The ethics of AI and its racial bias within the Legal system

Topic Outline

Artificial Intelligence has for years cultivated an idea of clean, machinery and as a future-oriented field. However, that’s not the case for a lot of the programmes being produced at the moment. A lot of these AI systems are perpetuating outdated and harmful ideologies, that clearly show tendencies towards racist and sexist views.

According to Data USA, a 2017 study shows 73.1% of people awarded a degree in this field in the US are men 1, where 63.6% of these are white 2. Back in 2017, only 1.42% that were awarded a degree were African American 3. This inequality in the working field is bound to have programmes that show inequality and bias towards a certain race and gender. This is not, however, a critique on the AI field alone. These discrepancies in percentages can be found in a lot of courses and areas. So, when advocating for change in these ratios and inequality, we are advocating for a reform of the educational system as a whole.

To put this into a more practical example, back in 2014 a woman by the name Brisha Borden and her friend saw a bike on the side of the road and took it, however, the owner of the bike was able to catch them in the act. She was later convicted for burglary and petty theft. Similarly, Vernon Prater was charged with a similar crime when he was caught shoplifting a Home Depot store. Prater had previously been charged with more serious crimes like armed robbery and served 5 years in prison. However, Borden also had a record but it was only misdemeanours committed as a juvenile. When both these cases were compared by an AI programme, on the likelihood of any of them committing a crime again, the programme clearly showed a racial bias when predicting this value, since Borden was a black woman and Prater a white man. It’s also important to note that 2 years later Vernon Prater committed a new crime and was arrested for it, whereas Brisha Borden has not. This is, however, one case, there are multiple other cases judged by an AI programme that show the same patterns and misjudgments 4.

I wanted to write about this topic mainly because I always had a big interest in the way AI processes information and how useful and helpful it can be when used right. I think that despite all the fascination with the concept and the potential of artificial intelligence, I think it’s important to talk about where this type of technology falls short. I also chose to combine the topic of AI with gender/race politics because I think some of the best essays I wrote are when I’m able to write about my political views. I’m a big advocate for equality and I hope I can translate those ideas through my practical work but also through my dissertation.

---

1 Integrated Postsecondary Education Data System (IPEDS) (2017) Gender Imbalance for Common Institutions. Available at: https://datausa.io/profile/cip/artificial-intelligence#demographics (Accessed: 30 March 2020);

2 Integrated Postsecondary Education Data System (IPEDS) (2017) Race & Ethnicity by Gender. Available at: https://datausa.io/profile/cip/artificial-intelligence#demographics (Accessed: 30 March 2020);

3 Integrated Postsecondary Education Data System (IPEDS) (2017) Race & Ethnicity by Degrees Awarded. Available at: https://datausa.io/profile/cip/artificial-intelligence#demographics (Accessed: 30 March 2020);

4 Angwin, J., Larson, J., Mattu, S., Kirchner, L. (2016) Machine Bias. Available at: https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing (Accessed: 2 April 2020);

Key Text

Ezekiel Dixon-Román is an associate professor at the University of Pennsylvania, teaching on the Social Policy Program at the moment. Him, along with Ama Nyame-Mensah and Allison R. Russell, wrote an article by the name Algorithmic Legal Reasoning as Racializing Assemblages 5, which relates to the topic I’m exploring. This written work examines U.S. statistics that predict the likelihood of a certain individual to commit a crime in the future.

I also want to point out that I choose an article that talks about U.S. data rather than the U.K. because I want to discuss the racial bias of AI within the legal system and the U.K. is only using these artificial intelligence systems to proof-read contracts/policies and other areas that don’t have a direct impact on someone’s sentence. However, it’s important to note that, according to Delloite and other workers in the area, the U.K. is going to start using A.I. judges to determine legal sentences in the near future 6.

The article begins by describing how algorithms work, and how both government and private organisations benefit from the use of A.I. as a source to get access to big data, such as user preference, in order to create better market strategies. The legal field is no different, they have been using AI to analyse for a long time now, Román explains that this act of analysing social behaviours in order to predict future actions through numbers has a strong relation to the philosophical ideology of positivism 7. Auguste Comte, author of many positivism writings, defends that there is “universal truths for human behaviour and social phenomena”, “empirical observation is the only rational means by which universal truths in the social world can be discovered” and “through the application of the scientific method, causal relationships among social phenomena can be established” 5. This ideology defends that it’s possible to analyse someone’s social behaviour through numbers and that this constitutes empirical evidence of such. If this is true then we can fairly believe the predictions of the artificial system since the data is to be assumed truthful. However, what if the data has been compromised by past acts?

During the late 1960s, urban crime and drug related crimes saw a massive growth in the United States with President Johnson. However, it was with president Nixon’s mandate that the infamous War on Drugs programme started 8. This programme targeted especially young African Americans and the anti-war movement. It was later revealed, by Nixon’s former domestic policy chief, John Ehrlichman, that this was indeed a targeted motion towards these groups of people. “You want to know what this was really all about…The Nixon campaign in 1968, and the Nixon White House after that, had two enemies: the antiwar left and black people. You understand what I’m saying. We knew we couldn’t make it illegal to be either against the war or black, but by getting the public to associate the hippies with marijuana and blacks with heroin, and then criminalising both heavily, we could disrupt those communities. We could arrest their leaders, raid their homes, break up their meetings, and vilify them night after night on the evening news. Did we know we were lying about the drugs? Of course we did.” 9

The damage done during this period can still be seen today, with black Americans being four times more likely to be arrested for marijuana charges than their white counterparts. African Americans make up nearly 30 percent of all drug-related arrests, even though they only account for 12.5 percent of all substance users.10

Román’s article also posses interesting alternatives on how to even out this racial disparity that AI machines have when it comes to predicting the likelihood of an individual committing a crime again. He frames a corrupted system, led by a colonialist mentality that lacks other communities voices. “Predicting risk [criterion of pathology] is a deficit-based framing masquerading as an asset-based perspective, which legitimates power structures by characterising individuals as unreliable, inherently “deviant” or “abnormal,” unable to speak for themselves, and dependent on the decisions, knowledge, and opinions of others.” 5

I think that this goes back to the debate as to whether the correctional facilities should serve as punishment or as a rehabilitation of the person. In the U.S. prisons serves as a form of punishment for someone’s crimes, that’s evident not jut in the way prisons are built but even the sentence length. Architecture and prison design play a big role in the way the country sees crime punishment. Halden Prison is a maximum security prison in Norway that has been recognised as one of the most humane prisons in the world. This prison is designed to mimic life outside of prison, Gudrun Molden, one of the prison’s architects, says that the punishment is to take away your freedom and that after that they need to focus on rehabilitating of the individual 11. And this type of penal system seems to be working for Norway with only 20 percent of the prisoners offending again. In comparison the U.S., according to a 2019 report by the Sentencing Commission, shows that nearly 64 percent of prisoners who had been charged with violent crimes were arrested within eight years and 40 percent for nonviolent offences. 12 Despite this, a lot argue that this type of humane prison wouldn’t work within the U.S. legal system since both countries have very different social and economic backgrounds. The economic differences between people within Norway are less greater than in the United States. The conditions of the Halden Prison are much better that a large part of U.S. housing, so this would become almost as an incentive to commit crimes, in order to live a better life.

To sum up, just like Román concluded, even if you come up with alternative ways to solve this bias within the Artificial Intelligence legal field, it still seems like a long way until we get equality because of the social relationship history. Román presents other alternatives just as interesting to explore, such as redefining and reframing the assessment model based on the risk the crime posses for society. He gives the example of white-collar criminals posing a bigger economic risk than marijuana consumption, however, these are the crimes that get more surveillance, mainly because white-collar crimes are normally committed by white businessmen, so they are not seen as such a threat. For these algorithms to work, we first need to fix the racial and gender bias in all social aspects, in order to use empirical evidence if that is even possible in a human society.

---

5 Dixon-Roman E., Nyame-Mensah A. and Russell A. Algorithmic Legal Reasoning as Racializing Assemblages Computational Culture 7 (2019);

6 Delloite, Report Developing legal talent: Stepping into the future law firm (2017);

7 Comte A. Introduction to Positive Philosophy, trans. Frederick Ferré, (1988);

8 Schoenfeld H. The War on Drugs, the Politics of Crime, and Mass Incarceration in the United States (2012);

9 LoBianco T. Report: Nixon’s war on drugs targeted black people, CNN (2016);

10 Pearl B. (2018) Ending the War on Drugs: By the Numbers. Available at: https://www.americanprogress.org/issues/criminal-justice/reports/2018/06/27/452819/ending-war-drugs-numbers/ (Accessed: 20 April 2020);

11 Benko J. (2015) The Radical Humaneness of Norway’s Halden Prison. The goal of the Norwegian penal system is to get inmates out of it. Available at: http://www.antoniocasella.eu/archipsy/Benko_26mar15.pdf (Accessed: 25 April 2020)

12 Deady C. (2014) Incarceration and Recidivism: Lessons from Abroad. Available at: https://www.salve.edu/sites/default/files/filesfield/documents/Incarceration_and_Recidivism.pdf ( Accessed: 25 April 2020);

References

Achoenfeld, HA & Schoenfeld, HA (2012) The War on Drugs, the Politics of Crime, and Mass Incarceration in the United States, Journal of Gender, Race and Justice, vol. 15, pp. 315-352;

Angwin, J., Larson, J., Mattu, S., Kirchner, L. (2016) Machine Bias. Available at: https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing (Accessed: 2 April 2020);

Benko J. (2015) The Radical Humaneness of Norway’s Halden Prison. The goal of the Norwegian penal system is to get inmates out of it. Available at: http://www.antoniocasella.eu/archipsy/Benko_26mar15.pdf (Accessed: 25 April 2020);

Comte A. (1988) Introduction to Positive Philosophy, trans. Frederick Ferré;

Deady C. (2014) Incarceration and Recidivism: Lessons from Abroad. Available at: https://www.salve.edu/sites/default/files/filesfield/documents/Incarceration_and_Recidivism.pdf ( Accessed: 25 April 2020);

Delloite (2016) Report Developing legal talent: Stepping into the future law firm. Available at: www.deloitte-uk-developing-legal-talent-2016.pdf (Accessed: 25 April 2020);

Dixon-Roman E., Nyame-Mensah A. and Russell A. (2019) Algorithmic Legal Reasoning as Racializing Assemblages Computational Culture 7. Available at: http://computationalculture.net/algorithmic-legal-reasoning-as-racializing-assemblages/#fn-3258-14 (Accessed: 27 April 2020);

Integrated Postsecondary Education Data System (IPEDS) (2017) Gender Imbalance for Common Institutions. Available at: https://datausa.io/profile/cip/artificial-intelligence#demographics (Accessed: 30 March 2020)

Integrated Postsecondary Education Data System (IPEDS) (2017) Race & Ethnicity by Gender. Available at: https://datausa.io/profile/cip/artificial-intelligence#demographics (Accessed: 30 March 2020);

Integrated Postsecondary Education Data System (IPEDS) (2017) Race & Ethnicity by Degrees Awarded. Available at: https://datausa.io/profile/cip/artificial-intelligence#demographics (Accessed: 30 March 2020);

LoBianco T. (2016) Report: Nixon’s war on drugs targeted black people, CNN;

Pearl B. (2018) Ending the War on Drugs: By the Numbers. Available at: https://www.americanprogress.org/issues/criminal-justice/reports/2018/06/27/452819/ending-war-drugs-numbers/ (Accessed: 20 April 2020);