UC Berkeley

The purpose of this dissertation is to study students' competencies in generating scientific explanations within the domain of complex systems, an interdisciplinary area in which students tend to have difficulties. While considering students' developing explanations of how complex systems work, I investigate the role of prior knowledge and how students' explanations systematically vary across seven problem contexts (e.g. the movement of sand dunes, the formation of traffic jams, and diffusion in water).

Using the Knowledge in Pieces epistemological perspective, I build a mini-theory of how students construct explanations about the behavior of complex systems. The mini-theory shows how advanced, "decentralized" explanations evolve from a variety of prior knowledge resources, which depend on specific features of the problem.

A general emphasis on students' competences is exhibited through three strands of analysis: (1) a focus on moment-to-moment shifts in individuals' explanations in the direction of a normative understanding; (2) a comparison of explanations across the seven problem contexts in order to highlight variation in kinds of prior knowledge that are used; and (3) a concentration on the diversity within explanations that can be all considered examples of emergent thinking. First, I document cases of students' shifting explanations as they become less prototypically centralized (a more naïve causality) and then become more prototypically decentralized over short time periods. The analysis illustrates the lines of continuity between these two ways of understanding and how change can occur during the process of students generating a progression of increasingly sophisticated transitional explanations. Second, I find a variety of students' understandings across the problem contexts, expressing both variation in their prior knowledge and how the nature of a specific domain influences reasoning. Certain problem contexts are easier or harder for students, depending on the kinds of prior knowledge resources activated. For example, in the sand dune problem context, where students are able to access a wide range of intuitive resources that are applicable at multiple levels, coming to explain decentralized causality is relatively straight forward. Third, I find that for these students' emergent thinking is not a unified entity. It is diverse in its nature and varies across problem contexts and across the kinds of prior knowledge that students evoke.

This dissertation illustrates the importance of students' prior knowledge resources in their understanding and developing explanations for how complex systems work. Combined, these results suggest that the fundamental diversity in explanations needs to be respected. Instruction should emphasize the generative process of explaining based on students' prior knowledge rather than any a priori taxonomy of forms of explanations to be learned.