Beyond Point Solutions: Formalizing Problem and Design Spaces for Visualization

May 8, 2023
10:00am EST
Cummings 265
Speaker: Camelia Brumar - Quals Talk
Host: Remco Chang


Quals talk:

Traditional approaches to visualization design encourage bespoke solutions based on a domain user’s data and task requirements. These approaches have produced powerful tools for specific domains and have spurred the development of foundational visualization theories. However, there is a need for visualization researchers to adopt new design approaches that can lead to generalized visualization contributions. Without advancing the current practices, it will remain difficult for the community to synthesize visualization solutions into generalized theory or knowledge. In this research talk, we propose a design approach to address this critical gap. Our approach formalizes the process of visualization design as a mapping between a problem space and a design space. In our formalization, a problem space consists of both a domain-level and a visualization-level description of the user’s task and data. A well-formulated problem space is defined to be sufficiently large as to encourage generalization while having the properties of being summarizable (can be succinctly stated), comparable (can be compared with other problem spaces), and indexable (can be indexed). Conversely, a design space consists of all valid visualization designs for the given problem space. We define a well-formulated design space to be composed of orthogonal, continuous, and semantics- preserving design dimensions with the properties of being descriptive (can describe existing designs), evaluative (can assess designs), generative (can generate new designs). We illustrate through an example that our proposed formalization encourages the generalization of the visualization design process.