Experts hope AI will reduce restrictions for PWDs

Australian researchers hope that a new educational program supported by artificial intelligence (AI) help reduce the use of restraints and seclusion on people with disabilities.

The program Positive Behavioral Support Practice is designed to help professionals create safer, respectful and constructive plans to guide caregivers' and support workers' responses to challenging behavior.

🇬🇧 Subscribe to the best newsletter about AI (in English 🇬🇧) 

(I.e. Sign it the best newsletter about AI (in Portuguese 🇧🇷)

“Challenging behavior” includes destructive acts, self-harm, aggression, or any behavior at a level of repetition or intensity that may place a person with a disability at risk of harming themselves or others.

Disability advocate Jarrod Sandell-Hay believes proposed changes to the National Disability Insurance Scheme (NDIS) would restrict independence, access to care and be more expensive for people.

“This behavior is people with disabilities using the best means they have to try to communicate that something is not right for them,” said Prof. Keith McVilly from the University of Melbourne, who worked on the project.

“They are in pain, they are sick… they are not being heard, and we are failing to provide the support they are looking for. These behaviors are a last resort.”

A royal commission into disability heard harrowing stories about people subjected to restrictive practices. In the 2021-22 reporting period, there were 1,4 million instances of unauthorized restrictive practices used against 8.830 NDIS participants and 5,58 million instances of authorized use against 8.685 people.

Restrictive practices may include confining a person to a room or physical space (known as seclusion), the use of a device to restrict a person's movement, physical restraint by another person, or the use of medication.

The NDIS watchdog has identified the abysmal quality of most behavior support plans – which are submitted to the commission as part of the regulatory process – as a key problem.

The commission's 2022 audit found that more than 80% of plans were weak or poorly developed, especially with regard to identifying positive or therapeutic ways to address the needs of the person with a disability, and almost 70% of plans did not include consultation with a person with a disability.

The new program allows behavior support professionals to upload segments of their preliminary plans into a tool – developed using natural language processing and machine learning models – which then analyzes the plan and provides feedback.

When poor practices or areas for improvement are identified, the program directs the practitioner back to specific components of online educational resources. “The idea is that they are almost being driven by AI in conjunction with educational resources, so that when they submit the plan, it should be functional or compatible,” McVilly said.

The program resulted from a collaboration between the University of Melbourne, University of Queensland and Flinders University, and was funded by the NDIS Quality and Safeguards Commission. It was developed in part in response to the increased use of restrictive practices.

Read also