Visualization recommendation systems must select appropriate visual encodings, yet few findings from visual perception are typically applied within these systems. Knowledge bases provide one way to inscribe perception guidelines, but how do we systematically translate the perception literature into a structured format? We present a literature review across 59 papers that study how to rank effective visualizations based on user performance in various visual analysis tasks. We contribute a comprehensive schema to collate existing theoretical and experimental knowledge and summarize study outcomes at three levels: between encodings, within chart types, and between chart types. We demonstrate how the resulting survey dataset can be utilized to inform automated encoding decisions with three representative visualization recommendation systems. Based on our findings, we highlight new challenges and opportunities for the community in collating visualization design knowledge for a range of visualization recommendation scenarios.
View on arXiv