Data Science in the Service of Architecture.

I've been carefully reading articles about the impact of data science in architecture but haven't felt any cover the topic particularly well. This concerns me insofar as these changes are zooming toward the design studio. By comparison, there doesn't seem to be a lot of haste in communicating to firms the tools they'll need to adapt. Furthermore, the big data landscape is getting increasingly competitive. Firms without experience in data science will be at a disadvantage. Data science conversations are becoming more and more commonplace within the AEC industry and general public and, secondly, it's important to recognize excelling at any one type of big data application can require a deep understanding of the underlying math and science behind the data. This is a specialist knowledge many big data firm already have, and small and mid-size architectural and engineering firms will struggle to get. Venture capitalist Matt Turck's 2017 Data Landscape poster graphically represents the competitiveness I'm trying to express in writing. There are an absolute ton of smart, driven, and hardworking firms coming for your dollars. My hope with this article is that a defence can start to be built, and, on a whole, we can get data science working for architecture.

Before proceeding, it's worth describing some reasonable constraints on our approach to the topic. The rapid growth of big data in modern life brings many different characteristics of the field forward; here we are going to discuss the topic strictly in terms of building design management, that being the study of architecture in terms of economics and business analysis. Left aside for the moment are more existentialist questions related data science's role in architecture, such as whether it's a good thing or not to let an algorithm totally determine the form of a building. Instead I favour of questions of adaptability. This technology is coming toward the design studio and we need to try to get out ahead of it. Where I need to admit bias is that I have a strong belief data science can help myself, and my readers, build more valuable architecture.

It's the math itself which really distinguishes the study of data science. Many fields of math intersect at various points with data science, any one of which is worthy of its own international conference. Every subdomain of the topic is deep. How to bring all of this together in a single firm is one of the main goals of building design management. There are three broad areas to consider if facing a big data request in an architectural or engineering setting:

1. Infrastructure
2. Computer Science
3. Analytics

Infrastructure. This refers to all the physical characteristics of the system or network to be used for work and is meant to be as broad as possible. Questions regarding multiple monitor computer setups all the way to browser-based BIM software hosted in the Cloud are all valid objects of study. Having a firm grasp of the computer and network infrastructure used in the AEC industry, including its costs and capabilities, and how it scales, all create the framework necessary to support technology users carrying out the following two points within a firm.

Computer Science. At the core of this category is supporting the needs of a high-performance computation architecture department or some other per project application of the subject. Distinguishing performance in this field is signalled by a depth of knowledge with several types of programming used in the industry (or fluency in several different programming languages). This includes familiarity with generative design (which encompasses branches of artificial intelligence, machine learning, and neural networks) and geometry (mostly within the field of finite mathematics, such as combinatorics and graph theory, but also classical and differential geometry). On the horizon, leaders will soon be expected to support additive construction techniques within a firm like architectural 3D printing or construction drones. Finally, how predictive mathematical models are developed and relate to statistics has a substantial influence on our last category.

Analytics. If above was about programming expertise in data science, this category is about getting the answers you want from your data. Having been involved in the field now for years, asking good questions of your data still seems to be more art than science. Several different branches of mathematics, such as linear algebra, matrices, and discrete optimization techniques, are involved in analyzing data. Network analysis can also be a powerful tool because as the data is analyzed, it often starts to reveal all sorts of complex connections within the data to that can be exploited (such as adjusting a product's supply chain). Building performance analysis in its many forms – thermal, structural, etc., – are all examples of activity in this field. Once all this data is in the model, it's time to analyze its form and predict its behaviour.

Conclusion. Ultimately, all three areas should be addressed on every project. This framework will help support better decision making about data science topics in architecture. Assuming one of the hallmarks of creativity is open mindedness, if the architectural and engineering field considers themselves creative at all, it's important we be openminded to welcoming data scientists and robotics engineers into the design studio. The creativity of the field can also be a source of adaptability.