National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026
Submission 3263
Dear Senate Community Affairs Legislation Committee,
I am writing to express serious concern regarding provisions within the proposed NDIS legislation that would enable automated systems and algorithmic processes to determine funding outcomes for disabled Australians.
I am an AI researcher, artist, and lecturer at RMIT University. My research and teaching focus on machine learning systems, algorithmic bias, computer vision, generative AI, and the social consequences of automated decision-making systems. I am also a disabled person with family and friends directly impacted by the NDIS.
Based on both professional expertise and lived experience, I strongly oppose the replacement of human-centred planning and decision-making with automated or algorithmic systems in the NDIS. Disability support is not a logistics problem, it cannot be solved through algorithmic means. Disability support is a human system dealing with complex medical, social, psychological, and environmental realities that often cannot be meaningfully reduced to data points or statistical averages. Automated systems are fundamentally incapable of fully accounting for the variability and nuance of lived disability experience.
Australia has already seen the consequences of deploying opaque automated systems within welfare administration. The Robodebt scheme demonstrated how automation, combined with institutional pressure and inadequate human oversight, can produce widespread harm. The Royal Commission documented severe psychological distress including suicide, financial devastation, and systemic administrative failures caused by treating algorithmic outputs as authoritative.
The proposed NDIS changes raise similar concerns, particularly in a context where funding decisions can directly affect a person’s access to daily care, housing stability, medication management, mobility, communication support, and safety.
I am especially concerned by reports that:
-
automated systems may be used to calculate funding amounts;
-
participants may have limited ability to challenge the reasoning or methodology behind these calculations;
-
human planners may be prevented from meaningfully adjusting or overriding automated outcomes;
-
the calculation process itself may not be transparent or independently reviewable.
These concerns are not just theoretical and they are reinforced against the history of robodebt. It has been proven time and again that Algorithmic systems routinely encode the assumptions and limitations of their designers, training data, institutional priorities, and historical biases. They can fail silently, scale errors rapidly, and create an illusion of objectivity around decisions that are ultimately subjective and political.
In disability contexts, those failures carry severe, life affecting, consequences.
National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026
Submission 3263
A person denied adequate supports cannot simply “wait for the system to improve.” Incorrect decisions may result in physical harm, loss of independence, social isolation, deterioration of health, institutionalisation, or preventable crisis.
The lessons of Robodebt should lead Australia toward greater human accountability in welfare systems, not deeper reliance on automated decision-making.
I urge the Committee to reject provisions that normalise or expand algorithmic determination of participant funding and support needs. If any automated systems are nevertheless pursued, they must remain strictly advisory only. Human planners must retain full authority to depart from algorithmic outputs, participants must have full rights to challenge both data inputs and calculation methodology, and all systems must be independently audited for bias, error rates, and disproportionate impacts. Disabled Australians should not become test subjects for another large-scale automated welfare experiment.
Thank you for considering this submission.