Towards an Abolitionist AI: the role of Historically Black Colleges and Universities
aa r X i v : . [ c s . C Y ] J a n Towards an Abolitionist AI: the role of HistoricallyBlack Colleges and Universities
Charles C. Earl ∗ Automattic.com [email protected]
Abstract
Abolition is the process of destroying and then rebuilding the structures that im-pede liberation. This paper addresses the particular case of Black folk in theUnited States, but is relevant to the global decolonization movement. Usingnotions of abolition and infrastructures of feeling developed by Ruth WilsonGilmore, I view
Historically Black Colleges and Universities ( HBCUs ) as aparticular kind of abolitionist project, created for the explicit purpose of nurturingand sustaining Black excellence particularly within the sciences. I then examinehow artificial intelligence (AI) in particular and computing in general have con-tributed to racial oppression and the further confinement and diminishing of Blackexistence. I conclude by examining how the space held by HBCUs in computingmight contribute to a reimagining of AI as a technology that enhances the possi-bility and actualization of Black life.
As a child, I attended the nursery school of Spelman College, a historically Black women’s collegelocated in Atlanta. One of my earliest memories is of witnessing a protest taking place on the campus.The young women of Spelman, in their proud defiant afros, were protesting the visit of NelsonRockefeller. They were demanding that the Rockefeller family – a major funder of the college –divest from all assets invested in the South African apartheid regime. Rockefeller’s grandfather,John D Rockefeller, had given the college its first major endowment, and what had been the AtlantaBaptist Female Seminary had been renamed in honor of John Rockefeller’s wife, Laura Spelman.In an age that demands resistance and the abolition of structures that continue to oppress Blackpeople, what is the role of historically Black institutions like Spelman College? And in particularwhat should be the role of those institutions in defining how Black people interact with and definethe development of AI? Given the outsize role these institutions play in the development of Blackcomputing professionals – 9% of Black undergrads attend HBCUs, yet HBCUs produce 37% of theBlack STEM graduates in the United States – it is important to assess how these graduates are pre-pared to address the inequities in computing that exacerbate these harms. Further, these institutionscontinue to receive significant funding from sources whose aims are at odds with liberation. In thispaper, my hope is to raise questions and initiate dialog about how these institutions might contributeto liberatory computing.
The geographer Ruth Wilson Gilmore has been one of the most eloquent and incisive voices ar-ticulating what abolition means in 2020. Her book
The Golden Gulag
Gilmore (2007) is both a ∗ All opinions and analysis in this work are the author?s alone.Resistance AI Workshop editation on the campaign to slow the growth of state and private prisons in California as well asan exploration of the harm done by removing scores of people from society for the sake of retribu-tion. In the specific case of prisons, Gilmore is asking us to confront what we think prisons are for –punishment? rehabilitation? justice? – and the reality of what they actually accomplish.Though Gilmore’s work focuses primarily on prison abolition, she articulates a broad, encompassingvision of liberation, developed out of her geographer’s view of the world. In her essay
AbolitionGeography and the Problem of Innocence
Gilmore (2017) she writes
Abolition geography startsfrom the homely premise that freedom is a place (Gilmore, 2017, pg. 227). Abolition then is overallthe affirming of life out of places intended for domination, exploitation, and ultimately spiritual,intellectual and physical death:If unfinished liberation is the still-to-be-achieved work of abolition, then at thebottom what is to be abolished isn’t the past or its present ghost, but rather theprocess of hierarchy, dispossession, and exclusion that congeal in and as group-differentiated vulnerability to premature death. (Gilmore, 2017, pg. 228)It is then this process of dispossession of life itself – epitomized by prisons – that is the enforcementof the inequality necessitated by capitalism. The central tenet of Gilmore’s work is the examinationof these qualities of existence. The carceral landscape of capitalism that demands an extractivehierarchy on the one hand, and the affirming alternate reality that the exploited imagine and createin real time and in real spaces on the other.But Wilson’s work sees liberation as a dual of abolition. Referencing W.E.B. DuBois’s classic
Black Reconstruction
If capitalism requires inequity, and if racism enshrines it Gilmore (2017), then the hyper profitmotive of AI applied to platform capitalism has been a successful effort to optimize racism. Theuses of AI – particularly in platforms that derive value from manipulative practices – are necessarilynot in the interest of Black people. Benjamin (2019), Buolamwini and Gebru (2018) and othershave documented how AI has been used to enforce racial hierarchies of power. Cottom (2020)further discuses how the logic of racial capitalism has been ensconced into the logic of all web scaleplatforms – it is at this point core to Internet economy.From above work we can distill a few kinds of technologies and practices that reinforce racial op-pression: 2
Carceral technologies : for example predictive policing and algorithmic sentencing are usedexplicitly for the purposes of capturing, isolating, and destroying Black life.•
Discriminatory resource allocation existing technologies are used to limit the access ofBlack people to resources from health care to credit, often optimizing for profit over well-being.•
Surveillance technologies are used explicitly for expropriation of behavioral surplus andcircumscribing autonomy by predictively tracking and controlling the behavior of thesurveilled.•
Racialized hiring and software engineering practices practices of recruiting, advancement,and project management within software organizations that enable the reproduction ofracial harms.•
Militarization of AI particularly in the secretive use of technology to further colonialistagendas on the global south – the use of this technology upon the civilian Black and Indige-nous population of the U.S. is particularly concerning.On the other hand, there are tools that Benjamin identifies as being able to dismantle these tech-nologies, mitigate their harm, or even better create the spaces that allow free spaces to come intobeing. •
Creating space for and voice to the marginalized , the practice of centering marginalizedvoices by enabling full participation, contribution, and equitable control of the developmentprocess.•
Algorithmic and process audits that identify the harms caused by algorithmic decision mak-ing and also identify the systemic issues in the software development process that allowthem to go unchecked.•
Racial abuse protection and mitigation technologies that intercept racial attacks and miti-gate harm (e.g. micro-aggression classifiers)•
Liberatory design practices which enable the development of software products center theaspirations and requirements of marginalized community members.•
Equitable work practices that recognize and reward the contribution of data annotators, con-tent moderators, and other "gig" workers whose value is extracted and erased by racializedplatform capitalism.
We can get some understanding of the degree to which abolitionist computing within HBCUs occursby looking at the research funding. The major U.S. federal agencies – among them Department ofDefense (DoD), National Institutes of Health (NIH), and National Science Foundation (NSF) – allo-cate large grants specifically targeted to improve the research and teaching capacity of HBCUs. TheNSF annually funds the HBCU Excellence in Research (EiR) program ($10 million per year), andthe Historically Black Colleges and Universities Undergraduate Program (HBCU-UP) ($55 millionper year).A close look at research activity at Howard University raises the other question of to what degreeBlack computing scholars are being trained to participate in the reproduction of racist technology. AtHoward, approximately $15 million awarded in computing research for 2020 (excluding on-goinggrants), 50% were from department of defense grants in bio-metrics and cyber defense (e.g. Howardwas awarded $7.5 million for a DoD Center for Excellence in AI/ML CoE-AIML (2020) ), 4%from educational awards for development of retention and STEM curricula, while 1% was awardedto research specifically aimed at addressing online abuse directed toward Black people (e.g. "DeepLearning of Sentiment Analysis and Codeswitching for Identification of Harmful Technologies" andother initiatives from Dr. Gloria Washington’s Affective Biometrics Lab ). Similar patterns emergewhen looking at other HBCUs like North Carolina A & T State University.That is, with heavy DoD and platform capitalism support Black students and researchers must navi-gate being participants in the reproduction of what Mohamed et al. (2020) refers to as colonizing AIespecially with respect to surveillance and military technology. This is concerning especially giventhe potential harms for Black folk in the U.S. and global south Garcia (2019).3
Abolitionists agendas for HBCUs
To begin in earnest the development of abolitionist approaches to AI within the HBCU landscaperequires a return to the source. Tapping into the infrastructure of feelings that already permeatesthese institutions and enables their success in nurturing so many impactful people. Within availableresources already devoted to capacity development, there could be informal meetups, conferences,gatherings across schools, across disciplines to explore the implications of abolitionist approachesto computing. The significant number of students involved in BLM actions combined with resourcesalready devoted to making computing accessible could be a powerful accelerator. The developmentof conferences and centers to promote tools for community audits and interrogation of softwareplatforms is another approach that could have wide adoption and immediate impact given the mil-lions of Black people are impacted daily by racialized ad targeting and unfair pricing experimentsby software platforms. Lastly, the Association for Computing Machinery (ACM) has long acknowl-edged the dearth of adequate ethics education in computing. For AI in particular, this is especiallynecessary given the potential for harm. Perspectives on ethics that would explore methodologiesfor assessing what should and should not be built would be an important preparation for students.Engagement with scholars of the social sciences, law, and healthcare – present at so many HBCUs– that could collectively work on the development of archives and models capable of addressingcommunity needs is also important.
I have attempted to consider the current state of affairs in AI viewed from the perspective of abolitionas articulated by Ruth Wilson Gilmore. Viewing AI as a carceral space that replicates and acceleratesinequity I have attempted to look at the way in which HBCUs might chart a new course for it. Thereremains much to be done.
References
Douglas L. Allen. 2020. Black Geographies of Respite: Relief, Recuperation, and Resonance atFlorida A & M University.
Antipode (2020). https://doi.org/10.1111/anti.12658
Ruha Benjamin. 2019.
Race after Technology: Abolitionist Tools for the New Jim Code.
Polity.Joy Buolamwini and Timnit Gebru. 2018. Gender shades: Intersectional accuracy disparities incommercial gender classification.. In
Conference on fairness, accountability and transparency .DoD CoE-AIML. 2020 (accessed November 17, 2020).
Howard University DoD Center of Excel-lence in Artificial Intelligence & Machine Learning . Tressie McMillan Cottom. 2020. Where Platform Capitalism and Racial Capitalism Meet: TheSociology of Race and Racism in the Digital Society.
The Sociology of Race and Ethnicity (2020).W. E. B. Du Bois. 1998.
Black reconstruction in America: 1860-1880 . New York : The Free Press.Eugenio V. Garcia. 2019. The militarization of artificial intelligence: a wake-up call for the GlobalSouth. (September 2019). https://dx.doi.org/10.2139/ssrn.3452323
M. Gasman and T. Nguyen. 2014. Historically black colleges and universities (HBCUs): Leadingour nation’s effort to improve the science, technology, engineering, and mathematics (STEM)pipeline.
Texas Education Review
2, 1 (2014), 75–79.Ruth Wilson Gilmore. 2007.
Golden gulag: Prisons, surplus, crisis, and opposition in globalizingCalifornia . University of California Press.Ruth Wilson Gilmore. 2017. Abolition Geography and the Problem of Innocence. In
Futures ofBlack Radicalism , G. T.. Johnson and A. Loubin (Eds.). New York: Verso, 225–240.Howard University Affective Biometrics Lab. 2020 (accessedNovember 17, 2020).
Affective Biometrics Lab Research . https://sites.google.com/view/affective-biometrics-lab-resea/contact?authuser=0https://sites.google.com/view/affective-biometrics-lab-resea/contact?authuser=0