Fleets using the latest technologies to track driver performance are likely already using machine learning, though some may not even know it.

Machine learning (ML) is often confused with artificial intelligence (AI). However, Brian Filip, chief technology officer with Idelic, explained the difference during a Truckload Carriers Association webinar on AI and ML in driver safety programs.

“People often conflate or use those terms in the same way,” he said. “They’re absolutely not the same thing.”

AI involves having a machine do something that would normally be done by a human. For example, Apple’s Siri making restaurant recommendations or reservations for someone. “We told the computer how to do something we otherwise would do [ourselves],” Filip explained.

Machine learning, on the other hand, describes technology that figures out or learns how to do something on its own. Think Netflix movie recommendations based on a user’s watch history. Within ML, you will find technologies that enable fleets to improve driver safety through “event prediction.”

Technologies that monitor driving performance and create a risk profile for each driver fall into this category. And fleets are seeing significant benefits from using those technologies.

Managing risk with technologies

Michael Lasko, director of environmental health and safety and quality with Boyle Transportation, said such offerings have allowed the fleet to become more proactive in managing risk.

“Early safety programs were very reactive in nature,” he said. They may have been able to detect things like hard braking situations, but lacked the context around what led to them. “We all want to say hard braking is bad,” he said. “In some cases, hard braking is good if you’re going to avoid an accident.”

Machine learning goes beyond simply reporting such events. It may, for example, incorporate video to show what led to the braking. This cuts down on “false flags” and allows fleets to direct scarce resources where they’re actually needed.

“We had relied on basic telematics, crash reports, motorist complaints – taking all this information and painting with a very broad brush,” Scott Reagan, director of health and safety with TForce Freight, said about traditional safety monitoring. “We would send a driver-trainer to a location and ride with everyone, because we didn’t know how to address that risk. Moving forward, [ML] allows us to be more prescriptive and preventive.”

Technology vindicating drivers

Even police reports can’t always be trusted after a collision, Lasko said.

“When we started getting cameras, we realized a lot of times the information on a police report was not correct,” he said. “Witness statements were not correct.”

When cameras were installed at Boyle Transportation, it wasn’t long before the technology vindicated a driver who’d been blamed for a wreck by law enforcement. That, said Lasko, put to rest any resistance drivers had towards the in-cab cameras.

“Drivers didn’t like cameras when we wanted to put them in the trucks,” he admitted. “Two weeks after we installed them in the whole fleet, we had the first incident where the police report said the truck driver was doing something crazy – but here’s the video footage showing the truck driver was doing everything by the book. It was like flipping a switch – drivers stopped hating the cameras and said, ‘There’s value here.’”

Fleets using the latest safety tools can be more targeted in their training interventions. This has helped with driver turnover. Instead of retraining all drivers in the fleet on backing due to an increase in backing incidents, the fleet can focus on only those drivers whose behavior put them at greater risk of backing incidents.

“As a professional driver myself, nothing was more irritating than getting training on stuff you don’t have a problem with,” Lasko said.

While technology may seem like an easy fix to a fleet’s safety problems, Filip said that’s true only if it’s utilized effectively. He pointed to three challenges a fleet must overcome when incorporating ML safety technologies: keeping up with the volume of data generated; consolidating data from different platforms in one location; and determining which data is important and requires action.

The human element

Fleets using ML safety technologies must also maintain the human element.

“This needs to be viewed as another tool in the toolbox to help that driver share roads safely,” Reagan said.

Filip agreed: “Keep it human,” he said. “If you go too far down the ML path and just tell people you should completely trust this machine and do exactly what it says, people do not trust that. It has to be part of a program. ML is there to assist human beings to make the best decisions.”

It’s about building trust, he added. “Don’t kick the humans out of the loop. They have to be part of this process.”

“You start to think that all this technology removes the human element. Sometimes, it helps expose it.”

Michael Lasko, Boyle Transportation

Data collected by safety technologies also give a safety manager confidence when addressing behaviors with drivers. They’re not acting on assumptions or incomplete data, Reagan said. Gamifying performance through driver scorecards created a competitive environment that pushed drivers to perform to their best, he added.

Lasko pointed out the technology can also help to identify some underlying issues that very much require a human touch. When one of Boyle’s safest drivers began showing uncharacteristic lapses in driving decisions, the fleet spoke with him and learned he was coping with a severe sickness in the family.

“That weighed on this driver. We were able to say ‘Let us help you out and find something more palatable for you to do so you’re not under so much pressure and making those mistakes on the road,’” he recalled. “You start to think that all this technology removes the human element. Sometimes, it helps expose it.”