These offshore detention photos are harrowing. But they’re not real
A new series of images depicting the experience of refugees detained on Nauru and Manus islands show the alleged horrors of offshore detention, including wet tents covered in mould, the aftermath of a sexual assault and the self-immolation of an asylum seeker.
But the 130 images were not taken by photographers; they were generated using artificial intelligence.
Maurice Blackburn principal lawyer Jennifer Kanis said more than 300 hours had been spent collecting the witness statements.
Based on the witness statements of 32 detainees, collected by lawyers from social justice firm Maurice Blackburn, the statements and images form the online exhibition “EXHIBIT A-i: The Refugee Account”.
The images have raised ethical questions about the use of AI.
Fake images were circulated on social media last month depicting former US president Donald Trump getting gang-tackled by New York City police officers and Russian President Vladimir Putin in a prison suit behind bars, prompting concerns about how the technology can easily spread misinformation.
Maurice Blackburn partnered with advertising agency Howatson+Company to develop the AI images, working with the witnesses who gave feedback at every stage.
Executive creative director at Howatson+Company Gavin Chimes said the images were a crucial part of storytelling.
“We knew that words alone weren’t enough to shift the dial,” he said, pointing to photos such as the tank man from Tiananmen Square of the Napalm-affected girl in Vietnam, which spurred public debate.
“We wanted to create images with the same weight and power as photojournalism.”
Journalists and photographers are restricted from visiting detention centres and detainees are not allowed to take photos and videos.
One woman alleged she witnessed the aftermath of a rape of a refugee woman on Nauru, as depicted in his AI-generated image.
KPMG chair in organisational trust at the University of Queensland Nicole Gillespie said most people across the globe were distrustful of AI.
“It is important to establish who is accountable for decisions made by AI systems, and to have in place clear processes that allow people to contest decisions made about them by AI algorithms that they think are not fair.”
There have been advancements in AI transparency, with tools emerging to help understand whether an image has been created by AI.
But Gillespie said there was more work to be done to advance these tools and implement legislation or regulation to mandate the use of those tools.
Detainees alleged humidity caused mould to grow in tents, with water dripping onto them as they slept as depicted in this AI-generated image.
“How do we know what’s true or not true any more? That’s what’s at stake,” she said.
To deal with the ethics of using AI, Chimes said it was important that the images were properly labelled as illustrations, were an accurate depiction of the witnesses’ experiences, and co-designed with them.
Maurice Blackburn principal lawyer Jennifer Kanis said more than 300 hours had been spent collecting the witness statements. “We wanted to find ways we could do justice to these stories, and [fulfil] the witnesses’ wishes that these stories be told,” she said.
One detainee alleged her baby learned to crawl on a sheet laid out in front of their room, with no other space for the baby to safely explore, as depicted in this AI-generated image.
The law firm had been pursuing a class action against the Commonwealth of Australia, alleging people were unlawfully detained between 2011 and 2020, seeking compensation.
The case was discontinued in 2021 following a High Court ruling that the detention of asylum seekers was legal, even when they had not been removed from Australia.
Without the exhibit, the witness statements would be archived and then destroyed in seven years under court rules, Kanis said.
“We really didn’t want that to happen, and we saw that as a great loss … There has been a lot written about the system, [but] nearly all of it was written by people who weren’t there.”
A survey of over 17000 people across 17 countries, including Australia, released in 2021 and authored by Gillespie, found just 39 per cent of people were willing to trust AI.
The images were generated from more than 300 hours of refugee testimony.
Independent, external oversight was expected by 71 per cent of people yet just 39 per cent believe the current laws and regulations are sufficient.
“There’s strong global consensus amongst the public about the principles and practises they expect to be in place to trust the use of AI, including high standards of data privacy, security and governance,” Gillespie said.
Respondents also expect the accuracy of AI systems to be robustly tested, including for bias before being put in use, that there is transparency around the use of AI, especially in the public sector, with risk assessments, and that the AI is explainable.
The Morning Edition newsletter is our guide to the day’s most important and interesting stories, analysis and insights. Sign up here.
Most Viewed in National
From our partners
Source: Read Full Article