AI in Social Justice from a High School Student That Evaluates the Criminal Justice System
Written by Ameya Kamani
Want to learn more about AI in Social Justice? Read on to learn more about AI in our criminal justice system and the ways social change is being impacted by it.
How does social justice merge with AI?
How does Artificial Intelligence play into bias via Criminal Justice?
What is social justice?
So…what is social justice? The formal definition says that social justice is the “process of building participation and challenging injustice both at the state and community level” (Coburn and Gormally 62). Basically, social justice is recognizing issues (whether based on race, sexuality, gender, socioeconomic status, ethnicity, etc) and wanting to rectify those issues.
Social justice is directly tied to one core principle: empathy.
Empathizing with people and feeling their pain and wanting to aid them is what it means to be an activist.
Naturally, social justice intersects with AI and coding because creating a program, website, or product which is meant to recognize or rectify injustice is social justice.
Often, we have a narrow view of social justice: we imagine activism solely as attending protests, but the world of helping people is endless !
How to apply AI to social justice
Essentially, AI’s purpose is to show patterns. Recognizing patterns may be able to help people who are wrongfully convicted of a crime or treated too harshly by the justice system.
Several people who are incarcerated have been involved with the justice system before, so having a program that distinguishes patterns in crime and criminal demographics can be used to reverse unjust indictments.
By consulting AI programs, “there is the potential to discuss the rules given by law, allowing for a more restorative process, rather than a process characterized by punishment” (Hadzi). Essentially, seeing bias within the justice system can help for restorative techniques through justice.
I wanted to give a specific example of this quintessential relationship between activism and coding: my Inspirit AI Criminal Justice project. The overall goal of the project was to recognize bias through AI recidivism predictors.
WHAT IS RECIDIVISM?
Well, recidivism is basically the possibility for someone to commit another crime. Obviously, there is an issue with recidivism predictors: racial bias.
Let’s compare the cases of two people who committed crimes : Vernon Prater and Brisha Bordon.
Brisha Bordon is a black woman with 4 juvenile misdemeanors and no subsequent offenses. Vernon Prater is a white man who has committed 2 armed robberies and did have a subsequent offense: grand theft.
Prater was given a “low risk” rating of 3 via Northepointe’s COMPAS program and Borden was given a “high risk” rating of 8.
Clearly, there is an issue with the COMPAS program.
The white man who did recidivate was given a significantly lower rating than the black woman who did not.
The model is clearly biased in the way it predicted recidivism, likely because of the programmers who created the model.
Whenever you write an essay or have a speech, inevitably your own opinion will shine through. The prejudiced values of the programmers was evidently shown in their program of predicting recidivism.
The Future of AI and Social Justice
Through this project as well as through an understanding of social justice, the relationship of AI in social justice is clear. AI is just one of the many modes of conveying social justice and activism. Every program has a specific purpose, so the possibility of what purpose your code serves is endless. You can be passionate about social change and still love to code: interests merge through the world of coding.
Sources
https://www.um.edu.mt/library/oar/bitstream/123456789/68179/1/ISEA_PROCEEDING_271020.pdf
https://docs.google.com/presentation/d/1ejQiMsf_r6Zg8C8mLoWceug1JRr6dnIemvEFITZpOTc/edit?usp=sharing
COBURN, ANNETTE, and SINÉAD GORMALLY. “SOCIAL JUSTICE AND EQUALITY.” Counterpoints, vol. 483, Peter Lang AG, 2017, pp. 51–76, http://www.jstor.org/stable/45177771.
