.


Optimization Overdose


October 6th, 2011, essays

This is just thinking out loud – don’t take too seriously. Read below if you really, really want to. 

Optimization and optimizing are now central features in any architectural discourse, alongside performance. The concept of optimization is associated with the continuous betterment of a certain design by finding the optimal values for the parameters describing it. Nevertheless, there are certain aspects which require critical reflection.

Optimizing a system which is essentially flawed in its design will not result in the absolute best solution for the given problem. In short, if the design is flawed, no matter how much we optimize, the solution will still be bad architecture because the parameters describing it are the wrong ones – or the less important ones.

What do we want to gain by optimizing a certain design? We can optimize it towards certain quantifiable values, like energy efficiency, structural resistance, etc. of which we know what we want from – less energy consumption, less material usage, etc. There are though a host of ambiguous variables for which we don’t have the correct studies to discern their optimal values. Integration of street networks is a good example: at first sight we want it to be as accessible as possible. Then again, in real-life examples show us that less accessible places provide shelter and quiet from otherwise busy surroundings.

Architectural projects aren’t yet quantifiable in their full complexity and will probably never be. Coupling this with the fact that optimizing one certain parameter, or group of parameters, can result in the complete “de-optimization” of an other set of parameters which are not embedded in the process, and thus ignored. How can we be certain of the godly powers of optimization when the consequences can easily vary wildly?

Constructal theory, which offers a predictive framework and an universal “optimization” goal for Manuel DeLanda’s reality as a flow of matter-energy, states that For a finite-size (flow) system to persist in time (to live), its configuration must evolve such that it provides easier access to the imposed currents that flow through it. Nevertheless, while this can be valid for non-conscious systems, I strongly suspect that anthropic evolution is based on a fine balance between inaction and action, between movement and cessation and should be treated as a special case.

It is thus important to critically reflect on our new found desire to optimize everything we can get our hands on. Good architecture and good urban design draws its positive qualities from the tensions embedded in it and how well modulated they are.

3 Comments. Wow. What do you think? Add your voice to the conversation below. Should be fun! Click to show the comments.

We had a really intensive and reflective discussion on this yesterday in the studio. You’re right optimization alone and excluded is useless or even worse can be easily abused for pseudo-scientific justification / masquerade (and therefore goes along with most of trendy computational tools).

When it comes to design a project with means of computational techniques and parametric models, architects tend to think we’re working in a revolutionary new mode on our products. In fact: we’re not. In every design attempt we’re trying to find a solution for a given problem. Consequently, problem definition becomes the key aspect of an early investigation. Asking the right question such as: what is disfunctional/needs to be improved, what kind of intervention / do I need an intervention at all? Second, this clear image needs to be translated into your design objectives – what do I want to achieve / what is my goal? Without this basic preparation every investigation becomes impossible to evaluate (and potentially to optimize).

Optimization is an old hat. I think to introduce this discussion it is necessary to clarify about what kind of optimization we’re talking about. Lets start with the ‘traditional’ way of iteratively adjusting your design and making it better. What do we want to gain by that optimization? Well, why do you never finish a competition proposal way to early before the deadline? Because it is one of the fundamental instincts of designers to find the best fitting solution. This is what distinguishes your project from others. The trick lies only within the definition of optimization: again your design objectives come into play (what do I want to optimize).

Talking about computational/automated modes of optimization: the quantifiability of architectural projects or certain design objectives is a valuable point – I agree that a building/concept/design is not de-composeable to all extend, but still you can justify your decision towards a specific solution, right? If we can rely on measurable/quantifiable objectives it definitely makes sense in my opinion to use advanced evolutionary algorithms to evaluate the generated solutions for optimization. The phenomena of eventually contradicting character of design objectives (minimizing facade area, maximizing floor area) is an important fact which needs to be elaborated. We especially stumbled across that issue while playing with multi-objective evolutionary solvers. As a matter of fact multi-objective optimization never leads to an ultimate single design solution. And imo this isn’t necessarily a bad thing.

To conclude: our brain is still (and will be if enhanced proportionally according to Raymond Kurzweil) more powerful than computers (they are just faster). Consequently, the precise definition and elaboration of the design problem at hand is indispensable. I would be also careful to neglect Freud’s introduction of the influential subconscious. Therefore, convictions (in life) are one of the most important things when it comes to design. Optimization shouldn’t be a thread to unpredictability and surprisingly simple and beautiful solutions.

The answer lies really within the question: what is the purpose of all this?

Borrowing words from science without appreciating delicacy of their meaning is a mistake easy to make. But the problem with architects is that once they hear it, they start vomiting around. Overuse + poor contextual understanding of term “optimization” made discourse suspicious. I agree with Patrick that optimization needs a reference – be it a problem you want to solve or multidimensional, dynamic environment to which you want to adapt. And even at that point you can never be 100% sure about the consequences of your design solution because of unpredictable “black boxed” everyday phenomena… But it also does not mean we should be suicidal “ex nihilo” designers. What happened with “sustainable” will happen with “optimized”. It is not the right moment to draw a conclusion, but shit happens a lot. I am not disappointed though. The shame of saying things too fast and too loud makes us more careful and critical.

What a refreshing thread, I honestly enjoy reading the thoughts posted here. The thing that I find most interesting about optimizations is that we have always been optimizing stuff since our ancestors discovered fire. And during much of the twentieth century, optimization has been key concerns on the minds of architects and engineers, from Henry ford’s invention of the modern factory line to the communist mass fabricated, and serial, habitats.
Optimization protocols are always internalized in industrial designs, and in a time at the brink of a second industrial revolution they will remain important. It takes a lifetime for a true paradigm shift to occur simply because the status quo is a hard nut to chew on. Critics and conservatives in every profession try to pull veils over innovation under the guise of isms. In my opinion architecture is not different.
So is optimization a.science by itself? Per haps, but their application to the real.of arts (which constitutes architecture) is precarious and plays into the hand of critics. On the other hand, optimization is vital to architectural design. Would there be a Beijing Olympic.stadium if Herzog and the Meuron wouldn’t have contracted Arup to optimize the “birds nest” structure using particle swarms? Surely, but at what price. And secondly, who are the unsung heroes at Arup that defined the state of the art?
Optimization should be understood both fundamentally and applied.

C'mon, write your thoughts in the box below.