Designing to Support Data Visualization as an Exploratory Tool in Science

• In Event: 29.027 - Designing Digital Environments to Support Mathematical and Scientific Reasoning: Theoretical and Disciplinary Perspectives

Abstract

Analyzing and interpreting data has long been a part of scientific practice, but technological innovation has fundamentally changed how this is done. Data visualization has become an important exploratory and communicative tool in science (Friendly, 2008). Students have few opportunities to learn about data visualization, despite its relevance for complex and civically important topics: climate, health, biodiversity (Edelson & Gordin, 1998; “Dealing with Data”, 2011). We are exploring curricula and tools to support middle school students as they work with and create their own data visualizations to investigate scientific systems. As part of this, we are designing a programmable sketchbook to create data-driven visualizations. Questions driving the study are: To what extent do learners use data visualization to examine relationships and structures in scientific systems? And, how can such use be supported?
Our design and research are informed by constructionist (Papert, 1980) and model-based (White & Frederiksen, 1998) theories of learning, which suggest learners construct understandings through the construction and testing of public artifacts and representations. We also draw on the theory of meta-representational competence (diSessa, 2004) which highlights learners’ abilities to interpret and create visual displays of scientific and mathematical information. When modeling scientific data, students’ meta-representational competence co-develops with disciplinary understandings of the relationships and structures that characterize the system being explored, driving further cycles of inquiry (Lehrer & Schuable, 2002). We conjectured that encouraging learners to invent and refine data visualizations, using sketching as an introduction, could extend their meta-representational competencies and their attention to the relationships and structures that underlie complex, multi-dimensional systems.
To investigate this conjecture, data were collected in three phases across classroom and interview settings. These include 42 classroom students’ drawn representations of scientific systems of varying complexity, 12 interviews with individual students and dyads as they interpreted and redesigned publically available data visualizations, and 3 interviews with pairs of students as interpreted with these same visualizations and created their own using the digital sketchbook. Interview transcripts and student artifacts were coded using thematic analysis methods (Aronson, 1995) for evidence of attention to relationships, where parameters of the system are expected to vary or co-vary (Carlson et al., 2002); and structure, the coherent linkages across those relationships to form a larger system (Stratford, Krajcik & Soloway, 1998).
Most learners attended to relationships, and many to structures, when creating drawn representations and interacting with existing data visualizations. However, these attentions were mostly absent when redesigning or creating their own visualizations. Upon further examination, we found that when learners focused on the datasets they were working with rather than the systems they reflected, they became overwhelmed and tried to create visualizations inclusive of all data. This challenges our approach, because learners must have access to their data when creating visualizations.
These findings have direct implications for our own design, but also for data visualization and data science education more broadly. In future iterations, we will explore notions of focus and linking (Buja, et al., 1991) from information visualization theory as principles to scaffold student design of representations.

Authors

• Michelle Hoda Wilkerson, University of California - Berkeley

• Vasiliki Laina, University of California - Berkeley