You are on page 1of 10

1

Maria Imperatrice
CST 300
11 February 2024

The Ethics of Facial Recognition Technology

Introduction/Background

Facial recognition technology has been utilized for a variety of different functions, from

unlocking our smartphones to helping law enforcement fight crime. Advancements in this

technology have taken place over several decades, from its initial development involving

computers mapping markers on human faces to the use of artificial intelligence to identify values

in large amounts of data. With these continuing developments, an ethical issue has arisen from

the use of this technology. The same data being used to train artificial intelligence in facial

recognition has created ongoing concerns around the existence of bias that presents itself in these

data sets. While the technology itself is not inherently bias, its development can and has been

embedded with bias. A notable example of this can be seen when mostly white male faces are

used to train this technology, leading to possible misidentification of women and minorities by

police and law enforcement agencies (Hassanin, 2023).

Development of facial recognition technology began in the 1960s. The earliest pioneers

of this technology were Woody Bledsoe, Helen Chan Wolf and Charles Bisson. This

rudimentary technology used computers to recognize the human face through manual marking

and measured the distance between these landmarks to determine identity (NEC New Zealand,

2020). Goldstein, Harmon and Lesk built on Bledsoe’s work in the 1970’s to incorporate

additional subjective markers, including lip thickness and hair color, in order to automate the

process in facial recognition. Time passed, and it wasn’t until the 80’s and 90’s that additional

progress was made. Mathematicians Sirovich and Kirby used applications of linear algebra to
2

solve some of the issues with facial recognition software at that time. Finally, in the 2000s, the

U.S. government began Facial Recognition Vendor Tests (FRVT) that were designed to allow

the government access to prototype technology and evaluations of facial recognition software

systems that were available for use (NEC New Zealand, 2020). It was during this time that 9/11

created a larger demand for this software, making its usage more widespread by federal, state,

and local government agencies.

Further advancements in this technology and its utilization by government agencies have

only expanded in recent times. In a 2021 Government Accountability Office report, 24 federal

agencies were surveyed and reported using this technology for digital access, cybersecurity,

criminal investigations, and physical security, including controlling access to buildings. The

agencies that reported its usage noted that it was also used to aid in their investigations. It was

used to compare images against mugshots and to identify victims of crimes through systems that

allowed them to compare against publicly available images from social media. Nearly half of

these agencies went on to say that they plan to expand their use of facial recognition technology

(U. S. Government Accountability Office, 2021).

Stakeholder Analysis

Two opposing stakeholders each hold values, position and claims regarding the use of

facial recognition technology and software. Those adversely affected by the use of this

technology are against its continued use by government agencies, while the same government

agencies that utilize this technology are in favor of its use in the commission of their work.

Stakeholder 1: Affected Parties

Values. Those affected most by the inaccuracies present in this technology also have the

most to lose when it fails. Due to the devastation it can cause those with the potential to be
3

adversely affected, the importance of addressing these concerns prior to deploying this

technology by government agencies is of the greatest importance. The bias in artificial

intelligence and machine learning software creates an ethical dilemma which can lead to

discrimination and false identifications (Hassanin, 2023). The utilization of this technology

without addressing these biases is irresponsible and will create issues in the future as artificial

intelligence becomes more accessible and widely used.

Position. The use of facial recognition software should not be utilized, especially by

government agencies, until biases have been addressed and there isn’t a discrepancy in accuracy

between race and gender. Laws must be enacted to regulate this technology. Continuing its use is

not only irresponsible, but potentially damaging to those adversely affected by the bias existing

in the data sets being deployed to train this technology. According to a paper written by

researchers at MIT, error rates for gender classification were higher for females than males and

were also higher for subjects that were darker skinned than those who were lighter skinned. This

paper went on to note that the data set used for this system was more than 77 percent male and

more than 83 percent white (Hardesty, 2018). This highlights a fundamental flaw with this

technology and the need for laws to regulate its usage.

Claims. The use of facial recognition software by government agencies should be

banned. This claim of policy mitigates the ethical issues surrounding the continued use of this

technology. Its discontinued use would stop any institutional racial, and gender disparities that

have been created as a result of its training.

Stakeholder 2: Government Agencies

Values. Using facial recognition software is beneficial to society as it permits law

enforcement and government entities to not only prevent crime, but also catch criminals.
4

Allowing the continued use of this technology will allow law enforcement agencies to keep

communities safe and aid in solving crimes that may not have otherwise been solved without the

use of this technology. It is the government’s responsibility to keep its citizens safe to the best of

its abilities and by using any means necessary. The importance of this goal can be seen through

the various government agencies that have been created in the commission of this goal, including

the NSA, Homeland Security, the Federal Bureau of Investigation, and countless other state and

local law enforcement agencies.

Position. While there may be mistakes and bias in facial recognition software, the

benefits of using this software outweigh the risks, especially when it is used to prevent crimes

and keep society safer. Utilizing this technology has expedited processes that traditionally have

taken countless hours; for example, in the past when the police would attempt to identify persons

of interest, they would manually look though hundreds of mugshots or search through databases

with minimal descriptions. Facial recognition software automates this process, allowing agencies

to dedicate additional time and resources to other cases. While this software can be prone to

errors and mistaken identity, so can eyewitness identifications. Mistaken eyewitness

identifications have been responsible for a large number of wrongful convictions in the United

States (Parker, 2020).

Claim. Despite the presence of bias, we must look at the greater good and the benefits to

society of utilizing facial recognition software and artificial intelligence. This aligns with a claim

of value, that it is better to keep society safe and catch potential criminals than to discontinue its

use due to any bias in data sets.


5

Argument Question

Should government agencies utilize facial recognition technology to help solve and

prevent crimes knowing the issues and bias that exist within the AI data sets used to train this

technology?

Stakeholder Argument

Stakeholder 1: Affected Parties

Government agencies should not use facial recognition technology given the bias

embedded in its software. Virtue ethics would dictate that there needs to be limitations and

safeguards in place to ensure that these concerns are addressed and rectified before using this

technology. This is because morality is guided by virtuous characteristics; being just, brave,

generous, wise, and courageous all play a part in this framework. Allowing discrimination of any

kind is inherently not virtuous. This ethical framework is said to have originated with Aristotle,

who noted that a person is seen as virtuous if they have ideal characteristics. These

characteristics are applied in a moral situation and focus on experience, ability to reason, and

sensitivity rather than rules or principles (Athanassoulis, n.d.). In that sense, it is virtuous to

discontinue the use of a system that perpetuates institutional discrimination.

The virtue ethics framework places an emphasis on virtue or morality as opposed to rules

or consequences. One must ask whether it is virtuous to allow for the continued usage of

technology that has shown large discrepancies in accuracies across race and gender. These

discrepancies can lead to misidentifications in best-case scenarios and serious long-term legal

consequences in the worst cases. Rather than focusing on the outcome of a given moral dilemma,

character and virtue should be used when considering what the morally correct course of action
6

is (Athanassoulis, n.d.). In this case, moving towards a society free of bias and discrimination is

the right thing to do.

This course of action is most correct because it emphasizes the role of virtue in this moral

dilemma. Discontinuing the use of this technology would effectively end yet another means of

institutional discrimination. Consequently, further action to address the issues in the software

would create a product that is not only more reliable, but free of embedded biases.

Those adversely affected by the continued use of this technology can see their losses in

the ongoing discrimination resulting from the inaccuracies in this software. Overall, bias in

artificial intelligence data sets has led to a series of problems, including but not limited to false

identification, racial and gender discrimination, and profiling (Harvard). Conversely, what is

gained by banning this technology is the protection of those in our society who have continued to

face institutional racial and gender discrimination.

Stakeholder 2: Government Agencies

It is in the government’s best interest to use this facial recognition software, as doing so

would allow for these agencies to continue fighting crime, catching criminals, and keeping our

society safer. This aligns with the ethical framework of consequentialism, which is derived from

utilitarianism. The consequentialist framework has been around for a significant amount of time,

and it’s the theory had roots in the utilitarian framework developed by Jeremy Bentham,

however, the term “consequentialism” was first coined by Elizabeth Anscombe in her essay,

Modern Moral Philosophy. According to this ethical framework, the consequences of an action

are what matter when making an ethical decision (Seven Pillars Institute, n.d.). In this situation,

the consequences of continued use of this technology can be seen in the reduction of crime, the

increased frequency of solving cases, and the creation of a safer society.


7

Although profiling, racial and gender discrimination are inherently bad, if utilizing

software in which these issues are present saves a life or protects citizens from harm, the end

justifies the means. This ties back to the idea that it is the outcome and consequence that matters

most when making an ethical decision. Despite the flaws and potential inaccuracies in facial

recognition software, is it not better to save lives or stop crime if given the means to do so?

Applying the tenets of consequentialism would indicate that this is the ethically correct choice.

Allowing for the continued use of this technology is correct due to the lives it has saved

and will continue to save. Ultimately, this technology has benefitted the government in a variety

of ways, permitting them to gain leads on investigations and catch criminals that may not

otherwise be captured. The consequences of its continued use outweigh any possible ethical

concerns given its role in keeping society safer.

Government agencies would lose a key resource in the fight against crime. The National

Institute of Science and Technology found that forensic examiners work best when facial

recognition technology is used as a support tool, increasing the accuracy of their performance. It

has a variety of benefits for government agencies, including but not limited to identifying

victims, catching terrorists, solving sex crimes, fighting human trafficking, and solving violent

crimes (Parker, 2020). To discontinue its use it would take away an effective means of fighting

crime and keeping people safe.

Student Position

Facial recognition software should be used to a limited extent; however, it should not be

used by the government as a means of policing its people. We should continue to use this

software and advance it, but also ensure that biases in the system are adequately addressed and
8

resolved. Regulations and oversight are essential to its continued use by government agencies,

especially when being used to identify potential criminals.

This position aligns most with those that believe the government should not be utilizing

this technology without first addressing the large inconsistencies and disparities between races

and gender. While the technology itself is not inherently biased, the artificial intelligence used in

this software is perpetuating racism and sexism. Despite these problems, there are means to

address these issues and they should be explored more aggressively. In no way should the use of

this technology be discontinued, nor should development be halted. Instead there are ethical

ways to not only address these ongoing issues, but to advance technology in a way that is

beneficial to all.

While it’s easy to make judgments on whether or not this technology should be utilized,

concrete action needs to take place in order to address the ongoing underlying issues. There must

be continued research into artificial intelligence and the ways in which bias can be minimized in

the data sets used. Additionally, there needs to be clear boundaries and limitations on the use of

this software, especially by a government agency. The lack of oversight and accountability must

be addressed through regulations and laws limiting its usage and mandating the disclosure of its

use in any system. Stopping the use of artificial intelligence in facial recognition software

altogether would be a disservice to advancements in this technology. Instead, creating ethical

limitations and continuing research into bias would not only mitigate the issues, but allow for

improvement of this technology as a whole by increasing its accuracy and decreasing its error

rate.
9

References

Athanassoulis, N. (n.d.). Virtue Ethics | Internet Encyclopedia of Philosophy. Internet

Encyclopedia of Philosophy. Retrieved from https://iep.utm.edu/virtue/#:~:text=Most

%20virtue%20ethics%20theories%20take

Hardesty, L. (2018, February 11). Study finds gender and skin-type bias in commercial artificial-

intelligence systems. MIT News; Massachusetts Institute of Technology. Retrieved from

https://news.mit.edu/2018/study-finds-gender-skin-type-bias-artificial-intelligence-

systems-0212.

Hassanin, N. (2023, August 21). Law professor explores racial bias implications in facial

recognition technology. News|University of Calgary. Retrieved from

https://ucalgary.ca/news/law-professor-explores-racial-bias-implications-facial-

recognition-technology.

Najibi, A. (2020, October 24). Racial Discrimination in Face Recognition Technology. Science

in the News; Harvard University. Retrieved from

https://sitn.hms.harvard.edu/flash/2020/racial-discrimination-in-face-recognition-

technology/

NEC New Zealand. (2020, May 26). A brief history of facial recognition. NEC. Retrieved from

https://www.nec.co.nz/market-leadership/publications-media/a-brief-history-of-facial-

recognition/

Parker, J. (2020, July 16). Facial Recognition Success Stories Showcase Positive Use Cases of

the Technology. Security Industry Association. Retrieved from

https://www.securityindustry.org/2020/07/16/facial-recognition-success-stories-

showcase-positive-use-cases-of-the-technology/
10

Seven Pillars Institute. (n.d.). Applying Utilitarianism: Are Insider Trading and the Bailout of

GM Ethical? Retrieved from https://www.sevenpillarsinstitute.org/ethics-101/applying-

utilitarianism-are-insider-trading-and-the-bailout-of-gm-ethical/#:~:text=Consequentialist

%20theories%20have%20been%20around

U. S. Government Accountability Office. (2021, August 24). Facial Recognition Technology:

Current and Planned Uses by Federal Agencies. Retrieved from www.gao.gov.

https://www.gao.gov/products/gao-21-526

You might also like