Notes

Outline

Trends in Future Web Designs Mary Czerwinski Microsoft Research

Overview A Framework for Evaluating Future Web Designs Too Much Visual Information Human Capabilities User Usage Patterns Technological Trends New WWW Interaction Techniques (visualizations) Empirical Findings

Slide 3

Visualization Problems Specific to the Web Users need global and local info when searching large spaces of WWW Global---used to guide lower-level, detailed tracking of information during a query; navigation Both levels of detail cropping up in new browsers Lack of empirical evidence of benefits Do they address classic search UI problems?

New Browsers:  The Claims Techniques exploit pattern perception to enable preattentive interaction (Eick, ‘97) Cognitive capacity freed up so user can  attend to relevant info related to search Rely on focus + periphery, animated transitions, enlarged, 3D spaces and moving points of view (Card, ‘97)

New Browsers: Intelligence User tracking and intelligence can: Help predict where user will go next What content user might need to see How busy the user is Whether or not it’s a good time to interrupt Help user remember what they’ve done before Suggest and recommend useful information

Examples of New Browsers

Revealing Things--Smithsonian without Walls Uses everyday objects to tell stories about people, their cultures, ... their possessions. Can’t maintain local OR global context

Slide 9

3D Hyperbolic Space

Semantic VRML Layouts

Natrifical’s “The Brain” http://www.natrificial.com/

Data Mountain (Favorites) + Implicit Query (Czerwinski et al. 1999)

Empirical Studies of Browsers Initial attempt to track user performance across info vis techniques during web queries Performance and preference data collected “Lostness” measures and spatial abilities tracked

Slide 15

Experiments 1--Results No significant advantage for either browser Item difficult drove performance Tree hierarchy good for tracking traversal path if used systematically Hyperbolic browser best for keeping global/local info in focus; category relatedness and size Neither browser caused “lostness”, but borderline

Experiment 2---Perspecta®

Experiment 2--Results 1 or 2 clicks? Lack of control in time and space Labeling issues Too many cross-refs Confusion: headers v. titles and links Small text difficult to read RSI from flying?

Experiment 2--Results User can “sniff” around without committing Use size of category to guide search Use related topics Use popup titles and info without flying in Cool!

Experiment 3 3D Hyperbolic Browser v. Snap.com v. 2D Hierarchical Tree Control Data was not hierarchical; 40,000 nodes Looked at wide variety of search tasks: targeted search, browsing, satisficing, etc. Targets with single or multiple parents Looked at tasks where subjects had to say where targets should be placed in web design

3D Hyperbolic Browser

Experiment 3--Results In all 3 browsers, subjects used search edit field first, then manipulated UI to hone in on target 3D hyperbolic browser faster across tasks, on average 3D hyperbolic browser significantly better when known target has multiple parents Non-targeted searching harder than directed search

Exp. 4--Data Mountain + Implicit Query Results

Slide 24

Conclusion—They Are Useful… Users like new web visualizations Best for seeing local/global focus, size of category and relatedness of items Can better detect targets with multiple parents, siblings and children Search hints with little user effort

Conclusion:  Problems Remain Navigation probs persist >2000 nodes Proper structure and labeling still crucial Need to keep the interaction “lightweight” Issues with spatial reasoning abilities Need multiple options for browsing web