Cognitive Fit in Visualizing Big Data

Abstract

This dissertation examines the consequences of cognitive fit in visualizing big data. Specifically, it focuses on the interplay between different types of business data analysis tasks and visualization methods, and how the defining characteristics of big data (i.e., volume and variety) moderate the outcomes concerning data analysis performance (i.e., solution time and solution accuracy). A 12-cell repeated-measures laboratory experiment (n=145) using eye trackers is conducted to test the hypotheses. Data analysis performance is observed to improve when the information emphasized by a visualization method matches the specific information requirements for a data analysis task. Such improvements in data analysis performance are further amplified when the visualized information has high volume and variety. This dissertation contributes to the literature in at least three ways. First, it improves our understanding of cognitive fit and how it manifests in analysts’ problem solving behaviors when using visualization tools. This is done by analyzing participants’ eye movement and gaze fixation patterns while they work with different types of data analysis tasks and visualization methods. Based on this analysis, this study proposes an objective method for assessing and measuring cognitive fit. Second, this study maps visualization characteristics to business data analysis task types, and informs the choice of visualization tools among an ever-increasing number of alternatives for supporting the complex problems faced by big data analysts. Third, this dissertation extends the cognitive fit theory to the big data context and highlights the relative importance of cognitive fit in this setting by demonstrating that increases in volume and variety amplify the task performance consequences of cognitive fit. The limitations of the experiment conducted for this dissertation and the future research opportunities they present are discussed. The findings of this dissertation also can inform the development of new visualization tools and techniques based on task and data characteristics.