While DoHDA may not be the worst offender, the IAT and I-CAN tools are examples of ADM that need to be better explained to the public.
A new report from the Office of the Australian Information Commissioner found that the use of automated decision making in 23 government agencies was muddied by a lack of transparency about how ADM was used.
Automated decision making is defined as “the use of technology, commonly referred to in Commonwealth legislation as a ‘computer program’, to automate a decision-making process”.
While the report does not name the government agencies it scrutinised, the Department of Health, Disability and Ageing has been facing criticism for its use of ADM tools in assessments in the aged care system – the Integrated Assessment Tool – and in the NDIS – the I-CAN tool.
The OAIC report identified 23 government agencies that have statutory authorisation to use ADM. It reviewed each of the identified agencies’ websites, AI transparency statements, Information Publication Scheme agency plans and performed keyword searches to try to identify whether agencies disclosed if they were using ADM.
“Our threshold was whether a member of the public, who wanted to know if an agency was using ADM, could reasonably do so by performing relatively simples searches on the agency websites,” said the report.
Although every agency published IPS information on their websites, there were other findings of concern:
- Just 17% (four) of the agencies authorised to use ADM disclosed that they use ADM in a decision-making process in their IPS information;
- 39% (nine) of the agencies (in addition to the four above) made reference to ADM in their IPS information but did not confirm that they used it;
- 43% (10) of the agencies did not mention ADM in their IPS information;
- The agencies that disclosed in their IPS information that they used ADM were not clear about how they used it; and
- None of the agencies had any published guidelines or policies on ADM.
In a case study published by the OAIC, the Department of Health, Disability and Ageing was found to have a website which for “automative administrative processes” which provided some basic information on ADM, the delegated person that authorises ADM, and some brief examples of where ADM could be used.
The OAIC described that as a “good practice” which “could be considered for wider adoption by other agencies when disclosing ADM to the public”.
Related
One unnamed regulatory agency stated in its AI transparency statement that it did not use AI in compliance or decision-making, but the OAIC found the agency used automated decision making to calculate fees on its online portal.
Another agency mentioned ADM in its data strategy, saying it was “embracing automation and artificial intelligence”, but the agency in question did disclose on its website if and how it used ADM.
The OAIC made four recommendations:
- All agencies authorised under legislative statutes to use ADM should publish this information as part of the IPS. This should include the statute that grants them this power, and whether or not they utilise ADM to provide information and services to the public;
- As part of their IPS, agencies should clearly state the types of ADM they use to make automated decisions, not just AI (i.e. from simple calculators to machine learning);
- Agencies that use ADM should publish, as part of their IPS, both a list of decisions they use ADM for and relevant and easy to understand examples, so the public can better understand how it is being utilised to make decisions that affect them; and
- Agencies that use ADM should publish, as part of their IPS, any policies that clearly set out the principles for when and how they use ADM to make decisions and/or recommendations affecting members of the public.
Read the full report here.
