Enhancing Data Usability for People with Visual Impairments

Human-Data Interaction (HDI) focuses on how individuals engage with, analyze, and extract insights from data. For blind and visually impaired (BVI) users, interacting with data, whether searching for relevant information from structured data (e.g., web data items) or interpreting visualizations to draw insights (e.g., data charts), presents significant challenges. These challenges arise from the complexity and sheer volume of data which cannot be effectively handled by assistive technologies like screen readers and screen magnifiers. Despite its importance, data usability, the ease, efficiency, and satisfaction with which BVI individuals can interact with the data, has received less attention compared to data accessibility. To address this gap, I focus on two key aspects of BVI users’ HDI in this thesis: Data Search and Data Sense-making. For Data Search, I focus on web data items (e.g., flights, products) typically found on e-commerce websites, where information is often distributed across multiple pages, e.g., item summaries on one page and specific details of items on separate pages. While this organization minimizes visual clutter for sighted users, it imposes significant interaction overhead for BVI users, requiring cumbersome to-and-fro navigation using screen readers. To address these challenges, I first conducted mixed-method studies to investigate micro-behavioral patterns using keyboard activity and screen reader logs on both familiar and unfamiliar e-commerce websites. Based on the findings, I developed usability-enhancing solutions, including methods to automatically extract key information from multiple pages and present them ‘all-in-one-place’ for assistive-technology-tailored, convenient BVI interaction. Moreover, these solutions support natural language queries, thereby enabling efficient information access to BVI users. For Data Sense-making, I address the underexplored needs of low vision users, given that prior research has largely targeted the needs of sighted and blind users. First, I investigated the graphical perception of low vision individuals. Based on the findings, I developed two solutions: (1) transforming static, non-interactive chart images into interactive and customizable visualizations, enabling users to selectively explore data points while maintaining visual context, and (2) combining magnified views of key data points with tailored audio narrations in a multimodal interface for enhancing overall usability.