National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026
Submission 3283
Please accept this email as a submission to the committee. While submissions were extended to Monday, the online form did not allow for making my submission that way for this enquiry. I am happy for this submission to be published, but without my name and contact information please.
To the Committee,
I wish to raise concerns about the aspect of the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026 (‘The Bill’) regarding the use of automated algorithms when calculating funding. I have concerns that are both on principle and in practice. I shall try to be brief.
About me —————————– I am writing as an individual who happens to feel that he has some relevant experience. I am a former database programmer and IT Professional. I have taught computing systems subjects, including project management, at both the secondary and tertiary level. I am currently a secondary school teacher of over 20 years’ service. I also have children with NDIS plans and others in the household (including myself) with diagnoses that don’t quite reach the level to attract NDIS support. I live in a rural city on a state border.
The principle —————————– The NDIS, when created, was sold to us as allowing for individualised care that suited the person, and not the system. I would describe this as equitable care (each to their needs), not equal care (each gets treated the same), and that is both as it should be and one of the main positives of the NDIS. I believe that resorting to an automated algorithm, especially one without recourse to appeal, is anathema to this central concept of the NDIS. I understand the desire to contain funding blow-out, but it should not be at the expense of that individualised support which is what we all rallied behind.
The practical —————————– In a country the size of Australia, there can be no one-size-fits-all approach. Rural citizens have costs that can seemingly bear no relation to what it might be in the capital cities. Even if the services cost the same, some services just frankly do not exist and would require major journeys to a capital city to access. Creating a system that ‘decides’ based on an algorithm without human oversight, or override, requires a vast capacity to deal with caveats, exceptions, and special cases. How can all possible cases be considered ahead of time? Geography and access to services is just one example.
Please allow me a non-NDIS example, and two NDIS counterexamples, to show how the current system doesn’t inspire confidence in this proposal.
Non-NDIS example —————————– Over 20 years ago, before Google Maps made things easier, I helped automate the ability for the NSW Department of Transport to process ‘special case exceptions’ to the School Student Transportation Scheme (SSTS) parameters. These were clearly identified exceptions to the 2km as the crow flies distance rule. They were: if the student had a medical condition (either temporary or permanent); if there was a safety issue (either temporary or permanent) regarding
National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026
Submission 3283
walking to school; or if the distance to safely walk was over 2.9km even though the as-the-crow flies distance was less than 2km. This is a straight-forward set of parameters but assessing them had required physically going to the venue at times – not optimal. My system noted if previously exceptions had been granted and if someone else nearby had been granted an exception already (for distance and safety considerations). Despite my best efforts, there were STILL cases that needed a person to decide, as automation could only SUGGEST a decision, not make it. This is what I believe is needed here – the algorithm can SUGGEST a level of funding, but then a person makes that determination (preferably after asking the intended recipient whether it seems reasonable). Without human oversight, an algorithm can deny a worthy application and allow an unworthy one. Edge cases are where systems fail, and this is where human oversight should be retained.
Counterexamples: —————————– I have children with NDIS plans. As a result of systemic thinking, they are considered separately and there seems to be no way to leverage the economies of scale provided by being in the same household. Given the genetic and hereditary nature of some conditions, being able to have flexibility to support the FAMILY and not the INDIVIDUALS could assist the NDIS budget, rather than hurt it, while still being of benefit to our children. Any system without human oversight is at risk of not noticing that possible advantage or assuming that advantage in all cases when it might not always exist. Same-ish inputs, different (desired) outcomes … how can an algorithm account for that? Processes set ahead of time do not allow for variation, and processes designed to adapt can develop in harmful ways without supervision.
With my children, if a plan change is made, I get a text telling me that “Recent changes were made to your NDIS account or to a NDIS account you have or had authority on. If you did not authorise them, please call us immediately on …” There is (or was) no indication who the plan related to, and logging into the portal there was no indication which had new updates. This requires me to make that phone call to work out what it was about because I can’t tell if I authorised it or not since I don’t know what it was. This wastes employee time (and therefore costs money) to ring and find out that it was a contact detail update (for example). I asked why there can’t be some indication in the message about what plan it relates to, and I was told that it was system generated and would be too difficult to change because of privacy issues. However, I got a text on another occasion telling me that a plan had been extended, starting with “Participant number: *****XXXX” (where XXXX were the last four digits) … this is all I wanted in terms to phrasing for the other messages! Why can it be done in one type of contact and not others? This seems to be automation in isolation without considering all the possibilities and is costing money - and it’s just a simple contact system. This does not inspire confidence in the idea of a funding algorithm being adaptable enough to anticipate every possible quirk and caveat.
Conclusion
—————————– I do not have an issue with trying to rein in costs where possible, we want the money to be spent better – but not at the expense of the people who need the support, and not by automation without oversight and the possibility of an informed override. We have seen examples of this in Services Australia already, as I’m sure other submissions will point out. I don’t feel the need to repeat them. From a systems analysis perspective, as well as a moral one, If algorithms must be used, let them inform decision-making, not be the decision-maker without redress.
Thank you for considering my submission.