Generative Modeling in Computer Science

• In Event: 63.025 - Supporting Modeling Epistemologies in the Science Classroom

Abstract

Objectives. We describe work in the intersection of computer science and modeling, supported by the Group-based Cloud Computing (GbCC) project. Portions of the activity sequence we present have been implemented in the UTeach Computer Science Principles (CSP) curriculum and in classrooms at middle-school, high-school, and university levels. Across this range of settings, they foster group-centered learning and teaching to broaden success and deepen interest in authentic STEM practices.

Theoretical Framing. Perhaps the most salient feature of classroom learning is its group-based nature. However, typical classroom instructions focus on individuals, emphasizing recall and procedural exercises and relying heavily on an Initiation, Response, Evaluation (IRE) pattern (cf, Wells, 1993). In contrast, generative approaches provide situations for groups to construct new relations between existing knowledge and shared experiences (Wittrock, 1991; Authors, 2004). Network-supported learning and teaching following generative principles has improved student outcomes in algebra and in science, balancing agent-based and aggregate forms of systems reasoning (Authors, 2014).

Methods/Data. The GbCC system builds upon the NetLogo Web (Author, 1999; 2015) platform, which we employ to support generative activities and engage learners in the core scientific practice of modeling through a computer science lens.
An introductory activity exposes learners to NetLogo language primitives and the GbCC tools for sharing learner-created artifacts: participants program computational agents to produce a display expressing “beauty.” Buttons introduce key commands, and a running record of executed commands bridges participants to text-based programming. Participants can freely share their work in a public gallery – publishing both drawings and the code to produce them. Iteratively testing code, they refine and share their “beauty” procedures. They can sample and remix classmates’ creations, incorporating and extending aspects of each other’s code.
Similar network-mediated sharing supports participants throughout activities that explore the spread of disease from multiple perspectives. An initial participatory simulation (Authors, 1999) offers an immersive environment for engaging with both the experience of disease and patterns in data describing its spread. Settings for the simulation allow the class to explore the effects of different features of pathogens or of a population experiencing an outbreak. Emergent lines of inquiry are explored here, generating and sharpening questions for further study. Subsequent activities explore the range of behaviors that an extensible model of disease can express; and here the artifacts shared to the gallery include the model settings and modular code extensions, as students create procedures to model infection-avoidance behavior in the population.

Results and Scholarly Significance. Leveraging the group’s diversity of thinking, programs for “beauty,” for disease spread, and for population behaviors during outbreaks are developed in socially-mediated ways, offering an image of computational modeling as a group-level phenomenon. This image suggests an epistemological shift, as the group acts as a bootstrap for generating distributed knowledge of the modeling language (“beauty”); as a representational infrastructure for understanding the phenomenon itself (participatory simulation); and as a community of inquiry and an authentic audience for exploring and extending a model to make inferences and claims about it (experimentation and coding activities).

Authors

• Corey Brady, Vanderbilt University

• Walter M. Stroup, University of Massachusetts Dartmouth

• Justin Cannady

• Anthony Petrosino, The University of Texas at Austin

• Uri J. Wilensky, Northwestern University