Dear Kari and Ken,
I'd like to jump in here, to make a recommendation, highlight
a little more some identified difficulties, and suggest what
else might be useful for design histories.
1 A Recommendation
Ken: your concise review of the beginnings and current state
of affairs in Cliodynamics (PhD-Design, August 8, 2012) nicely
shows, I think, that this is not a place in which to set of a
PhD student. It's interesting, exciting even, but not yet a
securely establish and widely accepted sub-discipline. To me,
what you describe and what you pointed us to shows several
signs of it still struggling to establish itself, and make a
good name for itself. It is, for example, still dominated by
it's [modern day] father, Peter Turchin, and seems not yet to
have had much impact outside it's own community of friends and
active researchers.
On the Cliodynamics website you can find enough reasons to be
worried, including what looks to me like quite heavy
propaganda, such as this, from "Why do we need mathematical
history?" <http://cliodynamics.info/MathHist.htm>
"Without mathematics (understood broadly) we are doomed to
make vague statements and to arrive at wrong conclusions.
..."
This, and other statements like it to be found on the main
website of the research field, serve as warning signs. They
are not signs of sincere and transparent argumentation. (I've
seen this kind of thing before in a youngish research field.
I worked in, and still work in, Artificial Intelligence. It's
a sign of insecurity, not of dishonesty.)
Despite these several signs of reasons to be worried, I DO NOT
want to say there is no good in any of this work. I am not
qualified to judge this. However, I DO want to say, that,
given these and it's current state of development as a
research field, it does not, in my view, offer safe ground on
which to build a PhD, not in design research, at least.
Indeed, I would class Cliodynamics as rather treacherous
ground, for both a PhD student and his or her supervisor or
supervisors. This does not, of course, mean other more
experienced design researchers should not attempt a foray into
Cliodynamics, but they do so at their own risk.
2 A Further Highlighting
Kari: You are dead right in making your amendment "Given that
enough good quality data is cost-effectively available ...".
It applies to any kind of modelling and analysis, not just
dynamical systems analysis, but let me stay with dynamic
systems modelling, which is what Cliodynamics is a kind of.
Histories are necessarily reconstructions (in some form) of
dynamic processes: processes that change over time. There is
nothing to be gained from building a history of something that
doesn't change; in fact there is no history in them. So, all
history making is some kind of dynamic system modelling.
Usually this is of a descriptive and qualitative nature, but
still good for supporting useful and insightful analysis and
diagnosis. (Or are we to believe that history research has so
far given us nothing? Or are we to believe it's doomed
because it doesn't use mathematics?)
I think your amendment, Kari, points us to the real question,
which is NOT, do we need dynamic systems models to do better
history, it IS which kinds of dynamic systems modelling
techniques might we most productively use? The choice comes
down to the balance between the effort needed to do the
modelling well enough and the value (in terms of new knowledge
and understanding) of the research outcomes that can be
reliably and robustly obtained from them.
In the case of Cliodynamics, or any kind of quantitative
dynamic systems model building using computational techniques,
the main difficulties are not, I would say, in learning and
knowing enough of the mathematics needed, though this is not
easy. The main difficulties are in what you point us to;
knowing that you have enough data and good enough quality
data. Neither of these are easy to do in practice. Nor are
the needed verification of the model and the validation of the
model. In my experience of using (complex) dynamics systems
techniques, all of these issues are fraught with difficulties.
They need a sound understanding of measurement and measurement
theory, a transparent and tested procedure for establishing
that the numbers you take to be data are really data, and
sufficiently free of noise and other perturbations,
disturbances, and biases. Just because you have or can turn
data into numbers does not mean you have meaningful data.
Then there is the issue of verifying that the computation is a
sufficiently good implementation of the mathematics of the
dynamic system model, which involves being sure that numerical
rounding errors and such like are not introducing artifacts in
the output. Then you must validate the model, which, in the
case of Cliodynamics, because it does not have other
independent sources of data to validate against, nor the
option of synthesising known to be good clean data, must show
not only that the computed output contains dynamic patterns
found in the history data being (descriptively) modelled, but
also that it does not contain patterns that do not occur in
the history data. I see very little convincing reporting on
any of these issue in what I have looked at from Cliodynamics.
And I find this worrying.
3 A Proposal for doing something different
Here, I'd like to go back to your history of the chair
example, Kari. I think it is a useful one as it contains what
I see as some important characteristics of design histories.
Do we know when the first chair or chairs were designed? I
suppose they were crafted, so they were designed and made all
in one process, at a time when designing meant something
different from (any of) what it means today. Still, since
that possibly unrecorded origin (or was it origins?)
designers, including the craft people and carpenters who made
them, have explored the space of possible chairs. Now,
because we are talking about designed objects, not naturally
occurring or naturally formed objects, the concept of chair is
defined by the set of all designed chairs, and not by some
established and accepted external definition of what is a
chair. This means that as chair designing goes on, so the
concept of chair changes. This is a dynamic that current
mathematics of dynamic systems cannot deal with well. The
quantities used to build mathematically specified dynamic
system models must all conform to an already established
external definition and one that does not change.
Note: I am excluding things we sit on, but which were not
design to be chairs. So the convenient stone we sit on at
the cave mouth doesn't count. And I am excluding things
that are not genuinely designed to be a chair: so not things
that a designer, or anybody else, just chooses to call a
chair with no other reason than to call it something.
Now, I think it would be nice, and insightful, to have a
better idea of where this exploration of chairs has gone over
time, who took it to these places, when, where in the world
they were working when they did this, and what else was going
on and being designed at the time. This is not a kind of
evolution, at least I don't think this is the right word for
this dynamic exploration over time and space. I imagine that
certain sub-classes, design styles, will have origins located
in certain places and times, but to then see how these spread,
and influence other chair designing in different places, at
different times, by different designers and craft people,
would, I think be interesting.
To see this, not just to read about, is what I would like to
propose we might do differently: to visualise design histories
using some of the now powerful visualisation techniques we
have today. Good visualisations--which take plenty of good
designing and making, and so is also not trivial nor effort
free--help us use our brains more and better to understand the
things visualised. So, I propose that it is not dynamic
models we need for design histories--we already have these--it
is design history visualisations.
And I think some work on this kind of thing could form the
basis of a worthwhile design research PhD.
Best regards,
Tim
|