Dear Ben, Tom, and Alan,

I hope I did not bring genetic algorithms across as something designers need
to be afraid of. My point was just that it is possible to generate with
configurational rules instead of bottom-up rules. For us, their purpose is
more to give you examples of what a set of parameters CAN mean in terms of
spatial layout, than to tell you what it MUST mean. This is of interest to
check assumptions about what morphological features syntactic parameters are
related to. If you analyse examples, you can find correlations, but you can
not necessarily tell whether the parameters you examine are constitutive for
the morphology, or whether they could also take on completely different
forms, depending on other factors.

In fact, I see this automatic generation relatively removed from a design
tool. What you propose, Tom, sounds better - actually, to integrate Space
Syntax derived visualisation in CAD is what we would like to go for as a
next step. Maybe, though, it's a bit rushing forward to ask for an index of
attractiveness to sit out on a sunny day. What would be interesting is an
in-between step, a visualisation that allows you to follow what changes of
intervisibility, spatial integration, etc, your changes in a plan produce. 

What effects these have on complex concepts like pleasantness is something
that, at this stage, is a matter of your interpretation, although it may be
informed interpretation. The strength of VGA and Space Syntax is that they
make it possible to visualise properties of space that are not
interpretation, but quite objectively there. The not-so-objective (or maybe
more the empirical) part is with what functions or likely human behaviours
you associate these properties.

I agree that a visualisation of these may enter the risk of being seen as
something more objective than it can be. However it would be a nice tool to
try out different mappings of relationships (maybe they don't have to be
fixed but could be set while visualising with all sorts of sliders etc..).
I think this could start at the more fundamental level of syntactic
properties visualised parallely to drawing. This would not tell you what to
do really, but it could help link your designer's intuition about design
interventions and what you think they will cause in terms of use, to what is
going on at the syntactic level of space - in order to inform your
interpretation of Syntax parameters, in an additional way to empirical
research.

By the way, I do not know if you had a chance to see the paper by Felix Kabo
for SSS5 on architectural programming, it might be of interest here. His
approach, as far as I can give an account of it, is one that tries to inform
you about the syntactical properties of interconnections in your design
while you program it. How these properties are evaluated is not fixed; in
his example they are rather used to compare the design to empirically found
examples.

Regards,


Joerg

> -----Original Message-----
> From: [log in to unmask] [mailto:[log in to unmask]] On
> Behalf Of Alan Penn
> Sent: Thursday, July 28, 2005 3:53 PM
> To: [log in to unmask]
> Subject: Re: I need advice please
> 
> I agree with all of this. Visualisation is a key component of both science
> and design, and if I am pushed I say that this is one of the key ways in
> which syntax analysis relates to design practice.
> 
> One way of thinking abut this is via Ian Hacking's description of the role
> of 'creating phenomena' in science - by which he meant the use of
> representations or experiments to 'show new things' that then become the
> subject for theorising. Another is via Donald Schoen's view of sketching
> in
> architectural design as a part of 'reflection in action' - where the
> designer uses sketches and sketch models so that they have something to
> think with. In both cases we work with visualisations and representations
> as
> a part of a kind of 'externalised cognitive equipment' to support our
> creative activity. The fact that we 'externalise' in this way also opens
> up
> the possibility of social discourse - representations and visualisations
> become the thing around which discussion can take place.
> 
> Having said that, the mapping between an analysis - which can be
> visualised
> - and some kind of human behavioural outcome, is only as good as its
> empirical support is strong. For 'simple' behaviours - how populations
> move
> through spatial systems on average - this may be relatively
> straightforward.
> For more complex behaviours (eg. Where they choose to stop, where crime
> takes place etc.) this can become difficult and so hedged about by caveats
> that its use in design may be limited. The risk in automating this kind of
> visualisation to my mind is that it can take on a status beyond its
> empirical support - just because it looks good and looks 'scientific'. It
> would be rather nice if there was a way of visualising both the prediction
> and its level of support at the same time - perhaps fog could descend and
> obscure poorly substantiated results?
> 
> A final point. The idea of working back from a desired result to the
> morphology of a design that would produce it is far from simple. This kind
> of thing is being done in the design of turbine blades to achieve specific
> air flow characteristics - however in comparison to human systems these
> are
> remarkably simple and well understood systems. And even for fluid dynamics
> this kind of thing is right on the edge of what is achievable. I suspect
> that human designers are safe from genetic algorithms for some time yet.
> 
> Alan
> 
> 
> > -----Original Message-----
> > From: [log in to unmask] [mailto:[log in to unmask]] On
> > Behalf Of tom lists
> > Sent: 28 July 2005 13:34
> > To: [log in to unmask]
> > Subject: Re: I need advice please
> >
> > Dear Jörg
> >
> > Your genetic algorithm sounds interesting – although I
> > understand what you say about the many other
> > constraints on design.  I sometimes wonder if there is
> > any choice at all after the regulations have been met.
> >
> >
> > This is why it sounds good to have a tool to provide a
> > visual interpretation of morphology with respect to
> > interaction.  I can imagine this might be like
> > parametric design in CAD, where physical boundaries
> > are allowed a range of dimensions within which they
> > can vary, and related elements move in response
> > according to rules.  If this were linked to
> > predictions about human interaction, perhaps using VGA
> > as you say, it could give responsive feedback to the
> > designer.
> >
> > Alternatively, there might be a representation half
> > way between the graph and the plan, laying out the
> > sizes and relationships between spaces without fully
> > defining boundaries.  This might help in the earlier
> > stages of design – I presented a possible outline for
> > such a process to SSS4 two years ago.
> >
> > Even then, the question is, ‘What have you actually
> > visualized?’  With VGA it would be fair to say that
> > you can visualize the pattern of co-visibility, but
> > what does this tell us about human behaviour?  I know
> > there are 'correlations’, but does it really show you
> > which pieces of space constitute a place which will
> > attract people to sit out on a sunny day? And which
> > will be safe from crime through natural surveillance?
> >  What are the behavioural parameters controlled by
> > morphology, which directly link to parametric
> > boundaries?
> >
> > It would be nice to imagine dragging a boundary line
> > on CAD and seeing colours change indicating changes in
> > security surveillance, habitability for gathering,
> > busy-ness of routes-through, etc.. . .  Perhaps Ben
> > will develop something like this?
> >
> > Regards,  Tom
> >
> > Thomas Everest-Dine
> > Architect,  London
> > [log in to unmask]
> >

--
Joerg Kraemer, Jan-Oliver Kunze
Technische Universitaet Berlin
Ackerstrasse 71-76, Raum 438, 13355 Berlin, Germany +49 30 31472748
[log in to unmask], [log in to unmask]