An Examination of Analytic Focus and Perception during Exploratory Visual Analysis

Visual analytics technologies utilize visualization tools to facilitate exploratory analytical thinking, allowing users to quickly identify new patterns and trends in complex data and develop hypotheses. However, this capability comes with critical risks, as people’s judgments and decisions may be based on imperfect approximations and heuristics. This can potentially lead to perceptual distortions and inaccurate conclusions. Furthermore, seeking insights from complex or large-scale data is challenging and it’s hard to efficiently identify relevant information. To address these challenges, this thesis presents a novel approach for visual analytic platforms to dynamically model an analyst's analytical focus and surface relevant information to mitigate the risks. First, we present a formal model to represent a user's analytic focus during data analysis and a prototype application to examine its accuracy and effectiveness. Next, we conduct a user study to investigate priming and anchoring effects within multi-view scatterplot visualizations, focusing on the specific task of cluster separation. This study aims to identify potential perceptual biases that could impact the accurate presentation of the analytic focus. Furthermore, we perform a systematic review of the literature to identify the state-of-the-art in how visualization systems surface recommended content during visual analysis and propose a design space for visual content recommendation interfaces. This thesis contributes novel visual analytics techniques that dynamically and accurately capture users' real-time focus and provide design guidelines for visualization systems to support more informed decision-making through effective content recommendation.