Information Visualization• “Ant-vision is humanity’s usual fate;
but seeing the whole is every thinking person’s aspiration” - David Gelernter
• “Visualization … transforms the symbolic into the geometric, enabling researchers to observe their simulations and computations” - McCormick et al
Visualizations in the Periphery
Questions• Do peripheral displays interfere with
other daily tasks?• What are the relative advantages of
different classes of peripheral displays?
• How does changing visual features of the display affect information processing and distraction?
Experimental Studies
Implications• Animated displays can be used without
negatively impacting other tasks • In-place animations preferable for
identifying changes, while motion-based are best for processing and remembering
• Understanding of peripheral animations affected by the size of the display
Results• Animated displays can be used without
negatively impacting certain other tasks • In-place animations better for monitoring• Motion-based animations better for
awareness questions• Small displays better for monitoring• Fast displays better for awareness• Balance productivity and preferences
What is Pre-attentive Processing?
•Refers to cognitive operations that can be performed prior to focusing attention on any particular region of an image
•Estimation based on viewing displays for <200-250 ms (qualifies as pre-attentive)
Pre-attentive Processing• Target has a unique feature (color) from the
distracters.Hence it can be detected pre-attentively.
Pre-attentive Processing• Conjunction of features: No unique feature distinct
from its distracters. Hence difficult to detect.
• http://www.csc.ncsu.edu/faculty/healey/PP/PP.html
Pre-attentive Processing• Example: Hue Vs Form• Feature Interference: Variation of form did
not interfere with hue segregation.Varying hue within a display region interfered with boundary detection based on form.
Emergent Features• Target in (a) contains no unique feature.• Target in (b) contains non-closure as unique
feature.
2D or 3D?• When the cubes appear "three-dimensional",
the 2x2 group with a different orientation is pre-attentively detected
• When three-dimensional cubes are removed, the unique 2x2 group cannot be pre-attentively detected.
Iconographic Displays• Lots of attributes.• Can’t detect the target
• Alternate design
Other Features• Motion and Depth• Texture for multi-dimensional data
Similarity Theory• As T-N similarity increases, search efficiency
decreases and search time increases • As N-N similarity decreases, search efficiency
decreases and search time increases • Decreasing N-N similarity has little effect if T-N
similarity is low; increasing T-N similarity has little effect if N-N similarity is high
Experiment
Goals:
• Can results from the experiment be used in building visualization tools.
• Can these results be extended for rapid and accurate numerical estimation?
• How changes in display duration and perceived feature difference influence a user’s ability to perform numerical estimate.
Experiment data• Salmon Migration
Numerical EstimationTask-relevant data: Landfall Task-irrelevant data: Stream
FunctionFigure (a) Figure (b)
• Color: Landfall Color: Stream Function• Orientation: Stream Function Orientation: Landfall • Target: Blue colored elements Target: 60 orientation
elements
Numerical EstimationTask-relevant data: Stream Function Task-irrelevant data:
LandfallFigure (c) Figure (d)
• Color: Stream Function Color: Landfall• Orientation: Landfall Orientation: Stream Function • Target: Blue colored elements Target: 60 orientation
elements
Numerical EstimationTask-relevant data: Landfall Constant data: Stream FunctionFigure (a) Figure (b)• Color: Landfall Color: Landfall• 0 Orientation: Stream Function 60 Orientation: Stream
Function• Target: Blue colored elements Target: Blue colored
elements
Numerical EstimateTask-relevant data: Landfall Constant data: Stream Function Figure (a) Figure (b)• Orientation: Landfall Orientation : Landfall• Red Hue: Stream Function Blue Hue: Stream Function• Target: 0 degree orientation Target: 0 degree orientation
Numerical EstimationResults:• Rapid and accurate numerical estimation can
be performed with display time: 450 ms• Accurate results with Stream function as task-
relevant data item suggesting preference for spatial arrangement of elements
• Accuracy didn’t differ between constant and variable trials during either hue or orientation estimate hence no feature interference in this task
Display DurationMethod:• Display duration was randomly varied among 5
values: 15,45,105,195,450 msResults:• Estimation Accuracy stable at all durations of
105 ms and above. Below 105 ms, error values increased rapidly
• Higher Average error from 15 and 45 ms display duration trials
• Minimum display duration using either hue or orientation lay between 45 and 105 ms
Feature DifferenceMethod:• 2 hues: 5R 7/8,5RP7/8 2 orientations: 0 and 5 degrees• 2 hues: 5R 7/8,10P 7/8 2 orientations: 0 and 15
degrees• 2 hues: 5R 7/8,5PB 7/8 2 orientations: 0 and 60 degreesDisplay duration: 45 ms and 195 msResults:Accurate hue estimate at 195 ms with 10P 7/8,5PB 7/8 and
at 45 ms with 5PB 7/8Accurate orientation estimate at 195 ms with 15 and 60 degrees and at 45 ms with 60 degreesNo feature interference
Conclusions
• Rapid and accurate numerical estimation.• Similar high levels of accuracy were observed
down to 105 ms• For >= 200 ms, target elements could be
reasonably similar to distracter elements while still allowing for accurate estimation
Discusser• Why use Pre-attentive Processing???
• Scalability (Is it really scalable??)
• Linear separation
• Future Enhancements
Infoviz in the Periphery An Evaluation of Information
Visualization in Attention-Limited Environments
Jacob Somervell, D. Scott McCrickard, Chris North, Maulik Shukla
Department of Computer ScienceVirginia Polytechnic Institute and State UniversityPresenter: Ndiwalana Ali
Discusser: Colaso Vikrant
CS5984: Information Visualization
Motivation– People do not always use information
visualizations as their sole or primary task.– How could information visualizations intended
for multiple-task situations be designed?– It is suspected that such visualizations are
distracting, but little is known about the degree to which it distracts users and whether users can overcome these distractions and interpret the peripheral visualizations.
– Peripheral visualization vs Standalone visualization
Overview– Information visualizations as secondary displays
(peripheral visualizations)– How quickly and effectively can people interpret
information visualizations (Secondary) while busily performing other tasks (Primary)?
– How can peripheral visualizations be designed to reduce distraction while maintaining awareness?
– Factors that might impact performance evaluated:
• Visual density• Visualization presence time• Secondary task type
Experimental Design – A 2x2x2 (time x density x task type) design– 28 undergraduate students from a cs class
participated in the experiment for class credit, 6 rounds each.
– Dual-task setting • Primary task – a video game • Secondary task – answer a question about info in a viz
that appeared while you played game– The experiment included three independent
variables:• Time 1 or 8 seconds time visualization was present• Density low or high low=20 objects, high=320• Question single or cluster find single or a cluster
Experimental Design (cont)– Each round started with the presentation of the question that
the participant would answer using the visualization.– The question was then removed and participants then played
the game. After 15 seconds of playing
the game, the visualization appeared on the screen. Incorporated in the visualization was the answer to the target question. The visualization remained visible for either one or eight seconds, depending on the test group, and then disappeared.
Experimental Design (cont) Participants then played the game for an
additional 10 seconds. The target question then reappeared along with 4 multiple choice answers.
4 test groups Group 1 – time of 1s, high then low density Group 2 – time of 1s, low then high density Group 3 – time of 8s, high then low density Group 4 – time of 8s, low then high density The initial task was chosen by the toss of a coin while device
ordering was counterbalanced
Low density
High density
Experimental Design (cont)
– Measure of primary task performance• percent of blocks caught both for the
time before the visualization appeared and for the time period after it appeared (including while it was visible)
– Measure of secondary task performance• correctness rate for answering questions
Results Performance (%)
– In the one second conditions, low density visualizations yielded better performance than high density visualizations. No effect on performance for density in the eight second condition.
– In the one second conditions, there was higher performance when locating a single object compared to locating a cluster. No effect on performance for question type in the eight second condition.
Average performance (after viz appeared) for 1 second conditions, based on high vs low density and single vs clustered question type.
ResultsCorrectness (%)
– As expected, those in the eight second condition answered more questions correctly than those in the one second condition.
– More people answered more questions correctly with low density visualizations than with high density visualizations in both the 1 & 8 second conditions.
– More people answered more correctly on “find cluster” questions than “find single item” questions in both the 1 & 8 second conditions.
Secondary task correctness based on viz density and question type.
Conclusions– Peripheral visualizations can be introduced without
hindering primary task performance.– Interpreting complex visualizations within 1 second in
a dual-task scenario can not be done effectively but with relaxed time constraints can be effective.
– Lower density displays can result in performance that is as good or better than high density displays in a dual-task scenario.
– Finding clusters of visually similar items is easier than locating a single item.
– Vikrant discussion– Expound on the last sheet and conclusions!