Flawed immigration detention risk assessment tool can’t be upgraded as ABF data ‘riddled with errors’

<span>The high-security main gate of the Melbourne immigration detention centre in Broadmeadows. </span><span>Photograph: The Washington Post/Getty Images</span>
The high-security main gate of the Melbourne immigration detention centre in Broadmeadows. Photograph: The Washington Post/Getty Images

The secretive risk assessment tool used in Australia’s immigration detention centres could not be replaced by a better model due to insufficient data collection by Australian Border Force, documents reveal.

The security risk assessment tool (SRAT) is meant to determine whether someone is low, medium, high or extreme risk for escape or violence. It calculates risk ratings based on factors including pre-detention history and episodes that can occur in detention, such as possessing contraband or refusing food or fluids.

In March Guardian Australia revealed serious flaws in the SRAT and its algorithm that have led to people being wrongly rated as high risk, influencing where they are placed in a detention centre and whether they are handcuffed. The tool has been criticised for inaccuracies and “unwarranted” escalations in ratings.

Related: Revealed: the secret algorithm that controls the lives of Serco’s immigration detainees

Griffith University researchers reviewed the SRAT for the ABF in 2019, concluding: “It is clear that the tool is not borne out of sound scientific research.”

Documents obtained by Guardian Australia via freedom of information laws now reveal that researchers also could not proceed with building “a valid and reliable risk assessment tool” in 2020 due to problems with how the ABF collected data.

“ABF data systems do not have the information required to conduct the necessary analyses to build a psychometrically sound assessment tool,” ABF minutes detail.

In 2019 researchers found potential errors with how detainee data was recorded. For example, a person was recorded as having been involved in “over 3,000 incidents” in a year – a scenario described as “incredibly unfeasible” in an email.

“This issue comes up with multiple cases, which is worrying as this means the whole dataset may have issues,” the researcher wrote.

The 2020 ABF minutes also refer to “extraordinary resourcing priorities” after the bushfires and Covid-19 that limited the force’s ability to proceed with “achieving a new and more reliable” tool.

Daniel Ghezelbash, deputy director of the Kaldor Centre for International Refugee Law, said the documents called into question the data used to make risk assessments.

“Here you’ve got data riddled with errors combined with a very primitive tool that is not backed by science,” he said.

“Those two things combined create what is potentially a highly inaccurate tool that appears to be identifying the vast majority of people as high risk, having significant flow-on effect on how they’re held and the restraints used.”

As at 18 January 2024, at least 75% of immigration detainees were rated by Serco’s SRAT as high or extreme risk.

Jonathan Hall Spence, a lawyer at the Public Interest Advocacy Centre, has represented clients who have allegedly been inappropriately handcuffed while in detention. He suggested the documents raised serious questions about the SRAT’s continued use.

“The Griffith report was produced in 2019,” he said. “At least from that time, the department has continued using a tool that it knows to be flawed.”

It appears Home Affairs has not completed a privacy impact assessment for the tool. An FOI request seeking one did not return any documents.

A spokesperson said the ABF “takes seriously the safety of detainees, staff and visitors”.

“The SRAT considers each detainee’s individual circumstances, including consideration of an individual’s capability (e.g. age, frailty, medical condition) and intent (e.g. immigration pathway, behaviour, prevalence of incidents) and is reviewed at regular intervals.”