Patterns in children’s online behavior and scientificproblem-solving: A large-N microgenetic study

The purpose of this study is to investigate the viability of a computer-based system for collecting and analyzing data from large-N microgenetic studies of scientific reasoning. Most studies of children’s scientific problem-solving strategies take place in classroom settings with established norms that homogenize student behavior. Small sample sizes and the limited range of settings limit the generalizability of the conclusions that can be drawn. The current study examined the problem-solving behaviors of 154 children using web-based science simulations. Of the participants who conducted genuine attempts using the simulations, 52 of the participants were members of a sixth grade science class that used the simulations as part of a curriculum unit on infectious disease and epidemiology while 42 of the participants were free-choice users of an informal science website. The data collected through the site and analyzed offline supported the hypothesis that previously reported patterns of scientific problem-solving do not represent the full range of scientific reasoning strategies that participants used when outside the classroom setting. Implications for the continued development of automated analyses of microgenetic studies and future research on children’s scientific thinking are discussed. The development of scientific reasoning in children has major implications for science education and, by extension, the public’s understanding and production of scientific knowledge. In instructional contexts, reliable assessment of scientific reasoning skills is very difficult because children’s strategy use is highly variable, both within and across individuals (Kuhn & Phelps, 1982; Schauble, 1990; Shavelson, Baxter, & Gao, 1993). Thus, it is particularly important to understand the processes by which scientific reasoning develops in order to design effective interventions (Chen & Klahr, 1999). Patterns in Children’s Scientific Problem Solving 2 Past studies indicate that students approach scientific experimentation from different perspectives that govern their strategy selection, metacognition, and personal criteria for successful completion of presented tasks (Zimmerman, 2000). For example, Klahr and Dunbar (1988) identify two classes of problem-solvers—theorists and experimenters. Theorists develop a hypothetical model of a phenomenon and design tests to analyze the suitability of their model. In contrast, experimenters manipulate values for available variables and subsequently generate hypotheses that can account for their findings. In an alternative description, Schauble, Klopfer, and Raghavan (1991) characterize children’s experimentation approaches as representing either engineering goals or scientific goals. Those students with engineering goals manipulate independent variables in order to generate a specific outcome (i.e. value for the dependent variable), such as designing the fastest possible vehicle in the context provided. Scientific goals, however, entail an attempt to understand the causal relationship between independent and dependent variables. Both dichotomies represent differences in fundamental approaches to scientific problem-solving that characterize students’ understandings of what “doing science” means. Further, they both draw an important distinction between students who attempt to test a causal model and those whose first priority is to generate experimental outcomes. Microgenetic approaches to the study of scientific strategy development and understanding typically provide the greatest insights into students’ problem-solving processes (Kuhn, 1995). Microgenetic studies entail high frequency qualitative and quantitative observations of a change process in very short time intervals for relatively small numbers of participants (Siegler & Crowley, 1991). The power of the microgenetic method for the investigation of scientific reasoning lies in its ability to detect subtle and transient changes that may not be evident in process outcomes. However, microgenetic methods have two weaknesses that limit the generalizability of their findings. First, sample sizes are quite small. Second, the collection of both qualitative and quantitative data usually requires the physical presence of the researcher in a setting that facilitates the systematic observation and recording of subjects during a pre-defined task. These constraints, in conjunction with other practical considerations, most often lead researchers to study children’s scientific problem-solving within the context of the school classroom (but see Gleason & Schauble, 2000). Within this context, various empirical findings have been replicated (e.g., Kuhn & Phelps, 1982; Schauble, 1990). However, little is known about the respective impacts of the context and the constraints that the setting may place on the range of individual differences among participants on patterns of behavior (Klahr & Simon, 2001). The current study uses an informal, web-based environment to examine children’s scientific problem-solving behavior through interactive simulations of infectious disease spread. A subset of the participants were students in a sixth grade science class that was studying infectious disease. The remainder of the participants Patterns in Children’s Scientific Problem Solving 3 varied widely in geographic location and age, and used the simulations without instructions to do so after school hours or on weekends as an independent activity. The study addresses two research questions: 1. Can computer-based data capture and analysis overcome the limitations of the microgenetic approach to studying scientific reasoning while preserving the richness and complexity of trends in iterative problem-solving data? 2. Does the context in which students engage in scientific inquiry affect their problemsolving strategies? It is expected that this large-scale microgenetic study will preserve the richness and complexity typical of smaller microgenetic studies. However, the use of algorithms to analyze patterns in performance data will allow researchers to inquire meaningfully into larger populations. Further, it is expected that this geographically distributed sample will demonstrate differences in problem-solving approaches when compared with the performance of participants in a classroom setting. Specifically, non-classroom participants are expected to adhere less to the models of reasoning described in previous research (e.g., Klahr & Dunbar, 1988; Schauble, et al., 1991).