Understanding the complexity of climate change with agent-based models: A study of contrasting learning designs


Research output: Contribution to conferencePapers


A centrally important issue concerns how to effectively teach complex systems ideas in K12 science classes. In this study, we explore whether a common pedagogical approach used by science teachers is sufficient for students to learn complex systems ideas, or whether alternative approaches may be more effective. We report on classroom-based research with ninth-grade students using agent-based computer models (ABM) to learn a core set of complex systems ideas as part of a unit dealing with climate change, earth systems, and sustainability. The learning designs were based on the Sequences of Pedagogical Structure (SPS) framework (Jacobson et al., 2013). Pedagogical structure is broadly conceived as structuring a problem, scaffolding, instructional facilitation, providing worksheets or scripts, and so on. The SPS framework has four core components: (a) high-to-high structure (HH), (b) high-to-low-structure (HL), (c) low-to-low structure (LL), and (d) low-to-high structure (LH). Direct instruction approaches such as those discussed in Kirschner et al. (2006) are classified as HH, and minimally guided approaches as LL. Examples of HL sequences include commonly used classroom approaches such as teacher led presentations (high structure) followed by students working on tasks in class (low structure). Examples of LH sequences are found in productive failure (PF) (Kapur & Bielaczyc, 2012), where the initial Idea Generation and Exploration (IGE) phase involves low structure and the subsequent Consolidation phase involves high structure. The experimental condition used a LH SPS sequence based on PF as previous research had demonstrated significant learning of challenging topics in science and mathematics with this learning design (Jacobson et al., 2013; Kapur, 2008, 2012). The experimental condition involved a PF IGE phase in which the students worked on a complex systems problem and complexity ABM and then a climate change problem with a climate related ABM; this was followed by a teacher led Consolidation phase that discussed the targeted complexity and climate concepts for each day. The comparison condition was based on a teacher’s suggestion that a more traditional teaching approach of teach first and followed by students working on problems (a HL SPS sequence) would be more effective. Students in the comparison condition had initially high pedagogical structure with teacher led instruction on the targeted complexity and climate concepts—essentially the same as the PF Consolidation phase—followed by low pedagogical structure with students working on the same problems with the same ABMs as in the experimental condition. Findings (see Table 1) indicate significant learning of straightforward ideas like “greenhouse gases” by both groups on the posttest. However, for conceptually challenging complex system ideas, such as “emergent properties,” only the LH experimental group showed significantly higher performance. Regression analysis showed that the complex systems score was a significant predictor of students’ performance on the climate ideas even after taking the group and other related variables into account. We hope these findings might stimulate further interest by researchers and teachers in learning designs based such as productive failure for learning challenging scientific knowledge about complex systems with agent-based computer models and in science education generally.
Original languageEnglish
Publication statusPublished - Apr 2016


Jacobson, M. J., Markauskaite, L., Portolese, A. E., Lai, P. K., & Kapur, M. (2016, April). Understanding the complexity of climate change with agent-based models: A study of contrasting learning designs. Paper presented at the 2016 AERA Annual Meeting: Public scholarship to educate diverse democracies, The Walter E. Washington Convention Centre, Washington, DC.

Fingerprint Dive into the research topics of 'Understanding the complexity of climate change with agent-based models: A study of contrasting learning designs'. Together they form a unique fingerprint.