Adam PhillippyMichael Schatz

CMSC 838SApril 4th, 2005

2D vs. 3D3D

2.1D Successes

Modest use of 3D to add highlights to 2D interfaces:

Raised/Depressed Buttons

Overlapping Windows & Shadows

Icons that resemble real-world objects

Now a standard component of desktop metaphor

3D Success Stories Natural 3D Visualizations

Medical Imagery, Architectural Drawing, Computer Assisted Design, Scientific Simulations

Continuous variables, volumes, surfaces, inside and outside, left and right, below and above are intrinsically meaningful.

Game Environments First person shooters, role playing

fantasy, virtual 3D environments Increasingly rich social contexts

based on social cognition

3D Failures

Air-Traffic Control Systems (ambiguity) Hierarchical Browsing (occlusion, navigation) Line & Bar Charts (distortion, ambiguity)

Digital Library (poor search, linking) Desktops & Workspaces (orientation) Web Browsing (screen space)

3D I

ssue

sIn

terf

ace

Issu

es

3D Ambiguity

Projective ambiguity 3D on a 2D display creates

ambiguity in all 3 dimensions

2D shadows help disambiguate x,y position

Orientation ambiguity 3D models provide limited

information Other icons may be

necessary to resolve

Information Availability

Smallman, H. S., St. John, M., Oonk, H. M., and Cowen, M. B. 2001. Information Availability in 2D and 3D Displays. IEEE Comput. Graph. Appl. 21, 5 (Sep. 2001), 51-57.

Empirical Results

Controlled experiment with 32 users performing search tasks across interfaces

Confirmed results of a prior study (orange), but that study compared across information visibility styles in addition to 2D-3D differences.

2D is clear winner when comparing with consistent information visibility (green).

Hierarchical Data

• Rooted, Directed relationships• File Systems, Organization Trees, …

• Traditional Node-link diagrams require space proportional to number of children at different levels• Overall aspect ratio grows exponentially with depth

Cone Tree & Cam Tree

“The clearest win in this technology is interactive animation. It is easy to demonstrate that animation shifts cognitive load to the

human perceptual system.”

Robertson, G. G., Card, S. K., and Mackinlay, J. D. 1993. Information visualization using 3D interactive animation. Commun. ACM 36, 4 (Apr. 1993), 57-71.

Cone Tree

Issues Occlusion Navigation Orientation

Contrast with SpaceTree

Same animation benefits

TreeMap 1,000,000 node

displays

Perspective Wall

Details are presented with overview via fisheye-like zoom for linear data

Sharp distortion at wall boundaries.

Robertson, G. G., Card, S. K., and Mackinlay, J. D. 1993. Information visualization using 3D interactive animation. Commun. ACM 36, 4 (Apr. 1993), 57-71.

XML3D

Visualize the link structure for web sites using hyperbolic zoom.

Support web content creators placing new content into existing hierarchy

Risden, K., Czerwinski, M., Munzner, T., Cook, D. An initial examination of ease of use for 2D and 3D information visualizations of Web content, International Journal of Human-Computer Studies, v.53 n.5,

p.695-714, Nov. 2000

XML3D Empirical Study

Controlled Experiment with 16 users and 4 tasks.

Measure performance relative to 2D hierarchical displays (Windows Explorer).

Statistically significant performance improvement for search tasks when category is present.

3D or not 3D Input

Mice offer only 2 degrees of freedom Output

Screens are planar User cognition

Naturally operate in 3D world Awareness, perception, reasoning, and judgment

Costs and benefits?

2D Navigation 3 degrees of freedom

1. Up / Down

2. Left / Right

3. Rotate XY

Input controls Mouse (2) Arrow keys (2)

3D Navigation 6 degrees of freedom

1. Forward / Back2. Left / Right3. Up / Down4. Pitch (transverse axis)

5. Yaw (normal axis)

6. Roll (longitudinal axis)

Input controls Mouse + arrow keys (4) Flight stick (5) 3D Mouse (6) Body Reference

Output Devices Flat monitor

Stereoscopic “3D” glasses

Relative motion Kinetic depth Motion parallax

Eye tracking Head mounted Retinal displays Holographic

Coupled Input/Output Ware and Franck

Find path of length 2 between 2 nodes 2D projection

~26% error 3D stereo with tracking

~8% error Timings roughly similar

Limited interaction Head/hand coupled Motion was effective, but

timing suffered

Summary Positives

3D information visualization has promise Eye tracking for parallax effect Stereo for depth

Missing features 3D input device for rotation

Negatives Uncomfortable for user Hardware not widely accessible

C. Ware and G. Franck, “Evaluating Stereo and Motion Cues for Visualizing Information Nets in Three Dimensions,” ACM Trans. Graphics, vol. 15, no. 2, 1996, pp. 121-139.

Spatial Memory Robertson’s Data Mountain

Leverage spatial abilities while keeping interaction simple Faster and more accurate than IE4 Favorites

Summary

Positives Leverages spatial and image memory

Users remembered their layout after several months! Simple navigation (point and click) Keeps user orientated at all times Limits occlusion and clutter Keeps text readable via pop-ups 3D audio enhances sense of depth

Missing features Auto alignment X-ray vision Dynamic filters

G. Robertson et al., “Data Mountain: Using Spatial Memory for Document Management,” Proceedings of UIST’98, 1998, ACM Press. 153-162.

2D vs. 3D Spatial Memory Where did I…

Leave Firefox? 2D window manager

Park my car? 2D (2½D) parking lot

Park my spaceship? 3D space

Which is the most effective for memory?

2D vs. 3D Spatial Memory Cockburn and McKenzie

Compare 2D vs. 2½D vs. 3D Data Mountain Both virtual and physical interfaces

2D vs. 3D Spatial Memory Users surprised by their spatial memory Subjective ratings

Preferred physical over virtual Physically least cluttered: 2D > 2½D ≈ 3D Physically quickly found pages: 2D > 2½D > 3D 3D felt “cluttered” and “inefficient”

Summary Skeptical of 3D document retrieval

As dimensionality increased Speed decreased User preference decreased

Spatial memory clearly effective But...

3D hindered retrieval, even in the physical world

A. Cockburn, B. McKenzie, “Evaluating the Effectiveness of Spatial Memory in 2D and 3D Physical and Virtual Environments,” Proc. ACM Computer-Human Interaction Conf. Human Factors in

Computing Systems, ACM Press, 2002, pp.203-210

Representation Matters

W. Ark, et al., “Representation Matters: The Effect of 3D Objects and a Spatial Metaphor in a Graphical User Interface,” Proc. Human-Computer Interaction Conf. People and Computers XIII,

Springer-Verlag, 1998, pp.209-219.

Information Visualization Success Success is often due to design features that make the interfaces

even better than reality

Interface controls are as important as the graphics display

Metrics help guide design

Usability testing is essential

“… it might be more important to fight for two versus three clicks than to debate 2D versus 3D.”

Shneiderman, B. 2003. Why Not Make Interfaces Better than 3D Reality?. IEEE Comput. Graph. Appl. 23, 6 (Nov. 2003), 12-15.

3D Guidelines

Use occlusion, shadows, perspective carefully

Minimize the number of navigation steps for users to accomplish their tasks

Keep text readable

Avoid unnecessary visual clutter, distractions, contrast-shifts and reflections

Simplify user and object movements

Organize groups of items in aligned structures to allow rapid visual search

Enable users to construct visual groups to support spatial recall

Allow teleportation, x-ray vision