Road safety authorities are right to target mobile phone use while driving. Distracted driving costs many lives, and strict enforcement mechanisms play a crucial role in removing reckless motorists from our public roads.
The deployment of automated cameras to detect mobile phone and seatbelt offences is a logical step. However, it would appear, the technology is imperfect, producing errors that punish innocent drivers..
The problem is not the intent but the execution. Artificial Intelligence systems designed to detect a phone in a hand or a missing seatbelt can misinterpret benign actions. There have been cases of receiving hefty fines for holding a sushi roll or a wallet. Others have been fined for driving without a seatbelt because they wore a dark jumper or shirt that obscured the black strap.
When the system mistakes a shadow for a device the consequences can be severe.
A recent case in Canberra illustrates the limitations of the technology and the human review process. A driver received a $548 fine and three demerit points for allegedly using a mobile device on Gungahlin Drive. His father, a former public servant with experience in artificial intelligence, examined the images and identified an optical illusion.
The driver was driving a manual Kia Cerato. His open hand was moving from the gear shift to the steering wheel. He was not holding a phone.
Despite presenting clear evidence to Access Canberra, including recreated photographs demonstrating the lack of grip action and the visible car seat webbing, the family has spent six months fighting the fine. Review officers dismissed the evidence across multiple appeals.
Access Canberra insists its system works properly. The agency states that camera provider staff first assess flagged images, and then specially trained Access Canberra officers perform a final adjudication.
They claim to adjust exposure, brightness, and zoom to verify the offence. Yet, this secondary human oversight routinely fails to correct initial machine mistakes.
When an Artificial Intelligence system flags a violation, a human reviewer must act as a sensible safeguard. Instead, authorities appear to blindly approve the camera's findings. Reviewers tell drivers they are definitely holding an object, even when those same reviewers admit they cannot identify the object. This unfairly shifts the burden of proof entirely onto the citizen.
Motorists must invest extensive personal time, stress, and resources to prove their innocence against a flawed algorithm that seems to regularly issue fines by default.
Authorities must maintain public trust to ensure community compliance with road safety laws. When drivers receive unwarranted fines and encounter inflexible, unresponsive bureaucracies, that quickly goes out the window.
If a citizen can demonstrate a camera captured a manual gear shift rather than a mobile phone, administrators should withdraw the infringement.
Prolonging such disputes because the bureaucracy won't admit an error is a clear systemic failure.
Governments must audit the algorithms powering these traffic cameras. Technology providers must refine the Artificial Intelligence to distinguish between a mobile phone and everyday items or actions.
Furthermore, infringement review teams require better training to critically assess generated images. They must also possess the authority and the willingness to dismiss false positives without dragging citizens through months of appeals.
A just legal system should not force innocent drivers to spend months proving a computer wrong.
