Drawing upon the literature in computational modeling, multivariable reasoning, and causal attribution, this study aims at characterizing multivariable reasoning practices in computational modeling and revealing the nature of understanding about multivariable causality. We recruited two freshmen, two sophomores, two juniors, two seniors, four master's students, and four PhD students in atmospheric sciences as participants. Participants' reasoning practices and understanding of multivariable causality were examined using semistructured interviews and recordings of their computer activities. Analyses show that participants with high expertise tended to take a mechanism approach to predict and identify relationships, focused more on multivariable relationships, and purposefully selected and tested variables. The findings also indicate that understanding about multiple causality involved recognition and identification of the integration rules of multiple effects and the attributes of variables (e.g., interactive and reciprocal) and relationships (e.g., direction and feedback loop). Additionally, this study suggests an interaction between participants' reasoning practices and their understanding of multivariable causality; participants' understanding about the integration rules and the attributes could initiate reasoning practices, and by the enactment of practices, the rules and attributes were confirmed and examined. This study provides insight into the nature of multivariable reasoning and the design of computer-based modeling tools.
ASJC Scopus subject areas