The variable third dimension

The variable third dimension 

I've been rereading Gibson's The ecological approach to visual perception, enjoying his relentless debunking of many things we accept unquestioningly, mainly because everybody accepts, believes and propagates them. I've grown rather wary of statements that include "of course", especially if people listen to them and just nod in acceptance. Automatically I make a mental note to examine them more critically.
One of the concepts Gibson attacks with conviction is the third dimension in visual perception. He suggests that thinking of depth as the third dimension comes from erroneously believing that what we perceive is a two-dimensional projection, just like in a camera. He points out how unnatural it is to analyse visual perception with such projections, while in reality we constantly change our viewpoint by moving our heads and bodies. One could therefore say that once again metaphors and implementation mechanisms distort our thinking and cloud our understanding.
In architecture the third dimension is different: not depth but height. The difference actually stresses the similarity of both cases: both start from a two-dimensional projection as the reality. The floor plan is one of my favourite representations, a marvel of overview and economy. However, any design described by a floor plan is three dimensional. What the floor plan describes is just some aspects of the design. The rest should be in the designer's mind or other, complementary representations.
That's why I feel inclined to roll my eyes every time someone claims that this or that adds the third dimension to drawings or designs, that it liberates architecture from two-dimensional tyrannies. Such claims merely reveal fundamental misconceptions and lack of understanding of one's tools. That's the real tyrannies.

Sketching

Sketching

Sketches may be sacrosanct in architecture but I never managed to see them like that. Sketching may be a great way to externalise design ideas and one can learn a lot from correlating these ideas to the sketches but in the end they remain a temporary product, an intermediate state in the production of a building design. So, I fully concur with interest or even fascination in sketching but I also refuse to revere them more than the design or the building.

For years I'd been working on the automated recognition of drawings (something that deserves far more attention than it's given) and every time I presented some part of that work, there was someone who pointed out that of course these recognition approaches did not apply to sketches. Without much thought I agreed with such remarks, until one peaceful summer afternoon in the garden, when I was doodling on a piece of paper. Suddenly, the inquisitive researcher in me finally woke up and asked the obvious question: it is so? Are sketches that different from drawings?

I've spent some time on that question and the brief answer is that from a representational point of view there may be fewer differences and more similarities than assumed. Sketches are fuzzier and compound but at the paradigmatic level (the level of symbols and primitives), they often come close to drawings. I just hope that I'll be able to test this hypothesis at a large scale in the future.

The moral of the story is not about sketching or representation; it's about insisting on questioning established prejudices and conventions. It may seem silly, fractious and obstinate at times but if done with real curiosity and interest in learning, it can be a source of great pleasure to the researcher. Why do research for anything less?

Solids & voids

Solids & voids

It's quite uncommon to see building elements and spaces represented in the same way. Most of the time, the one is implicit in the other (as with spaces in conventional architectural drawings) or is a product of the other (as with spaces in BIM). Usually, its the spaces that are implicit or derivative. In architecture, we may shape space but in fact we build solid elements; what is left over within the volume of a building is the 'voids' (a bad name because spaces contain quite a lot). A figure-ground reversal is seldom applicable to the duality of solids and voids in architecture.

This is one of the reasons why I'm interested in alternative representations like graphs. Using graphs one can describe both building elements and spaces in the same way, si that their complementary character becomes apparent. After that, it's possible to see different patterns emerge, patterns that describe and possibly explain stuff one's only vaguely aware of.

Theory and representation

Theory and representation 

BIM may seem to stimulate progress but in some respects it restricts progress. In the bad old days of CAD, no-one dared suggest that CAD was anything more than a technology. This created room for theoretical and methodological development as a complement, as the underlying reasons for using computers in architecture or at least as an intelligent approach to an enabling technology. The old diptych of design and drawing was interpreted as theory and representation, allowing room for exploration that did not necessarily fit the priorities or limitations of the technology.

By being both a technology and a methodology, BIM restricts this room to within its own framework. It forces one to work with a representation that often disappoints and requires much loyalty concerning key methodical aspects. In this sense, BIM is an old-fashioned technology and approach, a largely closed (proscriptive) system. It doesn't help that the current implementations are lacking in many respects, for example the central issue of shared models: to be able to collaborate online, people have to make many concessions concerning the size of files etc.

This brings us to an interesting situation: the currently dominant technology, which theoretically promotes inclusive integration and appears progressive, seems to exclude further development and variation. Of course, some would argue that add-ons, especially parametric programming tools, allow us to solve many problems but that's not the best way forward, especially concerning representation. If BIM has given us a usable theory of design, we now need a better theory of representation.

Symbolic representation

Symbolic representation 

I've heard it more than once: "So, what's new about BIM? We've seen it all before, with CAD libraries and the like." People who say such things are not only old, usually in their sixties, but also typically people with little if any hands-on experience about what they're talking. They're not people who spend time drawing, modelling or designing with computers, otherwise they might have noticed a fundamental difference between CAD and BIM: the level of symbols used.

In CAD, the basic primitives one interacts with, the ones that carry the essential information for the representation are graphic elements: lines, shapes, surfaces etc. In that CAD follows established traditions in architectural drawings, which rely on convention to support recognition. For example, two parallel lines close to each other and with a certain length must be a wall. So, in CAD we draw these two lines with vector graphics rather than ink on paper and if we want be sure that there'll be no misinterpretations, we group them together and possibly also label them as a wall. This changes little to the basis of the representation: it's two lines.

In BIM, one may still draw lines but works directly with symbols for architectural entities: one draws a line to indicate the axis of the wall but only to enter certain geometric properties. The type of the wall, its width and layers, are defined by type and not by the other lines. The appearance of the wall may be two parallel lines but that's just one of the many possible ways the wall may be depicted. What the representation knows is that there is a wall, not two or more lines.

This doesn't make BIM better than CAD, it just brings architectural representation closer to the structure of other digital kinds. Texts, for example, have been symbolic at the level of characters from the beginning - not complex pen strokes like in joined-up writing. When I type a letter on the keyboard, what the computer retains is an ASCII code for that character, possibly dressed up with a font, size etc. Just like a wall in BIM, this dressing up determines the appearance of the character; the character is not recognised on the basis of  its appearance.

The difference between BIM and CAD is therefore one of symbol level: BIM uses symbols for architectural entities, while CAD uses general graphic symbols, those belonging to the implementation level rather than the representation proper. There may be many things wrong with BIM but that it finally gives us a symbolic digital representation for architecture is significant.

The recurring fascination with generative systems

The recurring fascination with generative systems 

Soon after I had entered the area that used to be called CAAD (Computer-Aided Architectural Design), I realised it was motivated by two quite distinct and actually conflicting ambitions: to computerise drawing and to automate design. The design automators were in power back then, leading the area with creeds like "intelligent design versus stupid drawing". I didn't agree; I have always been fond of drawing and fascinated by visual representations in general.

Still, when the democratisation of computing meant that computerised drawing became commonplace and hence dominant, I was less than pleased for the basic reason that much in computerised drawing was not about new, intelligent representations but about producing the same old stuff on paper, only seemingly more efficiently (people think that efficiency is merely the ability to change easily the content of digital documents). What followed is what one can see in architectural education and practice today: an overemphasis on visualisation, combining poorly readable line drawings with lavishly rendered perspectives, meaningless or jittery walkthroughs and often no time to study all those images for the benefit of the design.

The funny thing is that design automation persists and is resurrected by every other generation of students and your researchers, who somehow inherit or more frequently rediscover the appeal of the same techniques for generating patterns one can pass off as schematic designs. Among them, those relating to space allocation never fail to come up: one makes a list of rooms, determines some clustering of these rooms, uses some elementary way of arranging them in a floor plan, usually without the benefit of architectural knowledge (e.g. spatial typology), makes some manual adjustments to make the layout look less like the product of a child playing the architect, and calls it a design. It's amazing how easily people believe in the same old, tired techniques. It could make one believe that they have some inherent power but I suspect that they're simply easy enough to be rediscovered again and again. That they lead to nothing and are forgotten for a generation or so is a truer indication of their power.

3D

3D

When most people talk about 3D, it's about depth perception, about combining two pictures of a scene (as we do with our two eyes) to actually see rather than infer the distance and relative position of objects. In architecture, when we talk about 3D, it's about the height of objects, the Z axis in our representations. This is arguably typical of architecture: we don't look at the world, we look at our own tools and conventions. Perhaps the saddest thing is that we don't appear to celebrate the ingenuity of these tools: 2D representations aren't boring, stupid means to be replaced by 3D representations as soon as possible; at least some of them are great ways of describing aspects of the real world, making them accessible and manageable.

No, I'm wrong: the saddest thing was that many years ago, when I was pushing for 3D and making people work towards it by putting floor plans on top of each other and cross sections at the right places, discrepancies of even 50 cm between floors were not uncommon. Of course, at the same time, when we put floor plans of adjacent buildings next to each other we discovered gaps of up to fifteen degrees. That's the saddest: not being able to use one's tools properly - and this won't improve with 3D.

Conceptual design

Conceptual design 

The purpose of a conceptual design is to provide evidence for the solution of a problem and to show what distinguishes this particular solution from the others. It is therefore important to know which features or other information should be present in a conceptual design to perform this dual task.

Firstly, we need to know what the problem entails, how it is composed from the goals and requirements of the brief, the constraints of the location, the challenges designers undertake (aesthetic, typologic, structural etc.). This calls for an inclusive approach that does not prioritise or neglect.

Secondly, we need some objective criteria for describing and evaluating design claims: a basis for analysing features of the design and linking them not only to behaviour and performance but also to similar designs in comparisons that explain. Typology used to promise a lot in this respect.

So, in conclusion, it may be time for different conceptual design representations, less elliptical in terms of information and more robust and analytical than the sketches and diagrams we have been using for too long now.