Foodstuffs Facial Recognition Trial

UPDATE 2024-05-25: I received a full response from Foodstuffs, which you can read here.

On a recent visit to my local PAK’nSAVE (one of three major supermarkets in New Zealand), I was greeted by the following poster on entry:

PAK'nSAVE facial recognition trial poster.

It took me a few weeks to get around to it, but I decided to register my displeasure with this by emailing Foodstuffs (PAK’nSAVE’s parent company). This post exists as a public record of these communications:

To: privacy@foodstuffs.co.nz

Hello,

I am writing to you to express my objection to your trial of facial recognition technology in some of your North Island stores [1]. I have grave concerns over some current issues with this technology as well as how it may be used in the future.

My initial objection lies in the inherent dehumanisation of decisions about access to a store, which provides life’s essentials, being delegated to an algorithm. Furthermore, the technology will increase discrimination against some of the most economically vulnerable people in our society.

Currently available facial recognition technology is widely known to display racial and gender bias [2], misidentifying women and people of colour. In your proposed deployment this could quite easily lead to a higher proportion of people from these groups being removed or barred from your stores erroneously.

In addition facial recognition technology is largely untested on the New Zealand population mix [3], so we really have no idea how it will perform in practice.

Doubtless, the point will be made that the use of facial recognition in your stores is clearly labelled and that the privacy policy [4] makes it clear that people who do not consent to being captured by your facial recognition systems should not enter the store.

I would counter this point by saying: “what happens when this technology is everywhere?”. If facial recognition is deployed in every supermarket in the country, there is no effective way to opt-out (and buying food is not optional). In a supermarket duopoly it is not hard to foresee this situation.

In fact, this is already the case. Auror [5] is a security system that is deployed in 85% of NZ retailers [6]. Although it is not directly spelled out on their website it seems likely that the Auror system is utilising facial recognition technology.

If this is the case, then a follow up question arises: why does Foodstuffs feel the need to develop its own facial recognition system? The privacy policy (referenced previously) makes it quite clear that the system is currently only being used for security purposes. However, this does not preclude other uses in future.

One such use would be to implement dynamic pricing [7] (transcript of audio [8]) based on facial recognition and electronic price tags (which I note are already deployed in your stores). In effect this would mean offering different prices to different customers either based on large scale generalisations or more invasive profiling of individuals. This approach could rightly be
referred to as ‘algorithmic price discrimination’ and would be disturbingly Orwellian if deployed, particularly in light of the technological biases discussed above.

I hope that you will take the above into account in your evaluation of this technology and halt its deployment. Additionally, I would call on you to review your use of the Auror technology in your stores.

I will be writing to my MP (David MacLeod) to discuss this matter. I also reserve the right to publish all communication regarding this subject to my blog or social media, as I believe it to be a matter of public interest.

Kind Regards,

Robert Connolly
ME (Hons) Computer Systems Engineering and Concerned Citizen


References:

[1] https://www.foodstuffs.co.nz/news-room/2024/Foodstuffs-North-Island-begins-trialling-facial-recognition-in-select-stores
[2] https://sitn.hms.harvard.edu/flash/2020/racial-discrimination-in-face-recognition-technology/
[3] https://www.rnz.co.nz/news/national/502445/facial-recognition-government-rolls-out-new-tech-despite-racial-bias-concerns
[4] https://www.paknsave.co.nz/privacy-policy
[5] https://www.auror.co/
[6] https://supermarketnews.co.nz/news/solving-retail-crime-through-ai-tech/
[7] https://www.npr.org/2024/03/06/1197958433/dynamic-pricing-grocery-supermarkets
[8] https://www.npr.org/transcripts/1197958433?ft=nprml&f=1197958433

At the time of writing, I have not received a response to this. Any responses that I receive will be published in a future post.

UPDATE 2024-04-24: I received the following acknowledgement of receipt from Foodstuffs:

Dear Rob,

We confirm receipt of your email, and will respond to your queries in due course.

Kind regards,
Foodstuffs Legal Team

While I was at it, I also sent a variation of the above to my MP:

To: David.MacLeodMP@parliament.govt.nz

Hello David,

I am writing to you to express my concerns over the trial of facial recognition technology in some of Foodstuffs North Island stores [1] (notably the New Plymouth PAK’nSAVE store is included). I have grave concerns over some current issues with this technology as well as how it may be used in the future.

My initial objection lies in the inherent dehumanisation of decisions about access to a store, which provides life’s essentials, being delegated to an algorithm. Furthermore, the technology will increase discrimination against some of the most economically vulnerable people in our society.

Currently available facial recognition technology is widely known to display racial and gender bias [2], misidentifying women and people of colour. In the proposed deployment this could quite easily lead to a higher proportion of people from these groups being removed or barred from stores erroneously.

In addition facial recognition technology is largely untested on the New Zealand population mix [3], so we really have no idea how it will perform in practice.

Doubtless, the point will be made that the use of facial recognition in Foodstuffs stores is clearly labelled and that the privacy policy [4] makes it clear that people who do not consent to being captured by facial recognition systems should not enter the store.

I would counter this point by saying: “what happens when this technology is everywhere?”. If facial recognition is deployed in every supermarket in the country, there is no effective way to opt-out (and buying food is not optional). In a supermarket duopoly it is not hard to foresee this situation.

In fact, this is already the case. Auror [5] is a security system that is
deployed in 85% of NZ retailers [6]. Although it is not directly spelled out on their website it seems likely that the Auror system is utilising facial recognition technology.

If this is the case, then a follow up question arises: why does Foodstuffs feel the need to develop its own facial recognition system? The privacy policy (referenced previously) makes it quite clear that the system is currently only being used for security purposes. However, this does not preclude other uses in future.

One such use would be to implement dynamic pricing [7] (transcript of audio [8]) based on facial recognition and electronic price tags (which I note are already deployed in Foodstuffs stores). In effect this would mean offering different prices to different customers either based on large scale generalisations or more invasive profiling of individuals. This approach could rightly be referred to as ‘algorithmic price discrimination’ and would be disturbingly Orwellian if deployed, particularly in light of the technological biases discussed above.

I call on the government to pre-emptively ban the practice of algorithmic price discrimination as well as the profiling of individuals based on facial recognition data. I also call for restrictions on the use of facial recognition technology by commercial organisations, which seem inherently prone to abusing such a powerful technology.

I have communicated my concerns to Foodstuffs and am awaiting a reply. I also reserve the right to publish all communication regarding this subject to my blog or social media, as I believe it to be a matter of public interest.

Kind Regards,

Robert Connolly
ME (Hons) Computer Systems Engineering and Concerned Citizen


References:

[1] https://www.foodstuffs.co.nz/news-room/2024/Foodstuffs-North-Island-begins-trialling-facial-recognition-in-select-stores
[2] https://sitn.hms.harvard.edu/flash/2020/racial-discrimination-in-face-recognition-technology/
[3] https://www.rnz.co.nz/news/national/502445/facial-recognition-government-rolls-out-new-tech-despite-racial-bias-concerns
[4] https://www.paknsave.co.nz/privacy-policy
[5] https://www.auror.co/
[6] https://supermarketnews.co.nz/news/solving-retail-crime-through-ai-tech/
[7] https://www.npr.org/2024/03/06/1197958433/dynamic-pricing-grocery-supermarkets
[8] https://www.npr.org/transcripts/1197958433?ft=nprml&f=1197958433

At the time of writing, I have received the following automated response:

Kia ora,

On behalf of David MacLeod, MP for New Plymouth, thank you for your email. Please accept this reply as an acknowledgement that your correspondence has been received.

While David considers all correspondence to be important, and all messages are carefully read and considered, it is not always possible to personally reply to all emails. As such, the following guidelines apply:
· Invitations and meeting requests will be processed and a staff member will be in contact with you in due course
· All media queries will be responded to by a staff member
· If your email is regarding a constituent issue that you need support with, please ensure your full name and address is included
· If your email expresses a personal view, or is copied to multiple Members of Parliament, then your opinion will be noted and your correspondence may be transferred to another office, or there may be no further response to you.

Thank you for taking the time to write.

Ngā mihi nui

Office of David MacLeod
MP for New Plymouth

As above, I will publish any responses in a follow up post.

Leave a Reply

Your email address will not be published. Required fields are marked *

Bad Behavior has blocked 422 access attempts in the last 7 days.