Coding Like a Data Miner: A Sandbox Approach to Computing-Based Data Science for High School Student Learning

Personal health tracking devices and internet-based digital platforms with the capacity to collect, aggregate, and store data at massive scales are examples of tools that have broadened priorities in computing to include data science. In response, there has been growing attention in research and practice emphasizing pre-college groups. This is partly because of the growing recognition-reflected in initiatives like CS4ALL, Code.org, Bootstrap: Data Science, Exploring Computer Science-that learning experiences before college are consequential in sustaining a robust pipeline of computer scientists and engineers. Despite these inroads, there is justifiable concern that existing efforts might not fully support learner development in the necessary conceptual, epistemological, and heuristic styles needed to productively parse and understand “big data.” This is because computing-based curricula that include data science often involve data curated by others (rather than learners directly), which results in simulated versions of practice instead of engagement that is realistically discursive and messy. This is further complicated by the persistent shortage of K-12 computer science teachers in general and even fewer who can design and implement curricula that support authentic engagement with data science. To address these issues, we leverage culturally relevant and constructionist perspectives in a sandbox (i.e., open-ended) science where tools like Scratch and electronic textiles (E-textiles) have had success expanding possibilities in computing to also include activities where learners can engage broadly along varied pursuits-and encounter challenges that spur computational thinking and problem-solving. The literature suggests that learning activities framed in this way encourage knowledge construction, practice literacies, and seriously impact learner attitudes, interest, and perceptions of growth in the field. This latter set of self-concept measures represents a few of many related key predictors of long-term field participation and persistence. In this work-in-progress scholarship of discovery research, we co-develop, with youth and educators, “Coding Like a Data Miner” (CLDM)-a sandbox approach to computing-based data science wherein learners access a social media platform, Twitter, to mine, analyze, and understand quantitative and qualitative data sources. In this preliminary work, we assess affordances in co-developing a curriculum that leverages sandbox approaches to data science. Ultimately (and what will be presented in our final submission), we aim to study learning outcomes when high school students' access, analyze and make sense of “big data” sets of their own. We collaborated with high school teachers in a West Texas/Paso Del Norte region where computer science educators are exceptionally scarce and where there is an urgent and persistent need to support underrepresented learner access to burgeoning areas of computing. Using mixed-methodological approaches (e.g., quantitative analysis of learner pre- and post-survey responses along with qualitative assessments of semi-structured interview data), we address the following research questions: (1) What affordances exist using co-design approaches to develop sandbox data science for pre-college learners? (2) Which computational concepts do students learn when carrying out CLDM activities, (3) Which computational practices do high school students enact when mining, processing, and analyzing big data sets in CLDM? (4) How do learner knowledge and perceptions about data science shift after participating in CLDM? We use contemporary perspectives in computing education, constructionism, and equity to discuss how open-ended sandbox approaches to computing-based data science support learner computational thinking, practice literacies, and field perceptions.

I. Introduction

Digital technologies-embedded with automated data collection capabilities-have transformed the nature of datashifting it from small data sets that can be readily processed and analyzed using traditional methods to massive compilations of information that can only be handled using computing. Concomitant with these developments is also a change in knowledge and skills needed to with and understand these data alongside their real-world applications [1]. This has ignited calls for a reevaluation of how data science might fit in pre-college education [2], including concerns about how I might deliver data science in ways that “count” as well as ways the field may be enacted in service of productive learning [3]. Computer science education (CSE) has shown viability to support pre-college data science education [4] because CSE provides opportunities for learners to conduct data science through authentic acts [5] whereby learners use computer code to harness data-generating technologies in order to access, collect and make sense of large amounts of information (e.g., Bootstrap: Data Science, Exploring Computer Science, etc.) that is tied to everyday contexts.