Measuring Implicit Bias in Graph Types and Studying the Relationship Between Information-Seeking Preferences and Graph Comprehension

• In Event: Scaffolding Self-Regulation, Co-Regulation, and Socially Shared Regulation of Future Learning: Affordances of Learning Analytics Dashboards

Abstract

This study employed an implicit association testing method to examine 1) individuals’ preferences for alternative data visualizations that could be incorporated into dashboards (i.e., line vs. bar graphs) and 2) how preferences for self-focused and comparative feedback predict the implicit preferences individuals present and their ability to comprehend data presented in visualizations.
Learning analytics research is reliant on visualizations to communicate messages about learning (i.e., awareness, Verbert et al, 2013) to spur learners into action (i.e., impact). This study adds to the growing research base exploring the impact of visualizations on motivation and self-regulated learning (e.g., Authors, Date) by incorporating an implicit association task to determine students’ implicit preferences between line and bar graphs, and whether motivation or self-regulated learning processes predict graph comprehension.
Research on implicit cognition has revealed that individuals possess subtle mental associations and biases that exist largely outside their awareness (Greenwald, Nosek, & Banaji, 2003; Nosek, Banaji, & Greenwald, 2002). The Implicit Association Test (IAT, Greenwald, McGhee, & Schwartz, 1998) is a widely-used measure of these associations, and been used to measure bias in areas focused on racial (e.g., Fazio & Olson, 2003), morality (e.g., Marquardt & Hoeger, 2009), and consumer preference (e.g., Friese, Wänke, & Plessner, 2006).
College students recruited via Amazon’s Mechanical Turk completed the IAT and survey measures.
An IAT to assessed the degree to which target pairs (e.g., bar graphs vs. line graphs) and categories (e.g., good vs. bad) were associated in implicit cognition. The idea behind the IAT is that one can more rapidly sort stimuli when pairings are compatible with one’s implicit associations. The Patterns of Adaptive Learning Scale (Midgley et al., 2000) was used to measure general academic goals; the Motivated Strategies for Learning (Pintrich & de Groot, 1990) measured effort regulation and test anxiety; the Motivated Information Seeking Questionnaire (under review) measured students’ preferences for self-focused or comparative information. Graph comprehension was measured by asking students to identify parts of a graph and infer relationships based on a graph. Items were scored for correctness.
Results indicated that students had no implicit preference for bar graphs when compared to line graphs (RQ1). Multiple regression analysis indicated that self-focused information seeking (i.e., students wishing to see academic information about them and them alone), was predictive of overall graph comprehension (Table 1, PDF).
This initial null finding for implicit preferences between bar graphs and line graphs suggests that students are likely to be unaffected by this design choice. Ongoing analyses with larger samples are required before we can generalize our findings. In contrast, the finding that a preference for self-focused information seeking predicted graph comprehension indicates that student preferences for some displays may have implications for their future learning. When designing dashboards, it appears that soliciting information about student preferences – including information seeking, mastery and performance goal orientations, etc. – may be important for ensuring that visual displays are optimized for use so that students are most able to make use of the information they provide.

Author

• Stephen J. Aguilar, University of Southern California