Is it time to Face Up?
I like to see myself as an 'early adopter' of new technology. Still, there are times when I have my doubts. One of those times is when I think of our pensions.
Have you been following the debate over the UN Pension Fund’s decision to adopt facial recognition technology? If not, read the article we posted last week by Elaine Fulz in UN Today (with thanks to Robert Cohen and Loraine Rickard-Martin for forwarding). * Also see ‘How We Lost Control of our Faces’ from the MIT Technology Review which we posted on February 9th.
As Loraine Rickard-Martin explains in her message, announcements by UNJSPF of this ‘convenience for retirees’ make little mention of the risks of mis-identification based on race and gender, the loss of privacy and the potential misuse of data.
We might add, ‘... and the loss of our faces,’
Some would argue that we lost ownership of our faces a long time ago. Drivers licenses, ID cards, passports, visas and many other government documents have long demanded our faces. More recently we have voluntarily given up our faces to Facebook, Google, Instagram and many other social-media sites. Those photos are used in ways we might never have expected. They are traded and sold (yes sold), hacked, morphed, manipulated, sometimes anonomyzed, and added to artificial intelligence databases.
The result - a world of faces - some real and others 'created' by artificial intelligence. Can you tell a 'real' face from a 'created' one? Find out here - Which Face is Real?. Could you tell the difference? Now, imagine this technology in the hands of a hacker or an intelligence service.
Whether we like it or not, there is little doubt that facial recognition is a fast-growing fact of our lives.
Still, the fact that our pension fund now wants to adopt facial recognition ought to give us some pause. Use of the technology may have grown rapidly, but it is still in relatively early days. Racial, ethnic and gender bias remain major issues, partly because the existing models used for ‘training’ systems were based on narrow models - often on mostly 'white' faces.
How are such systems ‘trained’? One example happened here in Colorado Springs in 2013. A professor at the local campus of the University of Colorado set up a hidden campus to photograph anyone crossing a public plaza. In all, he took 16,000 photos of people without their knowledge and identified 1,700 of them by matching to social media and other sources. Only last year did the public learn that the project had been funded by US intelligence and military agencies anxious to test their facial recognition algorithms. This was just one of many secret experiments carried out in cities around the world.
The major adopter of the technology so far has been China which has moved rapidly ahead in building a massive database of photos and biometric data of its citizens. Law enforcement organizations elsewhere have also been early adopters of the systems. There have been major hiccups along the way with false arrests, often involving algorithms wrongly identifying ‘non-white’ suspects.
Even if a harbinger of 'big brother' governments, the use of facial recognition for law enforcement may be obvious and at times beneficial to public safety.
But why would a pension fund want to adopt facial recognition?
Any pension system needs some ‘proof of life’ mechanism. Without such a mechanism, pensions could continue to be paid out long after the beneficiary has died. Fraudsters and family members could receive benefits not due to them, draining funds due to those really eligible.
The current system has its faults. UNJSPF sends out certificate of entitlement forms by post for us to sign and return each year. The system is at risk from delays, lost mail, and occasional fraud. Moreover, when a beneficiary falls ill, there is the risk that loved ones are unaware of the steps needed certify continued eligibility.
Aside from the UN, other public pension funds have been slow to experiment, much less adopt facial recognition as their ‘proof of life’ tool. In the US this reluctance reflects a strong negative reaction among the public. The irony here is that the expansion of such systems in countries like China has often been financed by western investors and US tech-giants anxious to develop and sell their systems to governments.
Search on the internet for something like ‘facial recognition by pension funds’ and you will find many upbeat articles from tech publications about the leading role the UN is taking with its pension system in adopting facial recognition, blockchain and artificial intelligence. The articles note the daunting challenge - a pension system covering over 70,000 pensioners living in 195 countries. The message is clear - if the UN can adopt these systems for such a dispersed global membership, why would other smaller funds not take a similar leap? In short, good advertising.
I am pleasantly surprised to find the UN taking a lead in anything, but why this? Is the UN Pension Fund, so recently unable to deal with simple matters like processing new retirements, really ready to lead the world in certifying pension eligibility?
New tech systems are costly and often take years to settle in. Consider the many past experiments in ‘modernizing’ the UN’s data systems. Consider the many times UNICEF decided to adopt new financial and programme management systems. In the end they may have helped, but along the way they caused considerable pain, extra work and delay for all concerned.
And at what cost? In the long-run such systems may save money, but they require big up-front investments. Where do the investments come from? - from funds set aside by and for us, the retirees, and those who will follow us.
Oh - and did anyone ask us?
* An update to the article by Elaine Fulz was posted on the UN Pension blog on 5 February and includes a response by Rosemarie McClean, Chief Executive of Pension Administration. While I am happy to see UNJSPF responding to some of the concerns raised, I remain concerned. Should the UN be a leader in this difficult arena? I would prefer to let some big commercial firm take the risks. Once the technology is proven, I would be more confident in letting UNJSPF follow.
Your concerns, Tom, about the facial recognition initiative by the UNJSPF brought out an issue that I was not aware of. I had missed the articles by Elaine Fultz and Deborah Raji and only read them now, as you recommended. Very well researched and certainly raising many questions.
ReplyDeleteI started the process last week and scanned my driver's license and took my photo. (I admit that I was concerned about the quality. The process went relatively fast.). I now have a video call appointment to verify ,my data on April 14.
I had to install a specific app for the Digital Certificate and it only works with my iPhone.
To answer Ms. Raji's questions: Since I applied, I obviously consented. Is data meticulously verified? I don't know, neither do I know if thorough documentation was required.
One thing I don't understand is how this system would use up so much energy (- Ms. Fultz mentions 14 gall of gas for one transaction). Maybe a better idea might have been to use the secure email system like the Credit Union and Cigna use.
Anyhow, I am in the loop - for better or worse...