Cameras in the Classroom: Facial Recognition Technology in Schools is the first report from University of Michigan STPP Program's Technology Assessment Project (TAP). September 2020.
>>SHOBITHA: Good morning, everyone.
Thanks for coming to our webinar today.
My name is Shobitha Parthasarathy and I am a professor in the Ford school of Public
policy and director of the science technology program here at the University of
Michigan. I am also the director of the technology assessment project. This is the
reason we are here today. Not discuss the publication of our recent report, cameras,
the classroom, facial recognition and technology in schools. Before we begin I want to
address the ongoing strike but University of Michigan graduated employees, that began
about 10 days ago. Use graduate students believe the pandemics reopening plans are
insufficient and dangerous. They want the University to cut off their relationship with the
police department. They want the funds of the University's division of Public Safety to
be reallocated. We are sympathetic to those concerns. We have fought long and
hard about whether to have the event today. We decided to go forward because the
issues we discussed in our report about law enforcement biased against people of
color, and racism and surveillance technologies. They are aligned with GEO's concerns.
Our overall aim today is to give you an assessment to the technology assessment
project. We call it topic. Methods of analyzing facial recognition and other echnologies,
to summarize our findings, and to talk about our conclusions and to knowledge E policy.
We will have time for questions in the end. Please leave questions and each time we
will get to them. Top as part of the University of message Michigan science and
technology program. As TPP is a unique policy center concerned with cutting edge
questions that arise at the intersection of technology, science, and policy. We have a
vibrant graduate certificate program, public and policy engagement activities, and
electro series. Our graduate certificate program is the jewel of our work. We teach
students about how values, society, and politics shape technology and science.
Nylund students also learn about the signs I technology making environment. They also
learn how to engage in it. Our alumni now hold jobs across government of the
private sector, and non- governmental organization. Our program is anger because of
profoundly disciplinary approach we take, to teaching students about the relationships
between technology, science, technology and policy. Also because of the into
disciplinary about the students who in gauge. It engages students from the school of
information to the business school. 26 percent of our students come from engineering,
for example. In recent years, given the growing interest in the social ethic on equity
interest of technologies, our program has been growing. At present gift 73 students. As
TPP launched the project in the fall of 2019. This technologically driven world is
increasing at unease. As citizens we are aware more and more of how technology is
shaping our lives. We are starting to see how it has disproportionate impacts. It tends
to not just affect but reinforce any qualities. At the same time policymakers are confuse
how to manage technologies. They worry they cannot properly anticipate the
consequences to regulate it properly. It seems like technological development move so
quickly. How can there be adequate time for policy discussion and legislative activity?
On what basis should they be making decisions? Because of these things we wanted
to do something. The idea of technology assessment as a general idea is not new.
They have used a variety of techniques to try do it.despite the social, environmental
problems. They use this analysis to inform their governments. We are try to do here is
something different. We are developing an analogical's case study method. Using
historical examples to inform our analysis of emerging technologies. I am a social
scientist myself, I use these kind of historical case study methods as my own
research, it is familiar to me. The basic idea is that the implications of emerging
technologies are much more predictable than we tend to think. We can learn from the
history of technology to anticipate the future. If we look at previous technologies we can
understand social patterns, and how technologies tend to be built, implemented and
governed. And the problems that arise. If we understand those things, we can do a
better job of anticipating those consequences. Before we get into the report, I want to
introduce our brilliant research team. They are all here today and you will hear from
each of them. Karen Forbes is a recent graduate of the Ford school public policy and
an associate in Chicago, background in tech. I want to emphasize, to analyze emerging
technologies. We also want to in Clement key disciplinary and writing skills. It is
important to note that Claire and Hannah did much of the case
development writing. As the top report is released at they have had the opportunity to
disseminate reports like these, and to engage with the media. other kinds of
commentary for public audiences. Of course, give presentations. Finally Molly
Kleinman is as TPP's program manager, she is also an expert in educational
technologies, therefore is a valuable part of the research team. We have a number of
people to hear from today. So, why facial recognition? And why schools? In so many
ways the idea of using a digitized image of your face to identify or verify your identity,
seems like the step of science fiction. It is increasingly used today from China to
Europe to the United States. It is used in a variety of settings, including most
notably and famously for surveillance and security purposes and law enforcement.
Also, for identity verification and our smart phones. Here in Detroit we have
project greenlight, businesses send footage to police. Digital images are checked
against law enforcement databases. The new use on the block, it's facial recognition
technology in schools. We have seen increasingly used across the United States.
It seems like a great solution. Perhaps most notably in 2018, a New York near
Niagara Falls, announced it would install a facial recognition system and its
schools. Cameras installed on school grounds would capture the faces of intruders.
Analyze if they match any person of interest in the area, have that much confirmed by a
human security person and that if there was a match, it would be sent to district
administrators who would decide what to do. The system was approved by the New
York State Department of Education last year, it finally became operational earlier
this year. Since and has been a subject of a lawsuit by the New York civil liberties
Union, and July the New York State legislator past a two-year moratorium on the use of
debates. At present, there are no national level laws. Exclusively based on recognition
around the world. Many countries are expanding their use of technology without any
regulation in place. Mexico City for example, invested $10 million and putting
thousands of facial recognition security cameras, to monitor the public across
the city. The state city system has been installed in 230 cities around the world,
from Africa to Europe. We did a comprehensive policy analysis of the national
international landscape. We wanted to look for policies that might be interpreted as
being related to facial recognition. We classified both proposed and past policies
into five categories. The first two are consent and notification policies which focus on
data collection, and data security policies that focus on what happens to the data, once
it has actually been collected. A number of policies in these two policies have been
passed around the world. Most notably the general data protection regulation in
Europe. Similar laws have been passed in India, Kenya, and also in the United
States, Illinois and California. European courts have found the GDP our club is facial
recognition. The third calorie is policies that Taylor use. We see this in the case of
project greenlight for example. The city of Detroit has limited the use of facial
recognition to investigate violent crimes, and also banned the real-time video
footage that was part of the original approach. Fourth, we see oversight reporting and
standard setting policies. These mandate different ways of controlling the
operations of facial recognition systems, including their accuracy. Most of these have
only been proposed. Finally, we see facial recognition use, bands, these have been
proposed at the national level in the United States. Where we see a lot more
policy activity in the United States is at the local and state level. Some states have
banned law enforcement use of facial recognition and body cameras. A number of
cities from Somerville Massachusetts, to most recently Portland, have enacted bands of
varying scope and strength. We have some policy activity but a lot of it is in progress.
None are explicitly at the national level. This raises some questions. We think about
developing policy for facial recognition, what should be thinking about? How do we
know what we should be thinking about? This brings us to analogical case study
analysis. Analogical case comparison, we mean systematically analyzing the
development, implementation, and regulation of previous technologies, in order to
anticipate how a new one might emerge, and the challenges they will pose. When it
comes to facial recognition, we look at technologies that seem to similar in both their
form, and their function. We looked at how close circuit television and resources that
schools have used. What kind of social and equity problems they have had.
Once we developed ideas about the kinds of inflammations and complications that
might have, we looked at other sorts of technologies that had those implications and
tried to expand our understanding in that direction, as well. For example. Facial
recognition might create new kinds of markets and data. Looks at markets that have
been created biological data like human tissue, and genetic information. We try to
understand the implications of that. We did all of this interactively and we ended
up with 11 case studies, until we could clearly see five main conclusions from
analyzing these case studies. Claire do you want to take it from there?
>> Hello everyone I am Karen Forbes and I'm going to talk about the first three of our
five findings, why facial recognition has no place in schools. First of all, facial
recognition is racist and will bring into schools. Not only will it discriminate against
students of color, it will do it in a legitimate affair. Technology is so often assumed to be
objective and highly accurate. You may be thinking, isn't technology objective? Or
more objective than humans? The answer is absolutely not. Technology does not
exist in a vacuum. Facial recognition is developed by humans. Based on data sets
compiled by humans and used and regulated by humans. Human bias enters the
process every step of the way. Discriminatory racial biases are no exception.
We came to this conclusion that facial recognition is racist by studying the
analogical rate case. The policy that police officers can stop citizens on the street by a
standard of reasonable –. They discriminate against people of color because bias
enters their views. Many would argue that the staff is neutral because, if you are not
acting suspicious you should have nothing to worry about. However, that comes to
become untrue. It has been consistently proven to be raised disproportionately against
black and brown citizens. Take New York City for example. Throughout the use of this
policy people of color were stopped at higher rates than white residents.
Compared to the rates of crime they actually committed. Because it was susceptible
to racial biases of officers, it criminalized and discriminated against people of color at
high rates. Even though it was to be a objective policy. Facial recognition is
similarly susceptible to racial biases. It unfairly targets children of color. Finally, and
condition it has been proven time and time again to be inaccurate. Facial recognition
algorithms show higher alert rates for people of color. Black brown and indigenous
individuals, especially women, are misidentified. This will create barriers for students of
color. Because they are more likely to be misidentified by facial recognition, they are
more likely to be marked absent from class, locked out of buildings, or floods as an
intruder in their own school. Altogether it is inaccurate and ask and it will create barriers
against students of colors in school, we believe it should be banned. This brings us to
our second finding, facial recognition will bring state surveillance into the classroom.
We expect facial recognition will use this technology liberally. Conditioning students to
think it is normal to be constantly watched until have no right to privacy at school.
An environment that is so supposed to be safe and constructive. This is proven to have
negative effects on children. The analogical case study of the use of closed-circuit
television in schools, let us to this conclusion. CCTV is using most secondary schools
in the United Kingdom, facial recognition is virtually the same technology with much
more powerful capabilities, we felt this case was a perfect example of how putting facial
recognition in schools, will play out. This case reveals less that one administrators
are entrusted with powerful surveillance systems, it is hard to control how they use
them. We call this mission creep is the use of surveillance technology outside of the
original agreed-upon intent. Interviews with students at the schools though
these symptoms were limited for security purposes they were used for behavioral
monitoring and control. It was supposed to be used to find school intruders but was
rather use to punish students who violate a dress code, tardy, or anything else.
Students reported this use of CCTV, made them feel powerless, criminalized by
their school, and mistrusted. They reported they would change how the outdoor
dress at school, to avoid punishment. This heightened anxiety reduced feeling of safety
in school, it is likely to affect a child's educational quality. Because CCTV is just like
facial recognition, except for the fact that facial recognition not also surveilled, it can
identify students. Just like CCTV, administrators for facial recognition will be unable
to resist the temptation to use it outside of its agreed-upon purposes. We are positive
constant surveillance will make students feel anxious, stressed, and afraid.
This is where they should feel as safe as ever. This brings me to our third finding.
Facial recognition will create nonconformity for creating barriers for students who do not
fit into standards of acceptable appearance and behavior.
>> This is only the tip of the iceberg. It also has heightened when used on gender
nonconforming students disabilities, and children. Yes children. This is problematic
when this gets implemented into schools. We are confident that these will create
barriers for students who may already be a part of marginalized groups. Nationwide
biometric system is the largest biometric system in the world. They have fingerprints
and high facial scans over 1 billion citizens. It is required to access many public and
private services, including welfare, pensions, mobile phones, financial transactions, and
school and work environment. Like facial recognition, it is designed in such a way it
excludes certain citizens. Specifically, citizens who cannot submit biometric data.
Such as manual laborers or patients you may have damaged fingerprints or eyes.
This means these individuals who are already disadvantage, are now unable to access
food rations, welfare, or pensions. Therefore, because these groups do not face her
and acceptability standard they face even more disadvantages in society, for no fault of
their own. We expect that facial recognition will replicate this in schools. When we
know from the CCTV example I discussed earlier that it is likely facial recognition will be
used in schools for behavior, speech, and dress code. We expect this will affect
students personal expression and get them in trouble, for not conforming to
administrators preferred appearance or behavioral standards. Altogether, because
facial recognition is likely to malfunction on students who are not white, or able-bodied,
we can expect facial recognition to affect marginalized students to be incorrectly
marked absent for class, prevented from checking out library books or paying for lunch.
We also expect children's will be directly or indirectly discourage for free public
expression. Overall facial recognition in school is for some students, and to exclude
or punish others. These will be characteristics outside of their control or their
personal expression. With I will turn it over to Hannah.
>> Thank you I am Hannah. There is so much detail, and great examples of everything
we are talking about today in the report. I encourage all of you to read it. I want to focus
on a few highlights for our cases, for the last two major themes we touched on. Let's
talk about how companies profit from data. A company cannot own your face, they can
own a representation of it. Like the one that is related by facial recognition algorithm.
A data point may not have may not be particularly valuable, companies can create value
by aggregating data, with a lot of data, or with different information that gives it context.
Individuals typically do not own any of the rights to their biometric data, despite this.
They often do not have a meeting will obey to give consent to collect and keep that
data. These new databases are vulnerable to subpoenas. We talk about the report
surveillance company called vigilant solutions. They set up a selling license plate
scanners to private companies. They connected all of the license plate data information
with all their customers. Knowing they all knew were all the scanners were placed and
how the geolocation information. They were able to recognize that they can use that to
build a new product, they gave real-time information about the location of cars
around the country. They packaged that product, they sold access to law enforcement.
The law enforcement product was so valuable they started to give away their original
license plate systems for free, to get even more data. They created a new data
market around license plate data that did not exist before. They were strongly
incentivized to expand it. We talked more about the implications of behavior like this in
the report, generally we do expect facial recognition companies operating in schools,
will also have a strong incentive to find secondary uses like this, for student data.
To get that valuable data they will try to expand their opportunities for collection, as
much as possible by pushing to get into as many schools as they can. Paradoxically,
US ports have traditionally held that individuals do not own. The more versus the
Regents of University of California, 1990. In that case, Doctor Gold of UCLA used
samples from his patients to develop a valuable research. He sold out and profited
off it without telling more about the research or potential profit. After he sold that he
would notify more of this use. Doctor Gold should've told him about the research
upfront, more tissue was considered discarded even when inside him, he could
not use it for anything on his own. He had no property right to it. Today a company or
researcher cannot use biometric data without consent. They do not owe anyone a
property state, or anything they develop using their data. Meaningful consent is
often limited, when it comes to complex technology systems, and especially in situations
like schools, where you cannot opt out of it. In the United States in particular, students
do not get a chance to engage with any of those questions because the family
educational rights and privacy act, allows schools to consent on behalf of their students.
Consent is not a topic for the student who may be surveilled. This push to expand
surveillance by companies, it will expand all of the reach of all the problems we already
mentioned. It also introduced some new issues of its own. Without strong data
productions this information is vulnerable to being stolen. It is impossible to replace
things like fingerprints and faces, once it is stolen it is out there. Second, data collected
for one purpose will be used for other ways, at the same level of scrutiny, that is the
mission crew we talked about. Thinking back to vigilant technologies, many of those
private customers who use those initial license plate scanners, may not have signed up
to share the information. If they knew they were going to be generating data, that would
be sold and packaged to the police. Even if a company does not actually package I
data, for law enforcement use, one that has been collected they can subpoena that
information, and it is a lot cheaper and less politically difficult to subpoena them
build a similar database on their own. That is what happened last year with
ancestry.com, when police subpoenaed information from their DNA database to solve a
crime, which would have been very difficult for them to put they database to gutter on
their own. You can easily imagine that police could subpoena facial recognition from
school to identify a student or track a child's whereabouts, if we are looking at the facial
recognition case. Our last theme is institutionalized in accuracy. In the report we
systematically unpack what accuracy really means, and what it doesn't mean.
Technology like facial recognition, we really talk about all of the ways that this idea is
complicated for a sociotechnical system. Accuracy could be in itself. I want to directly
tie what Claire said earlier about technology is being unusual from the human society
we are in. Thoughtfully and by sociotechnical system, directly to the questions about
accuracy, I want to show how these questions often get overlooked, leading to poorly
functioning technologies, becoming entrenched in daily operations. Components of
facial recognition also answer critiques about the technologies, by pointing out these
algorithms. They will get better and be improved by being used. Facial recognition
accuracy problem, deacons during the learning stage. That is early in development.
-- algorithm first to build a training data set, the system learns how to identify faces,
then you apply the algorithm to a testing side faces where the researcher knows the
identity, but the program does not. You see how often algorithm gets used. That is the
accuracy measurement that you are getting for most companies. The demographic mix
in the training site is going to determine how strong the algorithm is in various
demographics. Problems can vary. They are different from the real population
if you do not train with any women, you can still get your accuracy number two show
high levels of accuracy in your test. When you apply that system in the real world
where black women exist, you're going to have a problem. This is exactly what we
are seeing. The most common testing site is 77.5 percent male and 83.5 percent
whites. The main US agency, has not disclosed the Democratic makeup that
uses to test software it is difficult to test their accuracy. We do know that they built
in other database that was intended specifically to test how well facial recognition
performs, across racial groups and country of origin as a proxy for race. It did not
include any country is predominantly black, a high accuracy can also be hidden by
performance by one group. If that is outweighed by very high performance in
another group. You can begin to get the sense that it is very difficult to tell from one
or two numbers that I company might provide, or school district might have access to,
how accurate is this going to be across the population of your school? Another answer
to critiques like this, there should always be a human making final determinations.
However, when we looked into other forensic technologies, that use similar human
backstops like fingerprinting, predictive voicing or CCTV, it turns out across the board
when there is uncertainty in the process, forensic examiners tend to focus on evidence
that confirms their expectations. From CCTV studies we can get a sense of how well
humans might actually perform as safeguards for facial recognition, research shows that
trained observers identify individuals from footage, made correct identifications less
than 70 percent of the time. That number drops even lower when they are asked
to make cross racial identifications, to identify someone who is a different race of her
own. As a reminder., Even if a person incorrectly got identify they have already been
critiqued and sanctions that might have gone unnoticed. It also opens up a new
opportunity for human biases and punishment. In a facial recognition palette and South
Wales, the police revealed that in the process of making only 450 arrest the
algorithms falsely identified over 2000 people. We get a sense of public this problem is
or could be. This is another way technology is fundamentally a part of human society.
Humans who are most susceptible to human identification are also those who are most
likely to face outside punishment. Another issue is the question of who who decides
what is accurate. Courts end up being the main arguer. In the United States trial
judges and the United States court system, the most standard to determine whether a
non-expert witness testimony, about forensic technology is scientifically valid for
the case. That is done on a case-by-case basis. They consider things like
potential air rates, the technologies of reputation and scientific communities.
They do not have a minimum criteria for determining any one of these categories.
As a result the accuracy of fingerprinting is ultimately determined by the legal system of
the equality of the lawyer and the experts involved in the case. Some states currently
accept fingerprinting and extra scientific witnesses while others do not. Suggesting this
is not fully reliable. It creates two separate standards of evidence. One for those with
the means to mount a strong legal defense and another for those without such means.
Law enforcement may have incentive to use technologies that they are consistently able
to get it in court with them. There is a lot more to talk about, what I'm getting eyes
accuracy is much more complicated. These translate human biases into the software,
and into the system of the software. Despite this, people tend to perceive, technology
as objective. It is inherently free from bias. Please have a long history of leveraging
this idea of objectivity in court, to argue that some arrest could not be biased,
because they were based on our current the myth prediction. This usually holds up in
court. I am going to pass it over to Molly to talk a little bit more about we anticipate
>> Thank you. I am Molly Kleinman. Our recommendation is straightforward. Patient
and use of recognition technology in schools. In all of our research we are unable to
identify single use case where the potential benefits of facial recognition, would
outweigh the potential harms. This is true for the type of in person facial recognition
systems we are mostly considering in our research. It is also true for the kinds of
research that is expanding now in the online education schools, that are being used
during the pandemic. We have kids sitting in front of cameras all day. These
companies must not be allowed to collect biometric data for children, that have no
choice but to use their products. Furthermore, in this coronavirus situation we
are seeing other kinds of biometric surveillance expanding, such as thermal
tracking. Many of these systems have the same risk and dangers of facial recognition.
It would be best if countries banned them. In the absence of this, we are
recommending nationwide moratorium that would be long enough to give countries to
get advisory committees to investigate and recommend a regulatory system. When we
say an advisory committee. Talking about something that would be really
interdisciplinary. It includes experts in facial recognition technology and also in
privacy and security. Civil liberty laws, race, gender, education, and child psychology.
A moratorium should only and when the work of the committee is complete, and
the regulatory framework has be fully implemented. Separate from a moratorium
armband country should have security laws that address facial recognition and other
kinds of biometric data, if not already in place. The GDP art does not explicitly address
facial recognition, courts in several European countries have ruled that facial recognition
data is included under the GT PR personal data and is not permitted. Our report also
includes district level policymakers to help them provide effective oversight in the
moratorium. This is a situation we find ourselves in right now. As we discussed earlier
there is very little regulation happening at the national level or. We hope you look at the
report and ask questions. A list of Angelo's wills our goal of these listless to have them
ask questions even if it is only a single school building they are worried about. I'm going
to hand things over for questions.
>> Thanks Molly. Thanks for participating in the presentation, and talking about your
results. That is a pretty good place to start in terms of getting a general sense of
what the report talked about. As Hannah mentioned, we talk in much more detail
in the report about the recommendations. For each of these conclusions we looked at a
number of different cases for each. You will find those resources in the full report and
we have a shorter report. We have maps that I created a separate supplements.
As you folks are asking questions and the chat, I will say if you are interested in STPP
and its work, you can visit our website. You can see the URL here. You can follow us
on Twitter, and if you want to keep up to date on what we are doing and get our
newsletter, you can email us at STPP@UMICH.edu.
>> To think facial recognition should be banned across the world are just in schools?
>> Based on what we learned in our research, students are particularly vulnerable.
I cannot talk on if the drawbacks are going to outweigh the benefits, it is going to be just
as inaccurate, a lot of these things like feeling surveilled, and avoidance, a lot of the
cases we looked up we drew on cases that were not just in schools. We learned a lot
from people out in society, a lot of these learnings map directly onto other scenarios as
>> I would say I agree with that. I would like to add, in some ways someone who
thinks about case comparison, using schools is just one case, it is a hard case, I agree
there is a lot of vulnerability among children, therefore I think use of technology
has to be at a very high threshold. I also think, we have offered a lot of
recommendations and they are pretty detailed. I think use cases that are more
complex, used in law enforcement, or even identity verification, we provided resources
that can be useful for those cases as well. This idea that humans are partly technology,
we have to address that in some way, the idea of data markets, these are things
we all have to think about in a really serious way. Regardless of the use of facial
recognition, my concern is that we are not thinking about those sorts of things enough.
Regardless of the use, you see that even most recently in the context of project
greenlight in Detroit, which is not the use in schools, a black man was misidentified
because his image was captured and it was linked to some older drivers license photo.
That kind of problem that we are identifying is something we see across facial
>> The next question I have. Given private and public sector advocates of this
technology are likely to use the charge of inaccuracy and bias as leverage to highlight
how the tech industry is improving and training the human backstops is improving and
diversifying, the arguments against using the technology are being addressed, and will
soon be read the lesson relevant, so why banned this use outright?
>> I am going to take the director's privilege and say I think Hannah did a great job of
answering that question. The fundamental point we are trying to make is it
is impossible to reduce the inaccuracy in this technology. We cannot tech our way out
of this. There is no tech without humans and society, that introduces systematic,
structural bias, and individual bias. I am just using different words than what she said.
I do not think it is an argument. I hope our report provides some details to explain
why the argument is inadequate.
>> This is a question that is related to that one. Where do you think the attitude of
technology's objectivity comes from? Is there any sort of shift we are seeing in general
>> That is a great question. I could not tell you where that attitude comes from.
I think there is a belief in science tech and public policy field, we talk about the black
box of science and innovation that people assume what is done by scientists in a
lab is perfect. You assume if it is informed by science and research then it is not
likely to be inaccurate. I think it comes partly from black box policy. Chris is also STPP
argues for, we need to go ahead of this and regulate these things. Not just blindly trust
they are accurate. I do not know exactly where it comes from, but is it changing
generalization? We would all say yes. That is because we talk to STPP scholars all
day. I would like to think yes. Generations are getting more and more technology
advanced, for further use people use it every single day, and there is a better
understanding now that it is not perfect.
>> I do want to say that some groups have seen this as nonobjective.
It is not always news to every group. In general, a lot of people, scientists, and
tech, spent a lot of time in Silicon Valley. I still took my research now with people.
I very frequently find myself having to bring up the idea that tech is not
objective. I do not see there being a huge generational shift at this point, I do want to
point out that some groups that have often been harmed by things that are
called objective, have always known it is not objective. The idea that we are just
now learning this, is the idea that we are just not able to elevate this academically. This
is not just no, this is in science and technology studies as well. We are just now seeing
some of this be elevated in a more mainstream way. I think there's a lot of
power in the idea that your discipline is objective, the idea that we can build a new facial
recognition system or a new algorithm, the idea that it is objective gives us a lot of
power, to not only avoid some questions ut gives a lot of economic power to get
funding for projects like that. There is a lot of power that goes along with objectivity.
It does not mean it is intentional that people are trying to build up the objectivity, I think it
is a product of over time there is a lot of funding condensed around that.
>> I am calling this next question for myself. What do you make of tech that uses facial
recognition not in the name of security or law enforcement, more for specific teaching
and learning context that are often driven by faculty demand such as online exam
proctoring? I do not like those either. I think anytime you have a situation where you
are treating your students as probably trying to cheat, or probably criminals, you
are disrupting the relationship with your students, that is not a sound way to approach
education. I would argue that if you need this kind of policing technology, in order to
assess your students, you need to us assess your students differently. I realize that is
easier said than done, there are incidents where it is difficult to come up with
different kinds of assessments that would work in an entirely remote situation.
We are dealing with a humongous expansion of remote education that people were not
prepared for. These tools can seem really convenient, like it is going to make life
easier and it's time or nothing in life is easy. I want to think about with who is making
the life easier, I want to center on the students experience when those technologies
are being used.
>> I will briefly say that facial recognition technology is ready being used in schools
across the United States pre-pandemic, and as you said yourself is only expanding now
at colleges and universities. The things that seem like seductive solutions, have
other types of problems.
>> This next question says I am from a place overseas where the government is
beginning to roll out smart sensors capable of being used for facial
recognitions, there has already been resistance locally, are there any simple to
understand resources that we could recommend to use to spread awareness of the
issues raised? Our report is a great place to start. [laughing] There are some questions
at the very and that are tailored for individuals acing these kinds of deployments.
It is focused on the questions they can ask unless specifically about advocacy.
I think it is a pretty short step from some of those questions.
>> I would at the end of the report, we have the questions that can articulate
some of the kinds of questions people might ask. I believe this is on the website but we
will add it, a one page summary of the five conclusions that might be a good one pager
to distribute in terms of resources. Somewhere in that report, all of the sections are
hyperlinks, we gathered a number of resources of add advocacy groups and others
during work in this area. I pushed everyone to make people think we are
thinking about these things internationally, there are some international resources
there that might be useful for those folks around the world.
>> A number of technocratic solutions are focused on diversifying data sets to better
identify black people, which unfortunately represents a form of predatory inclusion. Are
there strategies that advocates can use to highlight racist aspects of many of these
>> I did see a lot of thought. Last year there was a big story about Google paying
a company to go into Atlanta and give homeless people five dollars to use
some kind of phone game, they were actually capturing their faces with a very
complicated consent, that they were not explaining. If the response to the
criticism of you being racist, is to go out and do something like that, it is hard to say
what is worse. That is a very bad outcome. I am not sure what I suggest as a solution.
If we can get this idea out there, not those clients of adding people to your database is
not sufficient to fix this problem, it would take a societal change to fix this. These
technologies are a part of society and totally indestructible from not. Next you can very
easily see that companies are not ready to not make a racist technology. I do not have
a specific solution, maybe a better understanding in general, it cannot stand on its own.
Plus I'm often accused of being a pessimist might come to technologies, I'm going to try
to be an optimist. When you see the report we offer policy recommendations. On the
one hand our instinct is to say there is something deeply wrong systematically and
socially, that cannot be fixed easily. We really need to think about stopping this.
We also understand what the real world looks like, and we want to address it. When you
think about a question like this, my initial instinct is to say we can use the analogical
case study approach that we are trying to pioneer here. I thinking through this I
thought to myself about two kinds of examples, one is infrastructure for
human subject research, it is certainly not perfect, it has a lot of problems. When you
think about the inclusion, dealing with exclusion by some sort of inclusion, or the
accretion of messed up incentives we have human subjects research system
has evolved, and has certain kind of institutional bias that we can take something that
has worked in society, and maybe innovate beyond that. Maybe that is one place to
start, if it is not a perfect solution. One of the ways that people have found not to
be a great solution, is the fact that it relies on individual consent and not community
consent. In recent years there have been ways to try to get community consent when it
comes to biomedical research. That is a another place but we might say okay, we
can think about how we will get community consent when it comes to issues
like trying to diversify data sets, of course, the diversification of data sets will not solve
this problem. If we are thinking about that we can use these analogical cases.
I also want to emphasize one of the things I am Trying to do, is to look across text
sectors and look historically. That is key. Too often we are sidetracked by different
areas of technology, when we can learn across them, one of my hopes is not in
this effort we can start to look at examples that might look really different, but can
actually teach us some interesting things.
>> We have some questions We were not able to get to today. I will see if we can
follow up with you guys to answer those questions. Feel free to reach out to
us and get in touch. The contact information is all in the report.
>> Thank you all for coming. You know where to find us. Thanks everyone.