A study of students’ reasoning about probabilistic causality: Implications for understanding complex systems and for instructional design

Abstract

Understanding complex systems requires reasoning about causal relationships that behave or appear to behave probabilistically. Features such as distributed agency, large spatial scales, and time delays obscure co-variation relationships and complex interactions can result in non-deterministic relationships between causes and effects that are best understood statistically. Causal Bayesian Research (e.g. Gopnik and Schulz in Causal learning: psychology, philosophy, and computation, Oxford University Press, New York, 2007) suggests that summing across probabilistic instances is inherent to human causal induction, yet other research (e.g. Schulz and Sommerville in Child Development 77(2):427–442, 2006) suggests a human tendency to assume deterministic relationships. Classroom science learning often stresses the replicability of outcomes, putting this learning in tension with understanding probabilistic patterns in complex systems. This investigation examined students’ reasoning patterns on tasks with probabilistic causal features. Microgenetic studies were conducted in multiple sessions over a school year with students in kindergarten, second, fourth and sixth grades (n = 16) to assess their assumptions when dealing with tasks from four domains: social; games; machines; and biology. Later sessions attempted to scaffold students’ understanding using connection making and analogical reasoning. This paper reports on the overall patterns and trends in the data. Most students held a deterministic stance at the outset; however, at least one student at each grade level reasoned probabilistically from the start. All students except one eventually revealed at least one topic for which they held a primarily probabilistic stance. The results have implications for how students reason about complex systems and for how patterns of co-variation and evidence in science are discussed.