Click for better resolution

Project: Preference Experiment.

Features:

• Comparison of different amounts of audience information in Portholes screen designs.
• Identify preferences for Portholes screen designs by first-time users.

Experiment: We explored the design space consisting of the old and the redesigned versions of Portholes with different options for visualizing the audience. We divided the design space along two dimensions (layout and audience — the two variables in the new Portholes design) with three and four variations, respectively. We had our evaluators conduct pairwise comparisons of all the permutations of the design (12 permutations in total resulting in 66 comparisons) to determine the factors that might influence their preferences. By revealing the design permutations in this manner rather than all at once, evaluators could carefully consider as to which of the two they preferred and the resulting preference judgements would be more objective in actual fact.

Analyses: The preferences were analyzed using multi-dimensional unfolding (MDU) (Shepard, 1972) to surface the evaluators reactions to the different layout and audience options. This was done in the belief that each person approaches the layouts from an individual perspective, as suggested by Coombs (1950) in formalizing unfolding techniques for a single dimension. Unfolding assumes that stimuli are aligned along one or several dimensions and that individuals observe the experimental conditions from their own perspectives. It places objects, conditions and evaluators, in a spatial model in such a way that the significant features of the data about these objects are revealed in geometrical relations among the objects. By unfolding these individual preferences and getting estimates of their position in the space, our intention was to reveal the factor(s) (layout or audience or both) and weights influencing individual preferences. Also, the technique could be used to reveal the various preference clusters that existed because we believed no individual design would be preferred by all evaluators.

Implementation: The experiment's implementation used basic Web technologies (HTML, java applet) to introduce, run, and collect data. Results of this study were incorporated into the redesign of NYNEX Portholes (known as Portholes^TNG).

Principals: Done in collaboration with Andreas Girgensohn (FXPal) and Althea Turner (Motorola).

Publication: Lee and Girgensohn, IJHCS 2002.

Work conducted: January 1997 — August 1997.