- vironments. Researchers (Bargh et al., 1980; Biswas et al., 2001) have shown that people learn more when they teach others as opposed to when they prepare to take tests them- selves. Teaching involves a number of constructivist activi- ties. Teachers prepare and organize their knowledge in antici- pation of the needs of their students. They provide explana- tions and demonstrations during teaching and reflect on the questions and feedback they receive from the students. Effec- tive teaching also requires the explicit monitoring of students during and after the teaching process. This helps teachers to evaluate their own understanding and the methods they have used to convey this understanding to students (Artzt et al., We have implemented a software agent, Betty’s Brain, that students can teach using concept maps (Novak, 1996). Once taught, Betty uses qualitative reasoning methods (Forbus, 1984) to answer questions. Students reflect on Betty’s answers, revise their own knowledge, and make cor- responding changes to the concept map to teach Betty better. Betty’s Brain The main interface to the Betty’s Brain system is shown in Figure 1. This interface implements three primary compo- nents of teaching: (i) teach Betty using a concept map, (ii) query Betty to see how much she has understood, and (iii) quiz Betty to see how well prepared she is to take the test, which will gain her admittance to the Science club. These components model the primary constructivist activities in- volved in various phases of the teaching process: preparation, teaching, and monitoring (Colton et al., 1993). Students teach Betty about entities in river ecosystems, e.g., fish, algae, macroinvertebrates, waste, and bacteria, and the relations between these entities, e.g., fish consume dissolved oxygen, algae replenish it, waste is consumed by bacteria to produce nutrients, and nutrients help algae grow. Betty is equipped to reason about the knowledge that she has been taught to answer questions like “if macroinvertebrates in- crease what happens to bacteria?” She uses qualitative rea- soning methods (Leelawong et al., 2001) to make inferences through chains of links. She determines that on the one hand macroinvertebrates create more waste and waste feeds bacte- ria, but on the other hand, macroinvertebrates also consume algae and dissolved oxygen, and, therefore, there is less oxy- gen for bacteria to breathe, and this inhibits their growth.