New Facial Recognition Technology Raises Concerns Over Racial Bias
The New Zealand government is set to roll out new facial recognition technology despite concerns over potential racial bias. The technology, developed by the Department of Internal Affairs (DIA) over a period of four years, has not been tested for bias on New Zealand’s specific population mix. This untested risk has raised alarm bells among critics who fear the technology may disproportionately affect Māori and Pasifika communities, who make up a significant portion of beneficiaries.
The facial recognition tool, called Identity Check, utilizes technology from Irish company Daon to match a live image taken on a phone with the individual’s driver’s license or passport photo in government databases. The Ministry of Social Development (MSD) will be the first to implement the technology on November 20, giving beneficiaries the option to use it or stick with existing verification systems.
However, a recent MSD report highlighted the unknown level of racial bias in the technology, labeling it untested and unconfirmed. The report emphasized the potential impact on Māori and Pasifika clients who could be subjected to fraud investigations due to any bias in the facial recognition algorithm. The report also cited US government tests in July, which suggested that facial recognition algorithms, including Daon’s AI algorithm, worked less effectively on people with darker skin tones.
Despite the concerns, DIA argues that racial bias is not an issue based on recent tests that showed the tool to be 90 percent accurate. While DIA has not specifically tested the algorithms against different ethnic groups, they believe the technology is successful in the New Zealand context.
Joy Liddicoat, an AI researcher at Otago University, criticized the rollout of the facial recognition system, stating that it should have addressed these well-known dangers before being launched. Liddicoat acknowledged the aim of making access to benefits easier but expressed concern about potential discrimination and emotional harm to Māori and Pasifika individuals who may face challenges in using the technology.
MSD defended their decision, stating that the risk of racial bias was acceptable due to ongoing efforts to consult ethnic groups and improve the technology. They also mentioned that the facial recognition algorithm used by DIA would be retrained to enhance performance among the New Zealand population.
The government’s plan is to make Identity Check the primary technology for verifying identities online, with the goal of improving access to services and e-business in a secure manner. However, concerns over racial bias and the need for comprehensive testing persist. The government must prioritize addressing these issues to ensure a fair and equitable implementation of the technology across all communities in New Zealand.
