Construction Knowledge From Interactive Visualizations: How Comprehension of Complex Relations Influences Representational Choice

• In Event: Learning With Static and Dynamic Visualizations

Abstract

Users often prefer representations containing task-irrelevant information (Smallman & Cook, 2011) even when they are experts (Hegarty, Smallman, Stull, & Canham, 2009). Representational choice, a component of meta-representational competence (diSessa, 2004), may also be influenced by conceptual understanding of what the display represents. Canham and Hegarty (2010) stated that an individual’s conceptual understanding should enhance their focus on task-relevant information. We argue that conceptual understanding enhances a focus on task-relevant information and accordingly requires less deselection of task-irrelevant information. Therefore, this study aimed to investigate the influence of an individual’s conceptual understanding on representational choice.
Two factors, Essay Task (content, surface) and Initial Display (overview, blank) were crossed yielding a 2 x 2 between-subjects design with two dependent variables: Representational choice and Performance. Participants viewed an Interactive Data Visualization (IDV) depicting indicators of human development and completed an essay task by conceptually interpreting the data (content) or by only describing the display’s functions (surface). To use the IDV, users had to either (1) deselect task-irrelevant information from an overview initial-display or (2) to select task-relevant information from a blank initial-display. Representational choice was measured as the amount of task-irrelevant data chosen to answer graph-reading questions. Performance was measured as time spent on correctly answered graph-reading questions.
Ninety-five undergraduates participated in the study. Participants began by developing domain knowledge by reading a passage about historical events and were assigned to one of two essay tasks (content, surface). Finally, they completed 12 graph-reading questions consisting of multiple-choice questions, using the initial-display of their experimental condition.
We hypothesized that participants assigned to the content essay task, by building conceptual understanding, would deselect less task-irrelevant information from the overview initial-display, but perform as well as users in the surface essay task.
Representational choice. Results indicated a significant main effect for Essay Task, F(1, 91) = 5.888, η² = .06, p = .017, however, this finding is qualified by the significant interaction with Initial Display, F(1, 91) = 5.027, η² = .05, p =.025. In the blank condition, both task groups showed low task-irrelevant data. In the overview condition, the content group chose representations with more task-irrelevant data as compared to the surface group, d(94) = .67, p < .01.
Performance. Results indicated a significant interaction between Essay Task and Initial Display, F(1, 91) = 5.723, η² =.06, p = .019. The fastest performance was accomplished by the surface group in the blank condition. In the overview condition, the content group, even in the presence of more irrelevant information, attained the same performance as the surface group.
The representational choice of the content group in the overview condition contained the more irrelevant data. However, their task performance did not suffer in terms of the speed for correct answers. Therefore, they seemed to be able to focus attention on the task-relevant data in the presence of irrelevant noise. In sum, our results can be explained by Canham and Hegarty’s (2010) finding about the effect of conceptual understanding on attention allocation.

Authors

• Ulrich Ludewig, University Tuebingen

• Erica de Vries, University of Grenoble

• Neil H. Schwartz, California State University - Chico

• Neil Garrett Jacobson, California State University - Chico

• Amy Rae Fox, University of California - San Diego

• Savannah Loker, California State University - Chico