Back to Blog

Artificial Intelligence Won’t Solely Protect Against Risk on the Road

Posted by Stephanie Forster on April 1, 2020 at 8:34 AM

Although there have been great technological advancements in transportation, advanced driver-assistance systems (ADAS), computer vision (CV) and other forms of artificial intelligence (AI) – on their own – aren’t perfect. Today we have semi-intelligent systems that can “see” but cannot replicate all human sensory risk inputs, limiting their ability to mitigate road risk. They do not change driving behavior; in fact, most drivers ignore when a semi-intelligent system is wrong, and also tend to ignore when the system is right. If a fleet is solely relying on artificial intelligence, specifically CV, its safety is at risk. Although these systems are part of a well-layered approach to mitigating risk on the road, humans still provide intelligence with the addition of their five senses, historical risk context for optimal road risk mitigation and coaching effectiveness.

Computer vision for transportation risk assessment is used to automate image processing/data capture to identify:

  • pedestrians
  • facial cues
  • traffic signals and signs
  • trucks, cars & scooters
  • and other objects.

Computer vision is also used in autonomous driving, ADAS (object detection, lane departure and pedestrian detection), parking assist, accident reconstruction/exoneration, claims processing and driver behavior analysis to detect distracted driving, drowsiness, mobile usage, drink/eating while driving and communication with other passenger and passenger activities.1 However, computer vision faces many challenges. As a result, data, alone, can be limited by the circumstances in which it was captured, limiting the effectiveness of CV in risk detection.

Safety programs that facilitate human-to-human coaching that can reduce risk long-term and modify a driver’s behavior are the solution. Humans bring context, five senses and historical risk intelligence that sensors and CV cannot fully replicate at this point. “When we see, our brain is using the context and experience of our entire lives to help, which is one of the reasons we can see so much better than computers.”2 Humans serve to validate the data so that a safety program is effectively using accurate actionable data for coaching and behavior change. For instance, humans can hear and see greater distances and process that a child with a rolling ball may go into the street after the ball and react proactively vs. a sensor that can only pick up within a specific range to react.

A managed service environment as a tool used by humans for coaching and risk management provides a tremendous advantage over computer vision, alone. Human review of the data capture can help eliminate false positives and add needed context for human-to-human coaching to achieve behavior change and risk reduction.

A managed service enables fleets to save time and money by taking the burden off the fleet and providing consistent, unbiased and professional reviewing, scoring and prioritizing of thousands of videos. Only a managed service allows fleets to focus on their business – not on reviewing video.

 

1 https://appen.com/blog/computer-vision-vs-machine-vision/

2 https://austinstartups.com/human-versus-computer-vision-94183d6737f6