3. Introduction
• Evaluation methods and information
visualization that count errors have been
criticized in recent years
• We encounter the errors as part of exploration
and sense-making processes
• This Paper demonstrates and outline a
methodology to get insights from the error
Errors Insights
4. • There are three separable views based on
cognition theory
• Micro View-it is about the perceptive
principles and processes like pattern encoding
and cognitive integration of graphical
components
• Macro View –this view track the procedures
which consists of sense-making theories,
problem solving activities and making
judgements
• Meso View –it provides insights and
interpretation into the data and the higher
level graph comprehension.
5. Three Levels of Errors
• There processing of data can be categorized into three
• Skills based Processing-The operations at this stage
are schematic there will be not much problems for
visual analysts to find the highest values and well-
scaled visualizations like bar-chat
• Rules-based processing-In this processing the rules are
generated by heuristics
• Knowledge based processing-in this processing the
classic reasonal and problem solving activities are
done by applying analyst knowledge and mental
models under abstract analysis.
6. • Skill based processing errors-The errors are caused in
skill based processing mainly because of memory
deficits. So there is need to switch them into higher
level of processing if there is a mismatch in the schema
• Rules based processing error- cause for errors here are
two types
1. The user might have applied bad or wrong
rules
2. Good rules worked in do not fit in to the
current situation
• Knowledge based processing error-The errors might be
due to user bias or they may only selcet subset of
problems which generates wrong decisions.
7. Three Levels of insights
• Data is categorized into three types of insights
1. Skill-based insight : It is result of trivial insights
such as finding the highest value in simple line
diagram or finding the pecularitites in a pattern,
this insights are highly routinized and automatic
2. Rule-based insights: Local signs and cues are
scaned to select a rule and then implemented.
This evaluation provides the outcome.
3. Knowledge based insight: The stored stack of
problem-solving routines are work using slow
and laborious resources.
8. Interaction and Learning
• Visual analysts need to have high skilled in graphic perception and
tool handling in order to generate the complex interactions.
9. Applicability
and
Challenges
Advantage: The model
provides a integral approach
for visual exploration and
they are seperable due to
the hierarchical structure
Challenges:
There exists no generally
accepted insight taxonomy
and additionally verbal
information is required to
decide the occurrence of
insights.
The data errors from written
reports, observation and
think-aloud data may not be
errors to the true insights.
10. Case Study
Mental Difficulty vs Visual Difficulty
• Pupil size increases based on themental effort
spent on the visualization.
• Visual Difficulty is based on the number of
fixations
• High number of fixations make the eye move
involuntarily towards many things increasing
time to fixate and the visual difficulty.
• Pupil Diameter and fixation duration both
indicates cognitive effort spent on making
sense out of novel visualizations.
11. Juxtaposition view vs Comet Plot
• Juxat position view shows the nodes and
connections between them with edges
undirectional and two networks visualizations
for fall and winter.
• Comet Plot shows nodes with the-directed
edges and color encoding for fall and winter.
• The comet plot is more efficient so, it takes
more cognitive effort to understand the
visualization
12. Evaluation
• While the users given task to analyze the think
aloud data was collected from them
• The verbal information from the users are
visualized in the time series
• The Following Visualization provides
comments, pupil diameter and Fixation at
various points.
First
Insight
13. Results
• Whenever there is a insight
the size of pupil increases .
• pupil is at peak when the
user founds interesting
questions.
• Fixation is high right before
making suggestions.
14. Conclusion
• This model clearly differentiates three levels of insights
• The Meso view is an intermediate view but it is rarely used for the
novel visualizations.
• Verbal information should have lighter comments while building a
lower level model
• Visual analyst must decide prior the amount of cognitive level to
develop model
• Most of users have prior knowledge in the evaluation of visualization,
The pupil diameter may vary at different point for different users