Turtles all the way down

A well-known scientist (some say it was Bertrand Russell) once gave a public lecture on astronomy. He described how the earth orbits around the sun and how the sun, in turn, orbits around the center of a vast collection of stars called our galaxy. At the end of the lecture, a little old lady at the back of the room got up and said: "What you have told us is rubbish. The world is really a flat plate supported on the back of a giant tortoise." The scientist gave a superior smile before replying, "What is the tortoise standing on?" "You're very clever, young man, very clever," said the old lady. "But it's turtles all the way down!" (Courtesy: Wikipedia)

This issue is one version of what is known as the problem of infinite regress and it occurs in many modeling problems.

Here is the formal definition of the issue of infinite regress (again from Wikipedia)

An infinite regress in a series of propositions arises if the truth of proposition P1 requires the support of proposition P2, and for any proposition in the series Pn, the truth of Pn requires the support of the truth of Pn+1. There would never be adequate support for P1, because the infinite sequence needed to provide such support could not be completed.

Distinction is made between infinite regresses that are "vicious" and those that are not. One definition given is that a vicious regress is "an attempt to solve a problem which re-introduced the same problem in the proposed solution. If one continues along the same lines, the initial problem will recur infinitely and will never be solved. Not all regresses, however, are vicious."

Let me simplify the above and put it in terms of an architectural modeling problem and here is a thought experiment for you. Imagine a line drawn somewhere and a man-in-white-coat (MIWC) standing on that line. (The reason for the "man in white coat" is because it is often a stereotype of a knowledgeable man in B-grade Hollywood movies) Now our MIWC has this capability of understanding reality, which is on the right hand side (RHS) of the line and he can place equivalent elements of that reality in a model which is slowly being constructed on the left hand side (LHS) of the line. Now let us make our MIWC do some activities for us. On the RHS we place a simple building and we request the MIWC to model it for us. The MIWC is now all excited because he just so happens to be an expert on building information models.

He takes one look at the building and says: "I can see four walls. So now, I'll place the modeling equivalent of those four walls on the LHS"

Then I say: "But, one of those walls contain a door opening"

MIWC, "Don't you worry. Here I am now placing the door opening in the model appearing on the LHS"

Then I say, "But the opening contains a door shutter"

MIWC: "Of course it does. I can place the model of the shutter into the modeled door opening on the LHS"

Now this goes on for quite a while, as our MIWC literally hauls the reality from the RHS into the LHS. Let us make a big assumption here and be very gracious to our MIWC. Let us assume that he has successfully modeled everything from the RHS and placed it on t LHS (i.e. the model is now complete) At this point, the MIWC steps away from the line and is so pleased with himself; he says: "There you go, now the model is complete"

And at this point I ask him to look back and see the Frankenstein he has created on the LHS. When he turns to see it, he finds that the model is now so accurate it has captured all the problems and confusions we had in understanding the reality which was there on the RHS. Why so? because the problems and confusions are also part of the reality and in the process of modeling everything our MIWC brought even those over to the LHS. In effect, by making the model so accurate; he had inadvertently destroyed its usability. The problem of the infinite turtles has come and bit him on the backside.

Yes: I know what you are thinking. When the MIWC is working, he will NOT be modeling everything from the RHS and be placing it on the LHS. So he cannot get into an exact duplication of the reality. To this, I ask: Who decides what is NOT to be modeled is irrelevant? Suppose I was indeed wanting the representation of that particular thingamajig the MIWC left behind on the RHS? What do I do then?

Reductionism
I played a small trick in the above thought experiment. I outlined only one of many possible approaches to modeling; albeit the approach I took is what is often taken in science. Such an approach of transferring model elements from the RHS to the LHS in smaller and smaller parts is called Reductionsim.

"Reductionism can either mean (a) an approach to understanding the nature of complex things by reducing them to the interactions of their parts, or to simpler or more fundamental things or (b) a philosophical position that a complex system is nothing but the sum of its parts, and that an account of it can be reduced to accounts of individual constituents.[1] This can be said of objects, phenomena, explanations, theories, and meanings." (Wikipedia)

Now Reductionism need not be a bad thing and there are some contexts where it can hold its grounds. From Wikipedia: "Reductionist thinking and methods are the basis for many of the well-developed areas of modern science, including much of physics, chemistry and cell biology. Classical mechanics in particular is seen as a reductionist framework, and statistical mechanics can be viewed as a reconciliation of ..." (etc.)

It is the view-point of many (I being one) that in the complex times such as ours where we find the intermingling of many issues, a reductionist approach should not be the approach that is to be taken by a model proposer in any field. In architecture, more so: Architecture affects the external skin into which human beings live. It is at the fountain head of the future of our earth.

Contradictory requirements
I believe the central issue in any modeling system is to develop the representation framework correctly. One needs to handle two quite contradictory requirements: On one hand, you need to leave aside some parts of the reality from the model just so that you don't get into infinite regress. At the same time, the model should surely contain everything that you are concerned about. Now the word "concern" bothers me a lot. Because the issue of modeling will then get intricately tied to morals of society. What is it that a model proposer should be concerned about? Who will bell that cat? Today I find that the IFC standard leans heavily towards architecture of the developed countries. I don't see it helping much in modeling buildings that are hand-made and/or crude as seen in many third world countries and poor environments. In much of reality, the issues to be modeled turn up only at the very last point in time; just as the modeling takes place. As a practicing architect I have seen many situations where the modeling requirement changes at the very last minute. (This just-in-time modeling issue deserves a separate chapter of its own. So more on that later.)

One more issue that turns up in architectural modeling is that architecture is not just that of physical entities but it also contains experiences and memories. One can imagine the geometry of entities being captured and brought into a model for future analysis. But how does one describe experiences and memories in a clear manner? I believe architectural modeling using BIM software currently does not go the full extent. Firstly, it takes a highly reductionist approach and can easily get into infinite regression. (This can be corroborated by the large number of entities and regression levels that are modeled in IFC) Obviously, they have not got into infinite regression ... and the only way that can happen in a reductionist model is to simply stop the modeling at some level of the regression.

More important than the aforementioned points, I believe BIM is not asking all the right questions. In architecture, I am not only interested in the "How?" and the "What?" when I develop a model. I am also interested in the "Why?" For example; when I look at a BIM model, I do come to know "What is being created?" and "How is this to be created?" but it does not tell me why was it that the architect wanted those elements to be the way it turned out. When I took a look at the IFC model, there are some cursory reductionist elements to handle the "Why" but nothing that came across as a systematic approach.

The memory of the evolution of a design is critical for future analysis of any design project. Why should that be the case?, you may wonder. Architecture at the level of experience cannot be plumbed very easily for flaws. When a building is finally built, it exists in a singular universe. It just is there with no alternatives existing in the same location for a human to make comparisons between the alternatives. Now this state of affairs is often cleverly exploited: How you experience architecture can be easily rationalized by the architect as something the architect "anyway wanted you to experience".

Before you think I am merely joking, read this entry in Wikipedia on Peter Eisenman's work: "Eisenman's focus on "liberating" architectural form was notable from an academic and theoretical standpoint but resulted in structures that were both badly built and hostile to users. The Wexner Center, hotly anticipated as the first major public deconstructivist building, has required extensive and expensive retrofitting because of elementary design flaws (such as incompetent material specifications, and fine art exhibition space exposed to direct sunlight). It was frequently repeated that the Wexner's colliding planes tended to make its users disoriented to the point of physical nausea; in 1997 researcher Michael Pollan tracked the source of this rumour back to Eisenman himself. In the words of Andrew Ballantyne, "By some scale of values he was actually enhancing the reputation of his building by letting it be known that it was hostile to humanity."

(Disclaimer: I am using Eisenman only to point out that this can happen even with architects with a good reputation. It is surely possible that the Eisenman has not done anything premeditated.)

More complications
Ah... now it gets more interesting. I saved the best for the last. You see, when you model architecture; one cannot talk about just one frozen reality at the end of the design cycles. The reality is growing in the mind of the architect who is holding the design in an embryonic stage and then at some point it grows up to such an extent that it emerges out into the real world. So the modeling system we need is actually a series of interconnected, yet different modeling approaches In the initial hazy stages, there are many elements and issues that are to be modeled. During the final stages the architects are concerned with other issues. Each modeling system in the series of interconnected ones must play catch with the reality being handled at that point in the design cycle.

End note
I hope I have been clear in outlining why I am skeptical about current BIM approaches. Unfortunately, whenever I lay down these points in my talks (and I hope I am not making a strawman argument here) I still get that little lady (even guys) who will get up; obviously impressed by the technological wizardry of these new software, and say "You are talking rubbish. You see, the model is kept on a silver platter supported by turtles all the way down"