Alternatives in Visual Analytics and Computational Design

Candidate: Victor (Yingjie) Chen
Type: Doctor of Philosophy (Ph.D), School of Interactive Arts and Technology
Date: November 23, 2011
Senior Supervisor: Dr. Rob Woodbury
Thesis: Download Thesis Document

Abstract

Clearly, one's ability to explore, build, and compare alternatives can lead to better decision making, problem solving, and design outcomes. However, I find that all too often many systems are still working in a single state mode where the user can only see the result from one set of inputs at a time. Here I propose a formalism designed to handle the necessary alternatives and the space for said alternatives within the symbolic models. A symbolic model is a graphic approach to a direct constraint solver. I choose the inputs (source nodes' independent properties) as the representation of an alternative, which I have labeled variation heads. A variation head may contain one or several inputs of the model. The information carried by several variation heads can be unified to create a new variation head. Then I define the concept of the variation space as a collection of many variation heads. A variation space carries a structure of an indexed array.  Two key operations, Index Unification and Cartesian Unification, can be used to unify two or more spaces. The user defines a series of variation heads as a variation space and indexes them based on his preference, uses unification to unify the many variation spaces to create a space of the inputs for the system, and then generates a space of results based on these inputs. This research adopts design science research methodology to iteratively refine the formalism through loops of problem awareness, design, and evaluation. A prototypical system has been developed as a formative evaluation in order to confirm, explore, and expand the formalism from a purely mathematical perspective by testing out the many varied and differing kinds of data organizations. To demonstrate its usage, I describe how this formalism can be used on a specific visual analytical tool (CZsaw) in order to create a space of visualization variations; I then explain both how this formalism can be used to enrich the user's interaction in the variation space and how the indices of the space can help the user to navigate through the space.