Introduction This study examines the utilization, challenges, and design principles of data visualization approaches, focusing on their applications within AI-assisted decision-making contexts, by reviewing relevant literature. We explore the types of visualization approaches used and the challenges users face. We also examine key visual elements that influence understanding and the evaluation methods used to assess these visualizations. Methods A systematic literature review (SLR) adhering to PRISMA protocols was carried out across five major academic databases, resulting in 127 relevant studies published from 2011 to July 2024. We synthesize insights from existing visualization approaches used in decision-making, and evaluates key aspects such as usability, interactivity, accessibility, and cognitive load management. Results We identified a range of visualization forms including charts, graphs, dashboards, and interactive platforms aimed at enhancing data exploration and insight extraction. The identified challenges include achieving a balance between complexity and usability, fostering intuitive design, and providing sufficient training to aid accurate interpretation of complex data. Specific visual elements, such as color usage, symbolic representation, and data density control, are highlighted as essential for enhancing user comprehension and supporting effective decision-making. Interactive and customizable visualizations tailored to individual cognitive styles proved especially effective. We further underscore the importance of diverse evaluation methods, including usability testing, surveys, and cognitive assessments, to iteratively refine visualization approaches based on user feedback. Discussion Our findings suggest that users benefit most from customizable, interactive approaches that cater to varied cognitive preferences and incorporate continuous training to reduce interpretive biases. This research contributes to best practice development for designing accessible, effective visualization approaches suited to the complex decision-making needs in data-centric environments.
Neri et al. (Thu,) studied this question.