Facial Recognition for Safety: Where Do We Draw the Line?
Freedom vs. safety. Both are conditions typically sought after by humans, yet we often find that they are at odds with another. With every bit of safety gained, we usually lose an equal amount of freedom, and vice versa. Finding a perfect balance between the two has been a longstanding struggle in human civilizations, and one that seems has not yet been resolved. This conflict has only been amplified with the recent explosion in technological progress, with modern smartphones being able to record and track your every location, call and text conversation. Governments have been able to use this technology to track down criminals and help keep us 'safe', but now it is harder than ever to stay off the grid. Now AI seem to be testing the boundaries further than ever before.
Just recently, London's Metropolitan Police Force (Met) has officially rolled out their live facial recognition system across London. This came after a multi-year trial period in South Wales where they installed special high tech cameras in targeted locations where crimes were most expected to occur. These cameras have access to a watch list with names and faces of wanted criminals. The Met decided to implement this system all across London after seeing promising results. So the trial period proved to be a success, helping to catch dangerous criminals and clean up the streets. That's great! So why would some people (myself included) be wary of this system? Well for one, the facial recognition technology is not 100% accurate. Certain reports even indicate it was only accurate 19% of the time. Some also worry that the technology might be biased towards certain groups, due to possible biased data sets that the AI used was trained with. The Met's response to these concerns is that being recognized by the system isn't an end-all where you'll be arrested on the spot. It will simply suggest to an officer that this person might be who you're looking for, and its up to them and their judgement to confront and contain them. This seems awfully similar to stop and frisk policies, however. With a system that is only accurate 19% of the time, it sounds like a lot more innocent people will be confronted and detained, which will only increase tensions between the public and the police.
If technologies such as these are going to be integrated effectively and ethically into our societies, it has become clear that strict and specific rules must be developed, especially with how rapid technologies advance. There are many potential problems that could arise if we are not careful and precise with a system like this. One example is the possible racial bias that was mentioned above. It should be mandatory for a continuous tracking of discrimination risks to be implemented at all levels of development. I think it should also be illegal for these systems to record data on everyday citizens that are not on the watch list. To allow otherwise would essentially let the police (and government) track our whereabouts everyday, which would feel like too much of an infringement on our personal freedoms, even if it would produce a 'safer' society.
If we don't put checks on systems like this right now, it can very well become abused in the name of safety. Take the situation in China, for example, where they have over 170 million facial recognition cameras dispersed throughout the country. Their law enforcement uses them to quickly catch criminals, but it is also common for people who were wrongfully flagged to be arrested. Even worse and seemingly pulled straight out of Black Mirror episode, the cameras are also used to regulate a Social Credit system that gives you ratings on things such as what you do and where you visit. If the cameras catch you doing something unfavorable or inconsistent, your score will lower which can impact the types of loans you can get as well as many other life impacting consequences. Scarily, these cameras were first implemented for law enforcement purposes only. What I fear about implementing technology like this in the US is that despite how much security it brings to us, it might be laying the groundwork for something much much worse. I'm not convinced the increased safety it brings us is worth the risk of the freedoms we could lose.
Just recently, London's Metropolitan Police Force (Met) has officially rolled out their live facial recognition system across London. This came after a multi-year trial period in South Wales where they installed special high tech cameras in targeted locations where crimes were most expected to occur. These cameras have access to a watch list with names and faces of wanted criminals. The Met decided to implement this system all across London after seeing promising results. So the trial period proved to be a success, helping to catch dangerous criminals and clean up the streets. That's great! So why would some people (myself included) be wary of this system? Well for one, the facial recognition technology is not 100% accurate. Certain reports even indicate it was only accurate 19% of the time. Some also worry that the technology might be biased towards certain groups, due to possible biased data sets that the AI used was trained with. The Met's response to these concerns is that being recognized by the system isn't an end-all where you'll be arrested on the spot. It will simply suggest to an officer that this person might be who you're looking for, and its up to them and their judgement to confront and contain them. This seems awfully similar to stop and frisk policies, however. With a system that is only accurate 19% of the time, it sounds like a lot more innocent people will be confronted and detained, which will only increase tensions between the public and the police.
If technologies such as these are going to be integrated effectively and ethically into our societies, it has become clear that strict and specific rules must be developed, especially with how rapid technologies advance. There are many potential problems that could arise if we are not careful and precise with a system like this. One example is the possible racial bias that was mentioned above. It should be mandatory for a continuous tracking of discrimination risks to be implemented at all levels of development. I think it should also be illegal for these systems to record data on everyday citizens that are not on the watch list. To allow otherwise would essentially let the police (and government) track our whereabouts everyday, which would feel like too much of an infringement on our personal freedoms, even if it would produce a 'safer' society.
What do you think the chances are of a Social Credit system being implemented in the US someday are?
ReplyDeleteDoes China's massive surveillance system (including facial recognition) make it better able to respond to a crisis (like the current SARS-CoV2 pandemic) than places like the U.S. that value individual privacy/freedom more?
ReplyDeleteIt did! Personally I feel a lot more safer in China than US. I am afraid of walk out of campus at night in US, but in China I know I will be safe at any public place (street, market, etc.) anytime. It seems not that helpful for pandemic though since the camera can not recognize patients. But the centralized power did helped in COVID, the isolation measures works better in China than most other country.
DeleteI think it would help with COVID-19. Let's just imagine that the U.S. have such system installed. If one person is diagnosed with COVID-19, the camera system would probably be able to trace back where this person have been and who they have interacted with, thus enable the ability to quarantine potential infectious sources.
DeleteI think another thing to keep in mind is how much surveillance there is already in the United States. There are security cameras in buildings as well as outside of buildings (like stop-lights). I think AI might be good at analyzing the devices that we have in place already, but considerations must be made as you have stated.
ReplyDeleteAdding on to this, is it morally okay to collect this data on cell phones? With most agreements being signed off as to being read without even a glance over, companies could be selling camera data with a large portion of cell phones using facial recognition now. This data provides a huge amount of recognition for the population, but is it crossing a line?
DeleteIt's crazy to think that even AI can have racial bias/discrimination. Your post reminds me of the book we read in Clauss' DATA100 class called "Weapon's of Math Destruction" that talks about a policing software that predicted where crime would occur within a certain city. The model was being reinforced with bad data and was sent into a negative feedback loop to the point where the model was racially biased to where it would predict crimes. I could definitely see something like that happening with a facial recognition model which I think would be one of the biggest concerns with employing something like that in the United States.
ReplyDeleteDo you have any proposal to how the 'rules' for the implementation of facial recognition technology?
ReplyDeleteWow, 19% does not inspire any confidence.
ReplyDeleteThis article reminds me of the "killbot" video that Dr. Stonedahl shared earlier in the term... The technology is amazing but it can be weaponized and made to do harm if it gets into the wrong hands. The implementation of AI in government tracking is super sketchy in my opinion, but there are pros and cons to everything.
ReplyDeleteFreedom vs. safety. I personally feel that a lot of people values freedom way too much because the government is always there to blame when some thing unsafe came to us. Recently, many people are protesting stay-at-home order, making crowds and taking the risk of contracting COVID-19. Do you think it is time to give up some freedom in order to gain more safety (not only for us, but also for our beloved)?
ReplyDeleteI am willing to sacrifice some freedom, but I must have my safety in return. I don't think the government would violate the freedom of ordinary people like us. So as long as we don't break the law or commit a crime, we should be fine.
ReplyDeleteHowever, the data of people face must be extremely secured. They should never buy and sell this kind of data and use it under any commercial form, for example marketing purpose.
This raises an interesting debate: safety vs. privacy. Privacy is recognized as a human right in the constitution, so that might mean that it potentially holds similar weight to life, another human right. How much are people willing to sacrifice to ensure their safety? Or are they willing to sacrifice safety for the sake of their rights?
ReplyDelete