The Power of Agent-Based Restructurations in Advanced Materials Science Learning

• In Event: Restructuring Concepts and Tools Through a Complexity Perspective

Abstract

Objectives and theoretical framework
As described in the first paper (Wilensky, here), the medium in which knowledge is encoded may profoundly affect its message. New kinds of representations with similar epistemic magnitude are also changing the nature of modern science and engineering, especially due to new computational representations (Oden et al., 2006).
Whereas traditional scientific practice leans heavily on formulas and differential equations that model the phenomena through a lens of macrobehaviors, new computational methodologies offer practices for modeling these same phenomena through microbehaviors (i.e. ABM). Within many fields in engineering and science, these types of representations have achieved a similar level of acceptance and reliability as their corresponding analytical representations (Raabe, Roters et al. 2004). Seminal research by Anderson, Grest and colleagues set into motion a vision of “computational alchemy” that has transformed fields such as materials science and solid-state physics (Anderson, Grest.et.al., 1985).
However, even as computational science transforms science, the same has not happened to the teaching of science and engineering at the college level: we are still pedagogically trapped in what Bachelard (1934/1966) called the “psychology of determinism” (see also Wolfram, 2002). A chronic problem in engineering education is that, by relying on an ever-expanding corpus of myriad domain-specific equations and heuristics, students’ knowledge end up being superficial, overly specific, ungenerative, and fragmented. Restructurations theory (Wilensky & Papert, 2010) helps us conjecture that part of these shortcomings result from cognitive discontinuity between, on one hand, intuitive forms of sense-making that deal with objects and actions, and, on the other hand, algebraic formulas traditionally used in this discipline.

Methods, techniques, modes of inquiry, and data sources
In this research, I examined how multi-agent models – taken as a new representations infrastructure for certain domains in engineering may offer greater cognitive continuity from sense-making to disciplinary knowledge. I empirically tested the hypothesis in an undergraduate class over a full semester, with 21 students divided into two groups, using a mixed-methods approach. I had as data sources semi-clinical interviews, artifacts, and logfiles from the modeling environment.

Results and scholarly significance
The treatment group, which used ABMs, engaged in sensemaking activities that were fundamentally different from the control group, which used traditional differential equations within a more traditional representational infrastructure. The findings point to the cognitive and epistemic advantages of the complexity restructuration for some types of content. Using ABM, students not only had a better recall of basic principles in solid state physics, but developed and later applied generative principles to new phenomena that would traditionally belong to different curricular units yet, it turns out, share in structure and function. That is, we have witnessed students linking productively among several phenomena that according to traditional taxonomy are ontologically disparate. As such, this study concurs with Jacobson and Wilensky (2006) to question the epistemic consequences of extant curricular taxonomy, showing the inferential power that agent-based perspectives bear as compared to macro, equation-based models.

Author

• Paulo Blikstein, Teachers College, Columbia University