Retail facial recognition trials in Australia spark accuracy, privacy and bias concerns

Facial recognition is now being trialed in retail in Australia to curtail crime and streamline the customer experience, in addition to its use by immigration and law enforcement agencies. The wider implementation of the technology, however, is raising questions about consumer privacy, algorithmic bias, and cultural concerns.

Despite the expanding use cases, many are worried about the technology’s inaccuracies in identifying people of color in particular. “The only safeguard [markets] have promised is no decision will be made to exclude anyone other than two staff members independently… checking that the matches are in fact accurate,” says Gehan Gunasekara, associate professor at the University of Auckland business school in an interview with ABC Australia’s Pacific Beat.

The problem with this is that “people tend to trust technology more than their eyes, and you’d have to be a very skilled individual to be able to tell that the biometric identification is inaccurate.”

Moreover, while it might seem promising to improve security by identifying criminals, “the objectionable part of it is everyone’s face will be captured because that’s the only way you will be able to screen somebody undesirable,” he explains.

He also notes that while stores promise to immediately delete data from those who are not shoplifters, such a commitment needs to be monitored. Indigenous groups also have cultural practices surrounding the capturing of their own faces that need to be considered.

Similar concerns about the impact of facial recognition in retail on indigenous people have been raised in neighboring New Zealand. Gunasekara notes that Australia’s privacy commissioner is closely following trials there.

Corsight pitches wide-ranging benefits

Facial recognition can be used to prevent retail theft by identifying repeat offenders as well as identify common scenarios that indicate a shoplifting or burglary risk, facial intelligence provider Corsight AI explains in an article for Security Journal Americas.

For instance, someone may return to the same store with high frequency in a short span of time. Two suspects may be seen cooperating or one may appear to interact with a staff member. Someone may wait outside during closing hours. Catching scenarios like these can help security teams limit losses by shortening reaction times.

Systems with facial recognition capacities can also be used to monitor employee presence for time and attendance management and authenticate them for access to restricted areas. Whitelist alerts could be set to be triggered when an unauthorized person is present in a restricted area.

Customers can also enjoy a more convenient shopping experience by using facial recognition to manage queues and adjust staffing to accommodate busiest times with the longest queues and minimizing staffing at quieter times.

By identifying loyal customers who are not necessarily customer loyalty members, stores can offer faster or preferred service to them, maximizing sales. Facial recognition systems could also track the age and gender of shoppers to facilitate insights.

Article Topics

Australia  |  biometric identification  |  Corsight  |  criminal ID  |  facial recognition  |  retail biometrics