Artificial Intelligence in Construction: Part III

Joseph A. Cleves, Jr. | Taft Stettinius & Hollister

As we noted in our first article on artificial intelligence in construction, artificial intelligence (AI) is a broad term that generally refers to technology that uses algorithms to process data and simulate human intelligence. In our first two articles, we discussed machine learning and then image recognition and sensors-on-site. In this article, we discuss two more AI-related topics: (1) building information modeling; and (2) smart contracts.

Building Information Modeling

Building information models (BIMs) are three-dimensional, digital, construction blueprints. BIMs allow numerous project participants to view and modify the same model. They are generally highly detailed, allowing users to access information on each building.

BIMs offer several benefits over prior practice. They provide participants with the capacity to visualize and comprehend a design much more easily. They also allow for better communication between participants by constantly updating the design as changes are made. BIMs can also lead to improved design quality, detail, and precision because of their digital form. Finally, because BIMs are constantly updated, they allow owners to monitor a project closely for deviation from the original plan. These benefits may also reduce the risk of liability in some cases.

However, BIMs also create several new risks of liability, two of which we note here. First, the roles and responsibilities of participants can become irreversibly intertwined in a BIM. In other words, it may become impossible to ascribe responsibility, and therefore fault, to the correct actor when numerous actors are given broad decision-making authority. Given all the hands touching the design documents, it is very important, in particular, to define the design responsibilities carefully. 

Second, BIMs create intellectual property right concerns. The traditional rule is that the party that creates the model owns it. But since BIMs are often compiled from information contributed by numerous sources and parties, the situation becomes more complicated. The solution to this issue is to address it in the contract. If parties fail to do so, they will be left to follow a convoluted web of information to locate the “true owner” of the model.

“Smart Contracts” and Blockchain Technology

“Smart contracts” is a phrase used to describe computer code that automatically executes all or parts of an agreement automatically and is stored on a blockchain-based platform. A blockchain refers to a decentralized, online ledger, which holds a time-stamped series of immutable records of data managed by a cluster of computers. 

There are two types of blockchains: public and private. Public blockchains are decentralized and allow anyone to join the network and participate in the blockchain. Once transactions are validated, the data is secured and cannot be modified or altered. Private blockchains, on the other hand, limit who can access and participate in the network and are typically controlled by one or more entities. Accordingly, only parties participating in a transaction can access private blockchains. Similar to public blockchains, once transactions are validated in a private blockchain, the data cannot be modified or altered. Large networks of computers are used to verify blockchain transactions and store blockchain data. The use of blockchain technology can eliminate transaction fees and the need for third-party verification typically required in certain transactions. 

Smart contracts, like traditional contracts, define the rules and liabilities of the parties. The difference, however, is that smart contracts automatically enforce their obligations and liabilities. Once operational, smart contracts generally require no human intervention to execute and enforce their terms. Smart contracts are currently suited for simple transactions. An example is automatically transferring funds from one party to another when specific criteria are met and imposing damages if certain conditions are not met. Hybrid-contracts, however, consist of a traditional written contract alongside a smart contract to cover an automated function, such as payment.

Implementing fully autonomous smart contracts in the construction industry presents many issues. One is the need for unique code to accompany every smart contract. While common transactions can recycle the code from other smart contracts, every construction contract contains significant elements that are unique that would require unique code. Another issue arises from the need for funds to be pre-loaded into a digital wallet so that the smart contract can automatically execute its payment obligations. Construction projects can be expensive. Requiring a party to advance funds into a virtual wallet until completion is likely not a viable option. Additionally, fully autonomous smart contracts would likely require some reliance on outside information that cannot be anticipated in the contract code. Data from image recognition and sensors-on-site, for example, will need to be continuously updated in a fully automated smart contract. Outside data reliance for smart contracts poses what many refer to as the “oracle” problem. Outside data can only be provided to smart contracts through manual input. Thus, it would require a party to hire someone who specializes in providing data for smart contract codes. Utilizing an oracle will create additional fees and less autonomy as the smart contract would fail if the oracle fails to relay the outside information. Further, parties using an oracle must trust that the oracle will adequately perform its duties as mistakes could render a smart contract useless.

Smart contracts also pose some important legal issues. Because data shared on blockchain technology cannot be altered or modified, it is virtually impossible to change the terms of the contract. Where modification is necessary, new smart contracts must be created. Additionally, courts will likely struggle to adjudicate smart contracts and blockchain technology due to a lack of familiarity with the new technology. One way some parties have nonetheless sought to take advantage of the benefits of smart contracts, while addressing their limitations, is to use hybrid-contracts that contain elements of both traditional and smart contracts. This allows some automation and provides security for parties by having a written contract that can be easily read and interpreted by a court. For this reason, hybrid-contracts appear to hold the most promise for industry-wide application. 

Artificial Intelligence in Construction Part II: Image Recognition and Sensors-on-Site

Joseph A. Cleves, Jr. | Taft Stettinius & Hollister

In this article, we continue our series on artificial intelligence (AI) in construction. Here we address image recognition and sensors-on-site. This technology uses cameras and other sensors to assess vast quantities of video, pictures, and other recorded conditions from worksites. Such technology has the potential to: (1) monitor worksite conditions for safety risks and hazards; (2) enhance equipment and material management, boosting productivity; and (3) improve worker safety by identifying unsafe behavior to inform future training priorities. 

Firms can leverage machine-learning techniques with image recognition programs to keep workplaces accident free and increase work efficiency. For example, Suffolk, a Boston-based general contractor with about $4 billion in annual revenue, is already developing predictive algorithms to monitor safety risks. Suffolk collected over 700,000 images, taken from over 360 job sites in the last 10 years, and uploaded them to startup Smartvid.io’s cloud-based platform. The algorithm analyzed the images to identify safety hazards, like workers not wearing proper protective equipment. Suffolk plans to expand the algorithm also to identify tripping hazards from tools and equipment lying around on sites. The algorithm will then compare the images scanned with Suffolk’s accident records to inform future training opportunities.

Suffolk is also exploring ways to use sensors-on-site and the internet of things to improve work efficiency. The advantage of having real-time data from connected devices is that workers can easily locate equipment on the job site and contractors can track materials from suppliers. Workers will not only find the tools and equipment they need on site faster, but more importantly, they will also know whether the tools are currently in use. Contractors will be able to track the location and arrival of important materials like concrete supply trucks en route to job sites. Knowing where available tools are and exactly when critical materials will arrive can reduce downtime and increase productivity through better planning and resource allocation. 

In the future, image recognition and sensor-on-site technology coupled with machine-learning techniques could also be applied to assess issues with quality control. By analyzing real-time data from sites, engineers can potentially detect defects in design. Catching project design deviations earlier creates a better opportunity to rectify them and limit any associated costs. With frequent monitoring, engineers may be able to detect the potential for critical failures or events with enough warning to limit, or even prevent, the occurrence. This can be applied not only to structures, but also heavy machinery.

A 2017 McKinsey report estimated that construction firms could increase productivity by as much as 50% through real-time analysis of data. The desire to capitalize on this opportunity to boost the industry’s generally low productivity is compelling construction firms to invest in AI. As construction companies incorporate more AI and machine learning into their business and worksites, new legal concerns associated with this implementation will arise. Some unanswered legal questions are the allocation of risk, responsibility for malfunctions and resultant damages, and the confidentiality and privacy of the data.

A primary concern for construction industry stakeholders will be what new duties and responsibilities will accrue to those who implement and use such technology. With the potential to identify safety hazards or unsafe working conditions, an open question arises: Who has the duty to observe or monitor the information? How closely and actively must sensors be monitored? Will there be a duty to act upon identified risks, or merely a duty to disclose? Will using this technology contribute to a party’s constructive knowledge regarding unsafe conditions that result in injuries or potential failures to meet contractual obligations? When something does fail, who will be responsible for repairs or the costs? How will the data be stored, who has access to it, and how will privacy and confidentiality be secured? In summary, contractors may unknowingly be opening themselves up to additional risks, liability, and greater responsibility with the information this technology provides. While most of these questions can be addressed through careful contractual drafting, stakeholders will have to think through these questions and possibilities. To reach acceptable risk allocation as AI usage in construction increases, parties should be prepared to negotiate these terms in any agreement very carefully.

Artificial Intelligence in Construction: Part I

Joseph Cleves, Jr. | Taft Stettinius & Hollister

Introduction

Artificial Intelligence (AI) is a broad term that generally refers to technology that uses algorithms to process data and simulate human intelligence. Examples of AI technology include machine learning, image recognition and sensors-on-site, building information modeling (BIM), and “smart contracts” stored on a blockchain-based platform. This technology can be used in the construction industry by way of design, operations and asset management, and construction itself. Construction leaders interested in staying ahead of the curve should consider its advantages, and the legal implications.

This article will discuss our first AI-related topic: Machine learning. In three subsequent articles we will discuss (1) image recognition and sensors-on-site; (2) building information modeling; and (3) smart contracts.

Machine Learning

Machine learning is a subset of AI, but it is the basis for the vast majority of AI technology. Machine learning at its core is a simple process: using an algorithm and statistics to “learn” from huge amounts of data. The data doesn’t have to be just numbers; almost anything that can be digitally stored or recorded can be used by a machine learning algorithm. This type of technology can be used to recognize patterns, extract specific data, make data-driven predictions in real time, and optimize many processes.

Machine learning’s ability to process and detect patterns in large amounts of data makes the technology ideal for data-intensive tasks like scheduling and project planning. To aid in project planning, machine learning technology can include the process of “reinforcement learning.” That is when an algorithm applies automatic trial and error. This is different than the usual process of humans collecting, labeling, and categorizing the underlying data that machine learning relies on. The autonomous process of reinforcement learning allows the technology to offer optimized suggestions efficiently and continuously based on previous, similar projects. It also allows the technology to help assess risk in a project, constructability of a project, and various materials and technical solutions for a project.

Firms can use machine learning to identify risks, such as when certain assets will need maintenance, by using data on various machines and equipment. The machine learning technology then analyzes the data to predict when preventive maintenance will be needed. This can increase efficiency by avoiding the need to take assets out of operation due to a breakdown. 

These examples of risk management and project and design optimization just scratch the surface of how machine learning can be applied. This technology can optimize virtually any process that generates data, such as bidding, pricing of fixed-price contracts, recruiting and talent retention, and inventory management. To begin implementing machine learning, firms should identify processes where optimization from this technology would maximize return on investment.

For companies interested in using machine learning, it will be important to address the issue of risk allocation in the contract documents because the state of the applicable law is not clear. The parties should map out precisely who will own the risk associated with the technology and what degree of liability a party is taking on. This issue is especially important depending on who owns the technology – the construction firm, or a third party. If a construction firm owns the majority of the risk associated with the technology, then the adoption of machine learning technology in construction may decline.

Lastly, the parties will need to determine who will own the data that the technology records and uses and whether the data needs to be protected. Parties will need to determine how that data can be used by the company supplying the technology or other third parties, if at all. That issue is particularly relevant if the technology is provided by third parties who want to use the construction firm’s data to refine their technology. And, if the data needs to be protected, the parties will need to negotiate contract terms that dictate the protection protocols. 

This article has just scratched the surface. Look for the next installment on image recognition and sensors-on-site next month.