Story
Idea 1: Collaborations between Science and Art.
- Goal: The creation of
tools for (identifying|highlighting|experiencing) dynamic
(patterns|characteristics) of data.
- Advances in data collection technologies has led to the creation
of huge datasets (Internet traffic, global transportation networks,
the telephone system)
- Often, these datasets record our comings and goings, in both the
electronic and the physical worlds;
- Traditional approaches to experiencing these data break down
because of their size and complexity;
- Why?
- We need new tools to experience large, dynamic data sets.
- As these data often involve human activity, there is a social
component to experiencing the data.
- Why sound?
- Sound can more easily express dynamic characteristics in data.
- Sound is an intrinsically ambient medium that we routinely use
for information gathering. We collect information about the physical
world through sound; the same can be true of the digital world.
(examples: car engine, wind direction, building occupancy)
- Sound is immersive, you don't have to be looking at a screen to
get it.
- What is the end product?
- We can't really speculate on future ``products'' in a formal
sense. This is forward-looking research.
- The math center and the statistics research group in particular
has a long history of contributing to a corporate ``bottom line''
through really fundamental research.
- Statistics is about data, about learning from data. The
telecommunications industry has produced some of the largest, most
complex data sets around.
- Anything we can do to enhance our experience of a data set,
making it richer or more expressive, offers the possibility of forming
new insights. This was true of the fundamental work on visualization
that came out of our group, and I am confident the same will be true
of this work.
- Having said that, there are naturally application areas that are
helping to inform our work. For example, we have already considered
Web site traffic; we will get data on network operations either in the
context of a network operations center or observations of traffic
patterns across the Internet. We have also plan to give the Java
virtual machine a sound, the classes and methods.
- This is, of course, forward-looking research. we have a lot to
learn in terms of what makes an informative audio representation; what
are the rules that govern how we perceive patterns in music or simply
sound? what are the principles that govern how best to amplify the
important features in data?
- Of course it is still early, but i think this work can be as
important as the early studies on data visualization that took place
at the labs.
- What?
- Passive monitoring of complex systems:
Network operations center collects messages from various
points in the network. Understading the patterns in these messages is
difficult. An audio representation can help controllers understand the
natural fluctuations in these data and more immediately alert them to
problems.
- Uncover patterns in web site usage:
- We can help web masters and content providers better
ascertain the characteristics of their traffic. A rich audio
experience can provide clues to behaviors that are difficult to
uncover through other methods.
-
Very few of us realize that there
are often tens of thousands of people all reading material from the
same web site.
- Provide users of these new technologies with more insight into
how they function and how others use them.
-
Realize your contribution to the traffic?
- See your
impact on the web site?
- Why would artists and scientists
collaborate?
What we are really talking about is a new way to communicate
information. Part of this must involve design and aesthetic concerns;
part of it will require new mathematical models and methods.
- Why involve a statistician?
- Statistics is about data. It is about learning from data,
extracting useful information from data.
- Models or mathematical representations for data are at the heart
of statistics. Models give us a way to identify important
characteristics in a data set, and can offer clues as to how we might
amplify them, draw them out, in an audio representation. Sometimes, it
might be important to simply listen to the information left over in a
data set after we have removed the effects of a model.
- Many music theorists suggest that our experience of music has
much to do with our expectation of what will come next, our prediction
of the next note or phrase. Utilizing these kinds of ``predictive
inferences'' is an intrinsically statistical exercise.
- Why involve an artist?
- An artist is not only interested in understanding or factually
communicating data, but also conveying aspects of the wonder and
amazement that these data exist at all; this is really a provocation,
it is about surprising people; it is about creating a concrete point
of contact with the data itself, so that the data become one step less
a total abstraction to people. It is putting people face to face
with this technological mirror
- The audio must be aesthetic for people to fully grasp the
character of the data being represented;
- Data visualization has benefitted tremendously from good design
principles that have essentially evolved from a basic understanding
about how people consume graphical displays.
- Future work
Web sites, NOC, architectural applications
- Preliminary experiments
- Like many corporate web sites, www.lucent.com is an information
source. The content is arranged heirarchically within the site. Top
level areas describe the various business units within Lucent; and as
you dig deeper into each, you find product information, etc.
- Our
initial audio representation expresses the depth that visitors are
reaching into the site.
- Accesses to high-level information contributed to overall
background drones; one tone per area we considered.
- More detailed information is captured by pulses that grow faster
and louder in proportion to the number of people at that depth.
- Statistics
Research: Comments
- Statistics is an empirical science. It is a mathematical
discipline, we build probabilistic models of the world.
- As a field, statistics is enriched by the applications we
consider. Each new problem provides us with new insights about how to
learn from data. The field advances by generalizing the lessons from
particular applications, by constructing new tools that can be used on
other problems. The newest challenges come from large complex problems
that involve massive amounts of data. Hot application areas include
data and telecommunication networks.
- So good statistics research is a balance between practical
applications, theoretical investigations and methodological
advances...tool building.
- ... and that's why i think bell labs is such a great place. the
telecommunications industry has historically been an excellent source
of challenging problems and complex data sets. Collaborations between
different areas in the labs are encouraged. A lot of really good
statistics research depends on borrowing tools from other branches of
mathematics, computer science, signal processing...even information
and coding theory.
- ... and in fact, this project is in some sense a good example of
that. the use of sound seems to offer us a new way to experience
complex, dynamic data. An experience that is difficult to achieve in
large systems through visualization alone.
- Alternative: Statistics is a rigorous science with a
heavy and healthy empirical component. As a mathematical field, it is
based on the modern probabilitistic, rather than the conventional
deterministic, thinking; It is enriched by the problems we solve.
These problems historically have come from gambling, insurance,
official census, genetics, industry quality control etc and now come
from fields as diverse as finance, telecommunications, public policy,
sociology, public health, and human genome project. It is actually
very hard to find a field where statistics is not useful. Anyone with
a good statistical training is a hot commodity in the market place and
will continue to be that way for a long time to come.
|