This post explores ideas and challenges of implementing A.I. within the building industry.
In the summer of 2021, I read an article about computer software bias that has had a significant impact on the ongoing development of BuildUSA’s Build Collaborative Environment (BCE). BuildUSA is very concerned about potential effects of computer software bias as the building industry grows its data capabilities. The building industry is one of the most data intensive industries in the world. Currently, however, industry data structures are too haphazard to implement long-term big data programs that can support the development of significant, far-reaching A.I. programs. But, once initiatives like BuildUSA are more fully implemented, data structures capable of supporting a long list of big data and A.I. programs will be readily available.
The article “Using A.I. to Find Bias in A.I.” outlines how bias creeps into databases and algorithms both intentionally and inadvertently. People train the machines. Therefore, our biases can creep into artificial intelligence systems. Additionally, a whole host of companies and industries may try to game the algorithms to provide preferential information for their products and services.
As the presence of A.I. and big data grows within the building industry, these biases will inevitably raise their unattractive heads., If not properly managed, A.I. and big data can create new problems as big or bigger than the old problems they were originally designed to solve.
As Common Data Environments (CDEs) like BuildUSA’s Build Collaborative Environment (BCE) become more prevalent and robust, the ability to apply serious A.I. to large, continuously evolving databases will grow. Creators will need to be careful to minimize the extent to which their own personal biases are digitized at a large scale. It is perhaps unreasonable to expect that the initial systems developed by a relatively small cadre of people will be able to avoid bias. One can easily conceive of systems that are expected to provide truthful information equitably to all Building participants, only to find that some parties — whether they be sectors of the industry (e.g., Architects, Engineers, Manufacturers, Developer/Owners, Operators, Contractors, etc., or just specific companies or organizations that have certain vested interests – have special privileges, access, or search results that benefit their position.
As the databases become more integrated and robust in size and detail, you can envision different industries that sell competing systems, products, and services attempting to input biases into search algorithms to benefit their systems, products, and services. There are an infinite possibility of hypothetical scenarios in which searches could be inappropriately weighted in favor of a specific position. The intent of this post is not to explore all these possibilities, but to consider how bias can be limited and controlled to ensure that trust and confidence will grow as these data systems come online. The goal is to give each party confidence in confidence that fairness and objectivity are built into the structure of A.I.- enhanced big data databases.
Certainly, one key ingredient is to bring a wide variety of interests into the discussion early on in the process. To ensure that the original mission statement clearly states that the collective work is to “provide an unbiased, trusted resource for all parties that comprise the building industry.”
A second would be an ongoing, regularly scheduled monitoring process. Comprised of a Board of diverse industry professionals who both understand their industry and will provide a check and balance on how data is updated, analyzed, weighted, and shared. The goal is to continuously improve the user experience and data metric quality, but to ensure that bias although present, will be continuously monitored and updated.
Future posts will explore specific ways in which A.I. could be applied to Building, and the types of insights into the industry A.I. could help identify.
