DUST OF SNOW_BY ROBERT FROST_EDITED BY_ TANMOY MISHRA
Visualization Lecture 2005
1. 19th April 2005
Advanced Human Computer
Interaction (HCI)
Week 7
CM30141-S2
Unit Lecturer: Dr Lisa Tweedie
L.A.Tweedie@bath.ac.uk
Unit Tutor: Chris Middup
C.P.Middup@bath.ac.uk
5. 19th April
External Representations
• Reduce Cognitive Load - tool for thought
• Act as a store for our knowledge over time
• Organize and structure information for us
• However can force us to look at information in
certain ways i.e. can limit thinking. Therefore we
need to have an appropriate representation for
the external representation to be useful.
6. 19th April
Characteristics of graphics
Need the right
representation
for the type of
data and the
questions the
user wishes to
ask of it.
7. 19th April
Characteristics of graphics
With the right
representation
inferences
often become
very obvious
Jon Snow 1845
9. 19th April
Characteristics of graphics
• Finding the correct representation is still
something of a black art
– Build on representations that have be used for
a problem before
– Think about the questions that need to be
asked.
– Think about multiple views of the data
10. 19th April
Interactivity
• Adding Interactivity to representations allows a
users to proactively ask questions of the data.
• In effect an interactive visualisation allows
users to scan many hundreds of static
representations very quickly - creates a dialog
between the user and their problem.
• Encourages iterative exploration of the
problem space.
• The locus of control has switched to the user
11. 19th April
Bertin (1977)
A graphic is no longer ‘drawn
once and for all it is
“constructed”
and “reconstructed” until all
the relationships that lie
within it have been perceived.
12. 19th April
Types of Representation - Bertin 1977
• Representations of Data Values
–bottom up
• Representations of Data Structure
– top down
13. 19th April
Representations of Data values
show relations between subsets
of the data
e.g. histograms, scatterplots etc.
16. 19th April
Brushing - linking attribute views
Can take multiple similar representations of all
the attributes in a data set.
In some ways Bertins distinction disappears - as you can
see the structure of the whole set and the subset
in context.
In effect the representation provides the structure
and the interactivity provides the querying of
individual values and their relations.
24. 19th April
Linking Multiple representations of data values
It is often difficult to anticipate the
questions a user would want to ask of the
data
Different representations might be suited
for answering different questions.
Thus brushing across different
representations is a logical extension.
26. 19th April
Representations of Data structure
Show relations within an entire set
Bertin identified five types:
– Rectilinear - ordered lists, tables
– Circular - Networks
– Ordered patterns - Trees
– Unordered patterns - networks and Venn
diagrams
– Stereograms - structure suggests a
volume e.g. 3D models
27. 19th April
Representations of Data structure
Whereas representations of Data values tend to
be used for analysis - representations of data
structure are often used for providing overview
and navigation around an information space.
32. 19th April
An early tree map
• Too disorderly
– What does adjacency mean?
– Aspect ratios uncontrolled leads to lots of
skinny boxes that clutter
• Color not used appropriately
– In fact, is meaningless here
• Wrong application
– Don’t need all this to just see the largest files in
the OS
33. 19th April
An early tree map
• Too disorderly
– What does adjacency mean?
– Aspect ratios uncontrolled leads to lots of
skinny boxes that clutter
• Color not used appropriately
– In fact, is meaningless here
• Wrong application
– Don’t need all this to just see the largest files in
the OS
34. 19th April
What would make it more useful?
• Think more about the use
– Break into meaningful groups
– Fix these into a useful aspect ratio
• Use visual properties properly
– Use color to distinguish meaningfully
• Use only two colors:
– Can then distinguish one thing from another
• Provide excellent interactivity
– Access to the real data
– Makes it into a useful tool
37. 19th April
Types of interactivity
• hiding/ filtering data
• labeling e.g. brushing
• reordering
• providing information scent and other forms
of more complex labelling
• animated navigation/ algorithmic transformation
38. 19th April
Information Scent
• Relates to the issues surrounding query
interfaces
• How can a user be given appropriate cues to
move towards their desired solution in the
problem space
39. 19th April
Traditional query languages
Problems:
1. The discretionary user must learn a language. Users are often not
prepared to do this. Even for simple query languages controlled tests
(Borgman 1986) have shown that even after an hours tuition on 25%
of University Students could use the library’s online query system.
And that queries created tended to be very simple.
2. Errors are not tolerated
3. Too few or too many hits often result from queries. There is no
indication how a query might be reformulated to access fewer or
more hits.
4. There is a significant time delay between the formulation of a
query and the delivery of the result. This definitely slows the problem
solving process and probably discourages users from exploring
extensively.
42. 19th April
The Model Maker
First Order
Terms
X1
X2
X3
X4
X1
X2
X3
X4
X1 X2 X3 X4
X1X2X3
X1X2X4
X1X3X4
X2X3X4
X1 X2 X3 X4
X1
X2
X3
X4
2
2
2
2
Second Order
Terms
Third Order
Terms
43. 19th April
Other forms of scent
• Social scent - e.g. recommender systems
- This is what others feel is valuable
• History (residue) - where have I been before?
- e.g. the blue text in the world wide web.
• Boolean colour coding and user defined labels
44. 19th April
Combining automation with visualisation
Algorithms can support users in performing their
task.
Simple algorithm animations - where the user watches
an algorithm perform (e.g. data mining)
- history can then be a starting point for interactivity
- ability for user to interact directly with algorithm
Algorithmic transformations which sort and order
data creating useful metadata.
47. 19th April
Where are the killers apps?
• Technology still not quite there
• These things are hard to design well - need to
keep it simple
• Humans take a long time to develop cultures
surrounding and learn to use new
representations
• matching tasks to representations still a black
art.
• The web is probably the domain where these
tools will emerge.