Measuring effectiveness of graph visualizations: ...

A cognitive load perspective

an article by Weidong Huang (National and International Surveys Program, Australian Council for Educational Research) and Peter Eades and Seok-Hee Hong (School of Information Technologies, The University of Sydney)

Abstract
Graph visualisations are typically evaluated by comparing their differences in effectiveness, measured by task performance such as response time and accuracy. Such performance-based measures have proved to be useful in their own right. There are some situations, however, where the performance measures alone may not be sensitive enough to detect differences. This limitation can be seen from the fact that the graph viewer may achieve the same level of performance by devoting different amounts of cognitive effort. In addition, it is not often that individual performance measures are consistently in favour of a particular visualisation. This makes design and evaluation difficult in choosing one visualisation over another. In an attempt to overcome the above-mentioned limitations, we measure the effectiveness of graph visualisations from a cognitive load perspective. Human memory as an information processing system and recent results from cognitive load research are reviewed first. The construct of cognitive load in the context of graph visualisation is proposed and discussed. A model of user task performance, mental effort and cognitive load is proposed thereafter to further reveal the interacting relations between these three concepts. A cognitive load measure called mental effort is introduced and this measure is further combined with traditional performance measures into a single multi-dimensional measure called visualisation efficiency. The proposed model and measurements are tested in a user study for validity. Implications of the cognitive load considerations in graph visualisation are discussed.

---