Worrall: Auditory display
as a tool for exploring emergent forms in exchange-trading data
Overview
of the research
The
material for this seminar is available at http://www.avatar.com.au/sonify/sonsem1/
The senses are suited to different kinds of tasks. Evolution would
suggest that that's
why we have them. The psychophysical Modal Specific Theory [Friedes D. (1974)]
recognises that that each sense has unique sensory and perceptual
qualities and distinct
patterns of transduction that are adept with different kinds of complex
information.
For example hearing is effective for monitoring sounds from all
directions, even
when the source of the sound is not visible [Cohen (1994)].
Becasue hearing is capable of a high temporal resolution, auditory
events ranging
from milliseconds to several thousand milliseconds can be distinguished
[Stuart R. (1996)]. Sound is
particularly important in situations
where the visual channel is fully utilised, which it clearly is in
trading (see later)
and when either more data needs to be displayed [(Wenzel E.M (1994)]
or in order to reduce modal fatigue.
There is growing experimental evidence that information embedded in
multivariate
multistream data, such as that used by traders, most often in realtime,
can sometimes
be better accessed cognitively when it is represented as sound than
when represented
visually either as numbers of graphically. The technique for producing
these sonic
representations as auditory displays is known, generally, as
sonification.
To date, trained users have observed patterning in market-trading data
using tools
which are primarily visual in the form of graphs and charts, or
numerical and computational.
The aim of this research is to sonify data from
trading exchanges so as to
make information embedded in the data more immediate and cognitively
accessible by
it's users. In doing so it is hoped to facilitate their decision-making
which is
based on that information.
The principle objectives of this research are to
ascertain whether, in presenting
this data in auditory displays:
1. patterns which trained users can recognise visually can be
recognised aurally,
and
2. other patterns in the data which cannot be observed visually can be
observed aurally.
Secondary objectives of this research are:
1. to enunciate new principles which are useful for the sonification of
multivariate
data sets,
2. to compile and encapsulate a set of tools to enable traders to
integrate auditory
display into their trading practice, and
3. to create a number of audio works based on the sonification of the
data sets under
study.
The three-year FTE study can be broken down into three
operational stages
of roughly equal duration:
1. Refine the topic and the establish a context for the research by
reviewing related
work by other researchers in the field,
2. Design the auditory display experiments, the auditory display system
in order
to undertake out these experiments, and then undertake those
experiments, and
3. and analyse the experimental results, create the audio works,
iterating over 2.
above as necessary. Writing up the thesis.
The Approach
I will adopt a "user-centered" approach [Preece, J., et. al. (2002)] to the development
and testing of the tools. .This process
consists of five basic steps:
1. Task analysis
2. Data Characterisation
3. Display Design
4. Display Prototyping.
5. Display Evaluation (Formative, Heuristic, Summative)
A number of peer—reviewed publications are planned,
including the presentation
of papers at the International Conference on Auditory Display (ICAD),
in a human-computer
interface journal, and an economics or econometrics journal.
This seminar is being presented after 9.5 months FTE
work, including three
major alterations to the original topic. This presentation will thus be
primarily
concerned with the first stage, which is not yet complete,
but I will also report
on progress made so far on some of the experimental work of
stage 2.
--------------
Cultural context: Pure Data (Mitchell reference?),
"Pure data, such as streams from the internet, can be sonified as part
of new
media production. Introduction of cognitively accessible points of
segmentation into
such data streams is probably essential for their impact (ref ?). In
addition, the
pure data of pre-existing recordings may be perturbed in performance to
create new
'meta-data' sound works (these range from wounded CDs, the efforts of
glitch artists
and scratch musicians, to literal or 'faulty' reading of computer
encoded data and
code), which also generate segmented impacts (ref ?)." from Nesbitt