---
id: "workshop:guidelines-for-promoting-privacy-in-facial-recognition-technology-"
type: "InnovateUS workshop"
title: "Guidelines for Promoting Privacy in Facial Recognition Technology"
url: "https://innovate-us.org/guidelines-for-promoting-privacy-in-facial-recognition-technology-"
date: "2024-04-30"
duration: "60 minutes"
language: "English"
instructors: ["Vance Ricks"]
recording: "https://www.youtube.com/watch?v=u7XkcRPJROk"
transcript: "automatic, from the recording"
source: "InnovateUS (innovate-us.org)"
licence: "CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org"
---

# Guidelines for Promoting Privacy in Facial Recognition Technology

InnovateUS workshop · 30 April 2024 · 60 minutes

Join us for this 60-minute workshop with Vance Ricks, associate teaching professor in the departments of Philosophy and Computer Science at Northeastern University, for a workshop on the use of facial recognition technologies and privacy. In this training, participants will learn about different kinds of facial recognition technologies and other forms of biometric identification, and will discuss some of the main ways of thinking about privacy and obscurity with the use of these technologies. Participants will learn about the challenges that facial recognition technologies pose to privacy and obscurity and will understand how governments can articulate clear and robust guidelines for the use of facial recognition technologies, in order to safeguard privacy.

## Learning goals

- Learn about different kinds of facial recognition technologies and other forms of biometric identification
- Learn about the challenges that facial recognition technologies pose to privacy and obscurity
- Understand how governments can articulate clear and robust guidelines for the use of facial recognition technologies, in order to safeguard privacy

## Speakers

- **Vance Ricks**, Teaching Professor, Northeastern University

Recording: https://www.youtube.com/watch?v=u7XkcRPJROk

## Transcript

Automatic transcript of the recording: names and terms can be misheard, and the times can be a few seconds off. Each time links to that moment of the video. The transcript alone: https://innovate-us.org/guidelines-for-promoting-privacy-in-facial-recognition-technology-.txt (plain text) and https://innovate-us.org/guidelines-for-promoting-privacy-in-facial-recognition-technology-.json (every segment with its start and end second).

**Unknown speaker**

[0:00](https://youtu.be/u7XkcRPJROk?t=0) Spotlight, you first.

**Vance Ricks**

[0:04](https://youtu.be/u7XkcRPJROk?t=4) That's me. Are we ready?

**Unknown speaker**

[0:08](https://youtu.be/u7XkcRPJROk?t=8) We're ready. Okay.

**Vance Ricks**

[0:12](https://youtu.be/u7XkcRPJROk?t=12) Thank you everyone for coming to this presentation. I'm going to go ahead and share my slides. What I've tried to do is prepare roughly 20 to 30 minutes worth of stuff for me to say to you, and then the more interesting and helpful part for me is going to be the question and answer period. So I hope you won't feel shy about jumping in with questions. So let's go ahead and begin. So as Hannah said in her introduction, I have a joint appointment at Northeastern University in Boston. That's the main campus. It seems to be springing and spreading up campuses in different places every other year or so, but I'm at the main campus. I'm also, as she mentioned, an editor and contributor to the Mozilla Foundation's Responsible Computing Playbook. This is a set of guidelines for instructors K through 12 and above who are interested in any sort of questions relating to how do they teach, broadly speaking, questions about the ethics of technology or the ethics of computer use to students. How do they learn that stuff? How do they get their principals or their deans to agree that those kinds of programs are ones that they should be running? I want to make clear also, I'm coming at this from the perspective of someone with a foot in philosophy as a discipline and a foot in computer science as a discipline. I don't have a background in the law or legal theory or in public policy directly. So I have a lot to learn, and that's part of why I was interested to be able to meet you all today for this event. This slide shows basically an overview of what I'm going to discuss this afternoon. A quick overview of what facial recognition technologies are because that's a sort of broad term that it turns out covers a bunch of different things. But the main ones that

[2:12](https://youtu.be/u7XkcRPJROk?t=132) I think people are concerned about are the ones that specifically seem to bear on questions of privacy and obscurity. So then I'm going to talk briefly about what those things are and why we might care about them and the ways in which certain kinds of facial recognition technologies raise challenges to privacy and to obscurity. Then I'm going to say some very brief things, and this is really more the part where I hope you all will supplement and fill in, when it comes to the articulating of guidelines for the ways in which people in public office and government agencies ought to think about their uses of facial recognition technologies. So when we think about facial recognition technologies, as I said, this is a kind of umbrella term and really tends to be ambiguous because there are three different things people might mean. There are facial detection technologies, there's face verification technologies, and then there's facial identification technologies, which are also sometimes called pair matching. Facial detection technologies are the ones that, in essence, we put into things like so-called self-driving cars or other systems that don't have visual systems in the same way that we have visual systems. And so we need them to be able to, when they're getting information about what's out there, to be able to tell different objects apart from each other. So you could imagine if you're a car, you might need to know what a stop sign is or whether you're approaching a stop sign. So when we talk about face detection technologies, what we're really talking about is a set of technologies that just say, okay, there's a bunch of things in my visual field.

[3:51](https://youtu.be/u7XkcRPJROk?t=231) Are any of those things a face or not? A different kind of thing that's under the umbrella term of facial recognition technologies is what's called face verification technologies. Some of you may be personally familiar with this kind of technological system because you might have it if you have a smartphone, whether it be an Android-based phone or an iOS-based phone. You can use this to unlock your phone similarly to the way that you might be able to use your fingerprint to unlock your phone, or you might have a six-digit code that you have to type in to unlock your phone. Face verification technologies work very similarly to the model of getting someone's fingerprint. That is to say that when you set them up, those of you who have this already know this, so you can sing along with me, there's, in effect, a kind of scanning that's done of your individual face and the different features of your face. It's mapped in such a way that the map is then stored on the device. Then when you show your face to the camera the next time around, it compares what it's seeing in the camera to what's stored in the device, similarly again to the way that it would compare your fingerprint now to the way it stored some prior image of your fingerprint. Then you can use it to unlock the device or maybe to get into a safe room or something else that you might need to be able to access. Then the third kind of thing, and this is, I think, going to be the focus of most of the rest of what I'm going to be talking about, is what are called face identification technologies. Face identification technologies are not just asking is this thing a face or even is this your face so I can let you into this room or let you into this phone, but there's a bunch of faces out here. Which one of them, if any, is yours or what's the probability that one of these faces is yours? Maybe I'm trying to identify someone in a crowd of people. That could be a reason that I'm using this particular kind of face identification

[5:52](https://youtu.be/u7XkcRPJROk?t=352) technology. Amazon sells a system that it calls Recognition, which has this capacity. Clearview AI, which is a product that many of you may be familiar with, either from personal acquaintance or just because it's been in the news a lot, is another one. PimEyes is a product that offers what they call a reverse face lookup. You can upload an image of a person's face, any person, and it will then be able to tell you, to some degree or other, who that person actually is. The way that facial identification technologies work is similar to the ways that we know about other kinds of things like large language models. There's some sort of training that you give it, whereby you give it a whole bunch of faces, images of faces, or perhaps even not still images, but maybe videos. Then features are extracted, stored in a database, so that then when you present it with some face in the future, it's able to come up with some degree of probability that the face that it's seeing right now matches one of the faces or some of the features of the face that are in the database based on the training that it's had before.

[7:10](https://youtu.be/u7XkcRPJROk?t=430) So, I talked about training. Where do the faces to do this training actually come from? Typically, there are three different answers to that question. A lot of them come from what are called publicly available sources. I put publicly available in quotation marks for reasons that I'll come back to in a moment, but this could include social media accounts where people have uploaded photos or videos themselves. It could include things like YouTube video or photos. It could include employee databases or someone's blog. There's actually a few databases that are publicly available specifically for the purposes of training these models, so that's one prime source of them. A different one is state-issued identification, so agencies like the Department of Motor Vehicles or whatever that equivalent agency might be called in your state, when they take a driver license photo, that's stored somewhere, and a lot of those photos end up being used later for facial identification purposes. Many public universities make IDs of their employees and of their students. Those also have been sources for the training of facial recognition systems or facial identification systems. More lately and more recently, some companies, and Microsoft is among them, have taken to generating faces, so these are not the faces of any actual existing human beings, but they are synthetic faces that are then used to train facial identification systems to be able to then identify actual human faces.

[8:46](https://youtu.be/u7XkcRPJROk?t=526) We're talking typically on the order of hundreds of thousands, if not millions, of images, and many of them could be images of the very same person taken from different angles, taken maybe one day when it's cloudy, a different day when it's sunny, taken from the side or taken from the front, different videos of that person, maybe one day they're wearing a hat, maybe a different day they're not, so even the size of these databases may or may not mean that there are a whole lot of distinct individuals, although it typically does. Now, I want to talk briefly about privacy and obscurity. Privacy, it turns out, as you can imagine, as probably you all have thought about as well, can be thought of in a few different ways, so I'm presenting here a few different ways that people have thought about privacy along with the ways in which they have said it's important or the reasons for which they value it. One way that we think about privacy, and you might say in some sense this is the original way, is as a kind of physical or symbolic boundary, so when I say that this is a private matter or when I say this is private property, there is either symbolically or actually a kind of boundary that's erected between this thing that I'm saying is private and you, that you're not allowed on this side of the fence or you're not allowed to have a certain access to something, some bit of information, for example, or some area or some relationship.

[10:18](https://youtu.be/u7XkcRPJROk?t=618) A different way of thinking about privacy has been articulated over the last couple of decades by a philosopher named Helen Nissenbaum. She's an American philosopher who teaches at New York University. The model that she uses is what she refers to as contextual integrity. Her idea is that look, we don't really, when we think about privacy, think about not making it possible for anyone to know anything about us. Rather, when we think about privacy, what we really mean is something like making sure that the right people have access to the right information in the right ways. So for instance, if you go to the doctor because you're not feeling well and the doctor asks you what's wrong and you reply that's none of your business, you're not doing it right. The idea is that in order for me to be able to get treatment, there are some things about myself that the doctor is going to have to find out. What I care about is not that the doctor has that information, but rather that the doctor keeps it confidential. And so we have law and policies and norms governing the use and collection and transmission of medical information. But we don't say that information should never be collected ever. We say, well, of course it should be collected because there's a genuine need that it's meeting to have that data collected or that information collected. What we care about, again, is it would be one thing if the doctor collected the information. It would be a different thing if the doctor then sold it or put it on a billboard or blasted it out. So those are the things, according to Nissenbaum, when we think about privacy that we really care about. That the information that's collected in a particular context stays within that context or that the rules that explain who is allowed to get that information and what they're allowed to do with it are adhered to or that they be adhered to.

[12:05](https://youtu.be/u7XkcRPJROk?t=725) So again, it's not a prohibition on the collection of information, but rather on how it's transmitted, who collects it in the first place and what they can do with it. A third way of thinking about privacy is actually quite skeptical of privacy as a concept. What it suggests is that when we talk about privacy, we're using that term as a kind of shorthand to talk about a whole bunch of other things that we actually care about. And privacy just becomes one convenient way to talk about a whole lot of different things at the same time. So we might be talking about control over your information or control over data that may be about you in some way. We might be talking about seclusion, right? So maybe someone might say, I need some private time. And what they mean is I need to be apart from the rest of you. It doesn't mean that the rest of you can't see where I am or even that you can't see what I'm doing, but it means that I'm just not interacting with all of you right now. I'm set apart in some way. Sometimes that's what we care about when we talk about privacy. Other times what we care about is that we're being monitored or surveilled in some way or other.

[13:07](https://youtu.be/u7XkcRPJROk?t=787) That may be for good purposes or for bad purposes. It may make us feel cared for. Maybe if we're a toddler, right, maybe we want our parents to do pretty intensive monitoring of our whereabouts. But if we're maybe a little bit older than that, it feels quite intrusive and maybe demeaning. A different thing that some people say we mean when we talk about privacy is what's called aggregation. Aggregation here is meant to refer to the fact that a lot of data can be collected about an individual. And seemingly unrelated bits of information about an individual can be assembled to form what seems like a more complete picture of a person. So that concern is not so much about whether someone's observing me or whether I don't have any private time to myself. It's more about the overall effects of a huge database consisting of information about me or images about me or other things relating to stuff that I consider personal in some way. Partly for the reason that privacy is contested in these different ways and it's not always clear that people are talking about the same thing when they're talking about privacy.

[14:16](https://youtu.be/u7XkcRPJROk?t=856) There's a group of scholars, particularly from the legal field, who have suggested that what we should be talking about is what they call obscurity. So one reason that they make this case is that, as I mentioned a moment ago, they think that the term privacy is just trying to do too much work. It refers to too many things that are related but different and that it would be more useful to pull those things apart. The other concern that they raise, however, is that in an age of big data where, as I just mentioned, it's pretty trivially easy to collect a lot of information about ourselves and each other and it's trivially easy for computer systems to crunch that data and find interesting correlations and patterns, that as a practical matter it's very difficult not to To leave a trail of digital exhaust wherever you go, if you are in any way online, and even if you're not online, if people in your life are online, it's very hard to avoid leaving some trail of digital exhaust. So, what these scholars suggest as an alternative to debates about privacy is instead thinking in terms of obscurity.

[15:21](https://youtu.be/u7XkcRPJROk?t=921) Obscurity, as the little graph here is meant to indicate, is a kind of function of or sort of point on two different axes. When we say something is obscure, part of what we're implying is that either it's hard to get to or it's hard to understand. So, I have a couple of potential examples here to illustrate what these scholars are talking about by obscurity. Take something like your physical address, your home address. It might not be very difficult for someone who wants that information to find it. Perhaps they could use their favorite search engine. In olden times, there was a thing called a telephone directory that you could use to look it up. I mean, people just put their physical addresses just in a book that any old person could pick up and read. So, in that respect, it's very easy to find that information and to understand what it is that I'm looking at. So, I understand how streets work. I understand what numbers mean and what addresses are. So, once I have the information, it's pretty easy for me to understand what it actually means.

[16:26](https://youtu.be/u7XkcRPJROk?t=986) There might be something similar when it comes to inside jokes, where here it's pretty easy maybe to overhear somebody telling a bunch of their friends some sort of joke. You don't get the joke. So, in that respect, it's obscure to you because, although you hear the joke and you see these people laughing, you don't understand what they're laughing about. You don't get the reference or you don't understand why the reference is funny. To the people in that group, they understand perfectly well why it's hilarious or why it's so amusing because it maybe refers to shared experiences or a shared perspective. So, in that respect, again, it's not that they're hiding from you the fact that they're telling a joke that's cracking each other up. It's that you can't understand what the joke is or even why it's a joke. Further along, we see things like maybe your email account password. If you followed all the guidelines that our IT professionals and cybersecurity professionals tell us, right, that should be something that's relatively difficult for other people to find. But once they crack it, suppose they do, it might not be difficult for them to understand, oh, this is a password or this is an email password specifically.

[17:31](https://youtu.be/u7XkcRPJROk?t=1051) And then a different example, I have just come up with the example of a secret treasure map that's written in code and it's buried somewhere. So, that's going to be something that's probably hard for someone else to find. And it's going to be hard for them to understand, even if they do find it, it's going to be hard to understand what they're looking at because they don't understand the code. So, this, again, is a different way of thinking about our relationship to information and to the information that other people might have about us. That these people are suggesting that it's one thing to try to fight people from getting information about us. Maybe what we should be focusing on is, to what extent can we control other people's ability to understand the information that they have about us? So, the easier it is for someone to understand that information, the less obscure it is. Whenever we think about this debate between the privacy scholars and the obscurity scholars, or about how we should think about privacy in the first place, we might have a prior question, which is, why does any of this matter? So, I have here on the slide a few different answers that philosophers and legal theorists, as well as psychologists, for that matter, have given to this question of what some of the things are that we value and how those values might be related to this idea of privacy.

[18:43](https://youtu.be/u7XkcRPJROk?t=1123) One of the main ones that people come up with is a fear of exploitation, right? That perhaps someone knows something about me, and their knowing that makes me vulnerable to them in some way. Maybe they're an employer, maybe they're a government, maybe they're someone who just now, by virtue of having that information, has a certain kind of power over me. There's a similar kind of concern about unjust treatment. So, again, it could be that on the basis of some information that someone has, or thinks that they have, I may be subjected, or a group to which I'm a member, may be subjected to unjust treatment. There are a couple of other ones that are a little more conceptual and more, I think, interesting in some ways. I'll talk about them all together. The idea that some people express is that whatever we think of when we think about privacy, if it's about access to information about us, or what people do with the information, or how much information they're allowed to collect in the first place, what's common to all of those things is that if I don't have the ability to have some discretion, or some say over the kind of information about myself that's collected, and what's done with it, that's going to make it difficult for me to develop a sense of self, because I'm constantly going to be performing, as it were.

[20:05](https://youtu.be/u7XkcRPJROk?t=1205) I'm only going to be worried about what other people might do with the things they know about me. And so, if I don't have some part of my life that in some way can be walled off from other people, or some part of my life over which I get to say who gets this information and what they can do with it, then it's hard for me to be a self at all. This might be especially germane in the context of close relationships. So, imagine whether you think it would be possible for there to be friends in a case, in a world where everyone knows the same information about you that every other person knows about you. Part of what it is to be a friend, for example, is not that I'm telling someone information that they couldn't find on their own, but rather that I'm the one telling them that information. So, disclosing the information, in other words, is a sign of trust in this person. It's a way that I'm signaling to them a certain kind of closeness and intimacy. And so, my capacity to do that requires that there be some information that this person doesn't seek out on their own, or even maybe can't get on their own.

[21:06](https://youtu.be/u7XkcRPJROk?t=1266) So, one concern then is if we lose the ability to control the kinds of information that's available to other people about us, that makes it difficult for us to form intimate relationships or close relationships. Okay, so I want to now put these two topics together. Ways in which facial recognition technologies may raise challenges to privacy and obscurity. So, first, I want to talk about the benefits that are often claimed. And so, I don't mean that these aren't actual benefits, just that these are claimed benefits for facial recognition technologies. And in particular here, again, I'm talking mostly about facial identification technologies. So, the ones illustrated here on the slide, we may not have to remember passwords. If there's a system that could just check your face and match it, then okay, that's one less thing that you have to remember. Similarly, Amazon apparently was trying out an experiment with stores where you didn't have to have a wallet, you could just have your face scanned. It turned out that there was a little less going on than what they claimed was going on there, so we can talk about that later.

[22:11](https://youtu.be/u7XkcRPJROk?t=1331) Or, a different case, we can avoid those daily embarrassing situations where the name of the person who's approaching you is on the tip of your tongue and you don't really remember it until after they've passed, so we could perhaps avoid a lot of those awkward situations. I'm also aware of some cases where there is facial recognition technology being developed for people who have visual impairments or who are blind that may allow those people, when they're out and about, that the system could just tell them who it is that they're approaching or who's approaching them. That previous slide was benefits to the individuals. This slide is about the benefits for sort of collectives in terms of public safety, public security. So, one concern is there may be people who have dementia, for example, who might wander off. Their families can be very worried and upset. Wouldn't it be great if there were cameras around and we could have facial recognition technologies be able to identify where somebody was and who that person was and maybe who they belonged to?

[23:14](https://youtu.be/u7XkcRPJROk?t=1394) Likewise, there's a similar kind of claimed benefit that this would be a great way of helping us decide or determine who belongs in a particular place and who doesn't belong there. So, you can imagine schools using this or, as is the case currently, Madison Square Garden has facial recognition technologies that helps them decide if you're an attorney who's part of a law firm that's suing them, they won't let you in to see whatever games or other events are going on in there. And then the main one that people often mention is not necessarily that it would prevent crimes, but it would at least make it harder for people to be able to get away with those crimes. Because again, we'd be able to identify them by rewinding the video, looking at the video, or perhaps having live, real time facial recognition to identify those people. This brings in train within a couple of ethical concerns. So, on this slide, I have a couple of headlines from the past year and a half. So, ways in which governments or private agencies might use facial recognition to engage in real time surveillance, but also the ways in which any person could could use this to look up an image of you. Right? So, maybe they're passing you on the street.

[24:33](https://youtu.be/u7XkcRPJROk?t=1473) They think, oh, that person's cute. So, they perhaps surreptitiously take your photo and then put it into PymEyes and look up and find out what your name is. And once they have your name, they can find out a lots of other information about you. Our related concern has to do again with how all of these faces got into the training sets in the 1st place. It seems impossible that we gave informed consent to that. Think about the social media sites that you may belong to, or where you may have accounts. The terms of service never said anything about the possibility that those sites would be scraped and that the images from those sites would be used to train software that could be used to recognize faces. So, there's a question of whether just because something is on the Internet. And therefore publicly available is that that's not the same thing as saying that it's in the public domain. Not the same thing as as saying that this is something that you can send it to. A 3rd concern has to do with questions of due process and social biases.

[25:35](https://youtu.be/u7XkcRPJROk?t=1535) So, 1 of the things that has been true, although is less and less true every day, I should acknowledge is that certain facial recognition systems typically have been better at recognizing certain kinds of faces than others, particularly the lighter your skin is, the more accurate these systems tend to be at recognizing accurately who you are, or whose face you have. Now, as I said, those disparities are going away as these technologies become better as they become better trained and as the data sets become larger. Um, but it is still the case that there's not necessarily equally accurate rates between different demographic groups. In the United States so far, all of the people who have been arrested after false identification by facial recognition systems have been black. Now, this is not just a failure of the systems. This is not this is partly a failure of the, the police agencies that we're using those systems in ways that they weren't supposed to. And what I mean by that is that they weren't following their own policies, their policy said that a facial recognition match or probability should not be the sole determinant of whether you consider someone a suspect, or whether you arrest someone.

[26:48](https://youtu.be/u7XkcRPJROk?t=1608) But in all of the cases, so far, what happened was, in fact, that the police departments did take that to be sufficient reason to to go and arrest these people. This is not just happening with public agencies like the police. It's also happening with private companies. So the, the Federal Trade Commission recently settled a suit against right. Chain of drug stores, banning them for the next 5 years from using facial recognition technologies. And here again, the accusation, or the allegation on the part, and I guess, right. I agreed with this as part of the settlement is that. These were used in ways that were not fair, right? That that were biased against particular employees or particular customers. So, I realize I'm going a little bit longer, so I'm going to bring things to a close. So these are just again.

[27:49](https://youtu.be/u7XkcRPJROk?t=1669) General very general guidelines for how, as a public official, or maybe someone who knows public officials, you might think about. The role that you want facial recognition technologies to play in your office and how you might want to use them. So, I think 1st, obviously, is to reflect very carefully on what the actual mission is of your office and also think carefully about the core values of that office, but also your own core values. Know what the relevant regulatory landscape is that could include general norms, but also laws and actual regulations. And then the final thing to keep in mind, and I'll talk more about this briefly is to avoid techno solutionism. That is to say, avoid thinking that, well, the solution to the problems that are caused by a certain technology is just some other technology. So, 1 thing to think about, and I hope maybe this can be a question we talk about during the question and answer period. What's lost and this again is something that I think you should discuss with others in your offices.

[28:52](https://youtu.be/u7XkcRPJROk?t=1732) Um, if you lose the ability to just walk around as a, as an anonymous person, right? A person who can't instantly be identified by their face and then have that connected to other information about you. If we are all going to lose that ability, then is that something of value that we're losing or not? How do we articulate the value of just being able to be an anonymous person? Are there policies in your own department or your own office that might, in some way, be eroding this capacity of people to be anonymous people in public. Again, keeping in mind who this audience is many of you work for state agencies. So, in some sense, you are agents of the state. And so what I'm also asking you to reflect on is before you start deploying systems like these. Um, what opportunity do the people that you serve in the wider public have to understand how these systems work understand why you're considering using them and to give meaningful feedback that will actually be taken on board.

[29:56](https://youtu.be/u7XkcRPJROk?t=1796) There are lots of different laws and regulations that are. So, in addition to other forms of biometric protection or biometric data protection, facial recognition technologies are falling under the scope of new privacy laws in several different states across the nation in the United States, as you can see on the chart, or, sorry, the map on the left side of the screen. There's, in a European context, the General Data Protection Regulation. You can see as well, so I'm in the city of Boston, so you can see there's a kind of push and pull where, on the one hand, the Massachusetts legislature is working on legislation, right, to impose strict limits on the ways that police forces can use facial recognition technologies. The actual city of Boston itself has banned the use of facial recognition technologies by public officials and public agencies. But in the meantime, the airport, which is not run by the city or the state, is using facial recognition technology for passenger control and security.

[31:09](https://youtu.be/u7XkcRPJROk?t=1869) Again, the last point that I want to make before we turn to question and answer is that this is not a situation, or this is not a technical system, where we're going to be able to tech our way out of it. So this is an image of a line of clothing, which is more of an artistic statement than a line of clothing, called Hyperface. These are clothes that are designed so that when you wear them, they're meant to scramble the ability of cameras to be able to recognize your face accurately. Now, I think the way that something like this works is if all of us wear it all the time. Otherwise, if you're that one person wearing this funky clothing, in a way, that's going to make it easier to identify who you are. So, again, I don't think that having us wear different kinds of clothing that's going to scramble the signals that are going into cameras is the most effective way, perhaps, to think about this as a community or as a society, about the roles that we want these technologies to play more generally in our lives.

[32:12](https://youtu.be/u7XkcRPJROk?t=1932) It will have to be not just a combination of clever technologies, but also, again, of really robust policies, and even more fundamentally, a clear understanding of what it is that we want to protect in the first place.

**Unknown speaker**

[32:27](https://youtu.be/u7XkcRPJROk?t=1947) Thank you so much, Ren. Let me just take us off recording, as promised.

* * *

Original page: https://innovate-us.org/guidelines-for-promoting-privacy-in-facial-recognition-technology- · Licence: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org
