According to the internet, artificial intelligence will save the world. Since construction is one of the most dangerous industries on the planet, it would be great to start there. According to OSHA, the rate of fatal accidents in construction is five times higher than any other industry. The good news is, one company is already developing AI technology to predict construction accidents before they happen.
The concept sounds good, but is it realistic? Maybe. To understand how AI can help predict construction accidents, it’s important to understand how construction accidents happen.
The dangers of working in construction
In construction, accidents happen quick and often. Workers fall from ladders, roofs, scaffolding, and equipment daily. Some workers are struck by materials being carried by machinery and falling debris. Workers are frequently injured when wearing improper protection or no protection at all. For those who survive, injuries aren’t minor and often require extensive recovery time. For instance, many suffer from traumatic brain injuries that result in long-term cognitive impairment.
The settlement amounts for construction accident lawsuits reflect how serious the injuries are. For example, Lipsig Law obtained a $20 million verdict for a construction worker who fell into a hole after the shoring collapsed. Another verdict of $8 million was obtained for a worker who fell due to defective scaffolding and needed multiple back surgeries.
These accidents seem impossible for a human to predict and it doesn’t seem like the workers are aware of the potential for danger. If a worker knew there was something wrong with a scaffolding, they wouldn’t use it.
Why construction accidents seem unpredictable
Accidents happen quickly because the injured party doesn’t usually have the perspective to see what’s coming. It seems possible to predict accidents with AI if the construction site could be monitored from different angles. However, even if AI can produce accurate predictions, how will it let workers know they’re in danger?
These are all questions Boston-based construction giant Suffolk has been trying to solve. Suffolk has been developing a predictive AI system for use in construction for over a year. The company has been in collaboration with computer vision company SmartVid to bring their vision to life. The project was recently discussed at the EmTech Next conference hosted by MIT Technology Review in June 2019.
Developing this project isn’t a small feat. Fully training an algorithm of this nature will require collaboration with multiple construction companies. To get a jump on collaboration, Soffolk invited their competitors to join their consortium called the Predictive Analytics Strategic Council. There’s no concern regarding competition when it comes to safety. If these construction companies can pull this off, it will be a big win for the entire construction industry. They will be heroes.
Predicting accidents is all about identifying patterns
Like any other AI project, Suffolk’s deep-learning algorithm is being trained to identify patterns in photos, videos, and accident records from construction sites. Identifying patterns that lead to accidents is crucial in training the algorithm to predict accidents. With enough training, the algorithm will eventually be able to monitor a construction site and flag situations that indicate a potential accident. For example, when a worker isn’t wearing gloves or gets too close to dangerous machinery, the system will flag the situation.
It’s unclear how the system would signal workers in time to avoid the danger, but once that’s sorted out, AI will play a critical role in worker safety.
AI is a supplement – not a replacement – for safety policies
It’s unrealistic to rely on safety policies to protect workers. Most injured workers were employed by companies with strict safety policies and they were still injured.
Safety policies are only effective when followed, and sometimes accidents happen beyond anyone’s control. Companies can try to change a worker’s behavior to force them to comply, but that can be futile. For instance, if an employee doesn’t want to wear hearing protection, they won’t. If a worker wants to perform a quick task without putting on cumbersome protective gear, they’re going to risk it.
Even well-trained, diligent workers are capable of making mistakes and taking shortcuts that put themselves or others in danger. Using AI to predict potentially dangerous scenarios will be the safety net that catches those unfortunate and uncontrollable situations.
If AI can make construction work safer, perhaps it really can save the world.