Methods Used in Interactive Visualization Research

Data Visualization overlaps with the area of Human-Computer Interaction (HCI) which is as broad as computing and a broad as all of human society. Researchers in the area of Visualization borrow techniques from Psychology, Sociology, Linguistics, Artificial Intelligence and Computer Science, as well as many other areas of knowledge.

The range of methods of analysis that can be applied ranges from techniques borrowed from the physical sciences to techniques that come from literary criticism.

Interactive visualization research is also constructive. Often the most valuable contributions involve developing new methods for solving problems. Nevertheless, simply developing something is not usually enough. It is also necessary to convince other people that what you have done is valuable and this usually involves both analysis and testing.

Here is a partial list of the research techniques that apply in Visualization research

Evaluating the Applicability of Theory to Data vis Broblem: E.g. theory of luminance channel being good for form perception applies to color sequence design.

Comparing solution A to solution B (atheorotetcal): E.g. David Laidlaw's study of flow vis methods.

Rapid Prototyping and Spiral Design: Often the most effective systems are first built using quick and dirty methods, then evalutated and modified in a cyclical spiral design process. Sometimes the early implementation can consist of a set of sheets of paper representing interfaces screens.

Proof of Concept Prototypes: This used to be the most common style of research. In essence this meant building a system and then describing it. In so far as the people who built sustems were also the ones to describe them, the descriptions were often overly positive. "Look, this really works well". Nevertheless, a constructive solution with a clear statement of reasons why is should be superior is an important contribution..

Graphics Algorithms: Graphics algorithms are needed to transform data into display. However, often the simplest algorithms are the most effective in visualization. The algorithms are not the problem. The most effective mapping of data to display is the real problem and this is a problem of perception.

Interactive methods: The most effective techniques for exploratory data analysis involve rapidly revealing and hiding data "on demand" using techniques such as brushing and direct manipulation.

Evaluation Techniques

Psychophysics: In psychophysics, measurements of human performance are made using the methods of physical science. The methods are useful to measure the human sensitivity to light or sound, for example. In data visualizaiton the methods relate to task like judging orientation in flow fields, finding hidden features, and so on.

Cognitive Psychology: In cognitive psychology the human brain is assumed to consist of a number of processing centers that interact with one another. Experimental methods are designed to test models of these processing centers. The classic example of a cognitive theory is the division of memory into long term and short term components. Techniques from cognitive psychology are widely used in HCI. For example, In dual processing tasks we can determine how much cognitive load is placed on the user by a particular aspect of a user interface.
Sociology and Anthropology: Artifact analysis: In anthropology, the goal is often to find out how some particular object or "artifact" is use in its cultural setting. It is religeous? Is it used for something practical? etc. The point is that in order to understand the object it is necessary to place it in a cultural context. In user interfaces, the computer program is treated as an artifact that is placed in the cultural context of its intended use. We study how it functions within the office or home work or play environment.

Scenarios: In a scenario, someone invents a system or program and then writes a story about how this might be used. Often, the result looks a lot like a film script. For example, A lot of work on the early develpment of Personal Digital Assistants was done in this way. A scenario may be a useful way of getting funding to build a prototype.

Role playing: Role playing involves acting out a scenario. It is usually coupled with a considerable amount of brain storming to try to find useful extensions to the object being designed. This is especially useful with meeting support tools.

Wizard of Oz: In the Wizard of Oz method a person is made to simulate how the computer should react in some yet-to-be implemented user interface. This is especially useful in designing help systems. In essence the computer becomes human powered rather than silicon powered.

Measurement tools

The actual measurement tools used in assessment are extremely varied
Some of the more common measurements are the following:

In addition there are many highly specialized methods borrowed from various sciences.