A Bi-directional Visualization Pipeline that Enables Visual to Parametric Interation (V2PI)

Typical data visualizations result from linear pipelines that include both characterizing data by a model or algorithm to reduce the dimension and summarize structure; and displaying the data in reduced dimensional form. Sensemaking then takes place as users observe, digest, and internalize any information displayed. The problem is that visualisations driven solely by algorithms or models may limit sensemaking because they have the potential to mask expected or known structure in the data. In this paper, we present a framework for creating data displays that rely on both mechanistic data summaries and expert judgement. In order for users to communicate their judgements, we present a new form of human-data interactions to which we refer as “Visual to Parametric Interation” (V2PI). Here, we develop both the theory and methods to create VA tools for users to adjust the parameter space while staying within the visual space. The coupled visual and parametric adjustments defines V2PI. When tools have V2PI capabilities users may not need to leave the visual space to explore data and test hypotheses. We demonstrate the benefits of V2PI in three examples.