Abstract
The IEEE VAST Challenge is an annual international competition that encourages innovation in visual analytics by presenting participants with realistic, complex, and data-driven problems each year. The teams which participated are challenged to design novel
analytical tools and visual solutions that transform data into meaningful insights. The 2024 Mini-Challenge 1 focuses on detecting and analyzing bias within CatchNet, a knowledge graph developed by FishEye International, a non-profit organization committed to combating illegal, unreported, and unregulated fishing activities in the Oceanus region. This challenge emphasizes the importance of understanding how bias can influence data interpretation and decision-making within large-scale analytical systems. Bias can arise from news articles, language model extractions, or human analyst edits. In this study, we developed an interactive visualization dashboard that allows analysts to explore and compare data across multiple sources, including outputs from the ShadGPT and BassLine extraction algorithms. The dashboard enables users to track how bias changes over time, visualize relationships within the knowledge graph, and identify unreliable actors. Overall, the system improves transparency, supports effective bias detection, and enhances the accuracy and reliability of CatchNet’s analytical insights.