Clue in the windscreen may be key in legal challenge to Queensland’s AI traffic fine system

AI SaaS

Evelyn Trueman did jury duty once but, other than that, has never been to court. Now the 79-year-old Brisbane pensioner is preparing to represent herself against the Queensland government in one of several cases set to test the artificial intelligence programs it uses to surveil drivers.

Trueman admits to being daunted. She had to enlist her son-in-law’s help to lodge the first online complaint – “he’s good with computers”.

“I don’t know how to go to court,” she says. “Should I just front up and give a magistrate my files and all my pictures? I’ve got them all in little plastic files, is that the way to do it?”

Trueman fell afoul of the state’s Camera Detected Offence Program (CDOP) and was fined $1,078 in March for being the owner of a car in which a passenger was not wearing a seatbelt correctly.

She is adamant the car in the series of images is not her vehicle and that she has never met the two women pictured.

The West End grandmother did drive along the Pacific Motorway with her adult daughter on the same day and around the same time as the fine was issued. However, she insists she was driving (her daughter does not have a licence) whereas the image shows the older of two women in the passenger seat, with the seatbelt on incorrectly.

While the overexposed images provided by the government may make it difficult for the court to discern if it is indeed Trueman pictured, she is hoping that a small detail in the photograph will help support her case.

Visible in the car’s windscreen is what appears to be a reflection of Trueman’s upside-down numberplate.

“How did that get there?” Trueman wonders.

It is a question she has asked of the Queensland Revenue Office (QRO) and one also asked by this publication.

This image, which the Queensland government says is of Evelyn Trueman’s car, has her numberplate reflected in the windscreen
This image, which the Queensland government says is of Evelyn Trueman’s car, has her numberplate reflected in the windscreen.

“We cannot comment on individual cases,” a QRO spokesperson told Guardian Australia. “Images are not altered or redacted before they are included on the infringement notice issued to the registered operator of the motor vehicle.”

Trueman’s theory is that a similar car also carrying two women must have been just behind her, catching the reflection of her numberplate in its windscreen. The computer has then recognised her numberplate and incorrectly attributed the fine to her.

So does the theory stack up?

Griffith University quantum physicist A/Prof Erik Streed says it would require a “very surprising set of circumstances” for the numberplate to be reflected upside down in the windscreen of the car behind it.

But he said that was still far more probable than the alternative – that, somehow, the car is reflecting its own numberplate in the windscreen.

Streed says the scientific principle of Occam’s razor – which prefers a simpler explanation over a more complicated one – is in Trueman’s favour.

“I guess the burden of proof is on the government, but in this case they’ve flipped that around and put the burden of proof on her,” he says.

Evelyn Trueman standing in front of her car, parked in front of her home
Evelyn Trueman says her $1,078 seatbelt fine, which she says was issued incorrectly, is ‘more than I get for a fortnight from the pension’. Photograph: Dan Peled/The Guardian

In its response to both Trueman and the Guardian, QRO references the AI software used to filter images from its highway cameras seeking out offenders. For Streed, this does offer one potentially simple explanation.

“If some machine learning algorithm has randomly gone out and been told: ‘Here is a batch of images, find a licence plate’, well here was a batch of images and they found a licence plate,” he says.

‘We are just the cannon fodder’

Queensland first began trialling the use of AI technology to detect mobile phone infringements in 2020, before widening it to seatbelt offences in November 2021.

Cameras – which can be fixed or mobile – take multiple photos of each vehicle passing by, including their numberplates and front seats. The images are then filtered by AI to look for indiscretions.

Authorities stress that a “trained and authorised officer” reviews AI-identified images before issuing an infringement notice.

“We use road safety cameras to save lives. We make no apology for prioritising the safety of Queenslanders on our roads,” a government spokesperson said.

The spread of new cameras, combined with large increases in fines, has led to traffic camera revenue increasing by nearly 70% last financial year, to $466m.

For Trueman, her $1,078 fine is “more than I get for a fortnight from the pension”.

But the technology isn’t faultless.

Earlier this year, it was revealed more than 600 drivers incorrectly lost their licence and almost 2,000 people were incorrectly fined after being erroneously sent double-demerit points for seatbelt offences.

And others are lining up to challenge the accuracy of the system in the courts.

Gavin Storkey is a former pilot turned Gold Coast small business owner who has opted to challenge his $1,161 seatbelt fine.

He also says he’s never been to court before and finds the prospect of representing himself daunting. But like Trueman, he feels he has no choice.

“We’ve got a family, kids, mortgage has gone up, everything’s gone up,” he says.

“That is why we are standing up for ourselves, especially when you believe you are wrongly convicted.”

Storkey does not deny the images sent to him by QRO show his car. It is clearly being driven by his wife, the passenger a friend’s son. But in the overexposed images several objects disappear or are near impossible to make out, including, he says, the accused teenager’s black tie and top half of his seatbelt.

Black and white image of two people in car
Gavin Storkey is taking the Queensland government to court over this seatbelt camera fine, which he says incorrectly found the passenger was not wearing their belt correctly

Since going public with his challenge, Storkey has been contacted by several other people complaining of similar stories.

“Everyone thinks this [system] was pushed out too quick and we are just the cannon fodder of them trying to correct it – and make some money along the way,” he says.

Another driver challenging her fine, who asked not to be named, is a professor in public health.

She was driving a fellow academic from Norway, who was the one accused of not wearing her seatbelt correctly.

“When you grow up in Norway, you wear seatbelts on the bus,” the Queensland professor says.

She was sent several images by QRO as alleged proof of the misdemeanour.

“We look like ghosts,” she says. “The quality of the picture is just ridiculous. I just don’t believe that they’re giving out fines on this. It is quite scary.”

Black and white image of two women in a car
The Queensland professor of public health driving this vehicle says the quality of photographs used to fine people for not wearing seatbelts is insufficient

The professor says she is supportive of public health measures and would pay the fine – if she were in the wrong.

“I know about carrots and sticks … I have published on driving, on women drivers and safety. This is my life, this stuff.”

What has her incensed is a program she says is “terrifying innocent people”.

Which is exactly how the professor feels as she prepares for the only option she believes she has left – a legal challenge.

“I am terrified about going to court and standing up in front of a magistrate,” she says. “It is just awful. It causes so much distress.”

Some mistakes will happen’

If the use of AI and big data in law enforcement and revenue collection is a cause for consternation, it is something to which Australians may become increasingly exposed.

University of Queensland professor of data science Gianluca Demartini says comparable countries such as the US have adopted these technologies far more extensively.

Demartini says AI promises a number of benefits for law enforcement, allowing it to scale up its ability to monitor behaviour – without requiring more boots on the ground.

Meanwhile, predictive technologies may anticipate geographic trends in crime, allowing police resources to be more efficiently deployed.

But Demartini says the enthusiasm to roll out new technology in law enforcement must be tempered by the understanding that no technology is perfect.

“Some mistakes will happen and should be expected to happen,” he says. “So it is not a great idea to use these types of technologies by themselves and trusting 100% their outcomes.”

The Queensland Council for Civil Liberties president, Michael Cope, says his organisation has questioned the accuracy of the seatbelt cameras and is concerned about a lack of transparency on AI in law enforcement.

“We’re not luddites, we don’t say police shouldn’t have access to technology,” Cope says. “But there are serious issues raised about this technology and, before it is introduced, there needs to be a public review of it.”

The lawyer says “critical decisions about people’s lives” are already being handed over to automated systems without sufficient oversight.

“These are tools which have great potential to affect people’s lives and we need to set up structures which control their introduction,” he says.

Trueman’s been told she won’t be represented by legal aid and that seeking her own lawyer would cost upwards of $5,000 – more than she has in savings.

So the pensioner has been forced to fight her charge alone. But she is determined to follow it through.

“I can’t pay [the fine] because it is wrong,” she says. “Why should I have to pay for that? I’ve paid enough getting photocopies as it is.

“It is a terrible feeling, not to be believed.”

Do you know more? Contact qlddesk@theguardian.com

AI SaaS

Leave a Reply

Your email address will not be published. Required fields are marked *