Making "Data Claims" as an (Inter)Disciplinary Practice in the Science Classroom

• In Event: Exploring Data Science Across the Curriculum and Across Grade Levels

Abstract

We have been studying how to integrate analyses of large public datasets into science curricula. There are growing calls for such integration (NSF, 2018); but our primary motivation is to disrupt the assumed ‘natural’ and ‘neutral’ status of data in today's world (Gillborn, Warmington, & Deback, 2017). Thus we seek to position second-hand data as a transformable and critiqueable source of scientific evidence (McNeill & Berland, 2017). This involves not only asking students to analyze and argue from provided data but also to assess the utility and trustworthiness of data in relation to investigative needs (Authors, 2018).

Our work draws from theory about how disciplinary norms and practices emerge in classrooms (Lehrer, 2009), and how interdisciplinarity and sociocultural factors influence what is valued (Bing & Redish, 2009; Bang, et al. 2012). A lack of clarity around interdisciplinary expectations can create pedagogical confusion: Scientific norms emphasize causal and mechanistic relationships (Russ, et al. 2008), while statistical norms emphasize quantitative, low-inference reasoning (Makar & Rubin, 2009). Assessing the validity of data can bridge this gap, but requires disciplinary flexibility. Thus we ask: How do students navigate the interdisciplinary expectations of data-intensive science investigations?

We conducted a task analysis of science units enacted with multiple teachers in a large city in California. Each unit lasted 1-2 weeks and was driven by a causal question. This presentation will report on five seventh-grade classrooms in a suburban school with predominantly Latinx students who explored the question, “Did wolves change the rivers of Yellowstone?” Students used publicly available datasets, a visual analysis tool (CODAP; Finzer & Damelin, 2014), and prior content knowledge (predator-prey dynamics; energy webs) to address the question. Two analysts content logged screencast recordings of 36 student groups during the unit.

Both analysts independently identified data claims as one way students manage the interdisciplinary expectations of data-driven science investigations. These are “low inference” (Kerlin, McDonald, & Kelly, 2010) claims about what conclusions can be drawn directly from a dataset. Most groups struggled to make causal claims but frequently made valid data claims that reflected productive paths for investigation.

Instructors overlooked or dismissed these claims, which are more granular than the causal claims most valued in scientific argument, we suspect for both disciplinary and raciolinguistic reasons. One example of such a dismissal is: "Your claim isn’t what the graph says, that’s evidence. I don’t want to see any numbers in this claim box." [P1T1D4]. These dismissals represented missed opportunities to support student work, and interruption to students' progress. When data claims were supported by instructors, students tended to make progress and later link their data claims to causal inference: "There was 12 [beaver colonies] by 2009 and they might be making dams." [P3T5D4].

Data claims represent an intermediate step between the complex interdisciplinary expectations of analyzing data and making inferential claims. In this way, offer connective tissue between “statistical/data science” and “science” practices. They should be better understood and explicitly supported through instruction and curricular design.

Authors

• Michelle Hoda Wilkerson, University of California - Berkeley

• M. Lisette Lopez, University of California - Berkeley

• Julio C. Jaramillo, University of California - Berkeley