Investigação da adaptação de testes de usabilidade para avaliar ferramentas de visualização de redes multivariadas

Abstract

Data visualization tools promise to enhance decision-making by offering new ways to understand the world. However, factors related to human experience and the way data is presented significantly influence users’ interaction with these systems. The visualization of multivariate networks, which involves abstract data structures containing data items, their relationships, and associated properties, is widely used across various fields. However, research on the evaluation of multivariate network visualizations has primarily focused on two usability aspects: efficiency and effectiveness in tasks related to network topology, often using fictitious networks and users without domain-specific expertise. This work presents an investigation into adapting usability testing to include network analysis tasks based on taxonomy and specific domains, with an emphasis on domain-expert participants and real data. The goal is to assess not only the visualization performance but also the usability of the tool in a more comprehensive manner. The proposed method was applied to evaluate a web tool in the field of computational pathology, which uses a multivariate network to represent renal lesions and diseases. The results indicate that the evaluation was effective in identifying interaction issues, understanding user expectations, and including new functionalities in the tool. Additionally, lessons learned throughout the process are discussed, which are valuable for improving the usability evaluation of data visualization tools in general.