Strongly supportive
In early February, Transparency International New Zealand (TINZ) prepared a submission supporting the work of Stats NZ to create an Algorithm Charter. The charter aims for a more consistent approach by Government agencies in their use of algorithms in “a fair, ethical and transparent way.”
“We strongly support the primary purpose of the charter – to put the right safeguards in place when algorithms are used so that data ethics are embedded in the work, and so that the people and communities being served by these tools are always front of mind,” says Julie Haggie, Chief Executive Officer of TINZ in a related media release.
Critical areas for improvement
TINZ would like to see the Charter go further, in four key areas:
- Include a commitment to cross agency engagement on matters such as algorithmic literacy, frameworks for algorithmic hygiene, governance frameworks, whistleblowing, peer bias assessments and public consultation.
- Extend the scope of the charter beyond people, to include the data related to natural systems.
- Include algorithm use or development by external organisations for any provision of public services or projects.
- Include a commitment by all government agencies to apply their knowledge in support of the development of industry standards in algorithm use.
Significant risks
The use of sophisticated data modelling puts a lot of power in the hands of a few. Background reports and international research shows the risks of using algorithms. These risks include unintended bias, persistent error, breaches of privacy and restriction of choice. It is critical that there are systems in place to monitor and review this use, to make sure it is fair, unbiased and transparent. Also, researchers, journalists and civil society groups have an important role to play in shedding light on the use of data, as has been evident overseas.
Trust
There are different levels of transparency. The general population may not find use in seeing raw anonymized data, and may not understand the associated algorithm itself. But they do need to have trust in the fair use of algorithms that particularly apply to them. Transparency International supports the use of a principle called the ‘right to explanation’. Using this principle, organisations can show what the outcome would have been for an individual if they had had certain different attributes. This allows people to understand a decision and to challenge it if they think it is not fair.
TINZ also strongly supports the inclusion of digital literacy within a general framework of literacy. Students should be encouraged to understand what algorithms do and how they are used in both operations and analysis.
Private sector critical to effectiveness
The Stats NZ proposed Charter applies to government agencies, but the main use of algorithms is in the business sector. Examples of this include in recruitment, financial services, product marketing, social media and the private health industry. TINZ calls on the business sector to show a similar commitment as that provided by government agencies, including developing and improving industry standards.
