This piece is also posted by Stop LAPD Spying Coalition here.
In an increasingly policed and surveilled world, algorithms have become critical sites of power, struggle and resistance. By telling the story of our fight against PredPol, a predictive policing technology, we are introducing an abolitionist model for analyzing algorithms: The Algorithmic Ecology. The Algorithmic Ecology is both a framework and an organizing tool that can be critically applied to any algorithm. This model decenters the algorithm itself, looks at the different actors that shape the algorithm, and illustrates whose interests the algorithm serves, with the ultimate goal of dismantling the actors creating algorithmic harm. The Algorithmic Ecology is also a reminder: We must critically examine what lurks beneath “scientific,” data-driven policing, and we must go beyond technology-centered critiques of algorithms and “dirty data.”
The Algorithmic Ecology has served the Stop LAPD Spying Coalition as an essential tool for understanding and communicating our fight against the pseudoscientific PredPol program in Los Angeles. The Algorithmic Ecology of PredPol is a story of how race and poverty are policed in the United States. It is a story of how land and bodies are contained, controlled, and criminalized. It is a story of a land grab, of displacement and banishment, and of the continuing violence of settler-colonialism and white supremacy. The Algorithmic Ecology ultimately serves as a roadmap to lay bare PredPol’s intent to cause harm to Black, Brown, Migrant, and poor people, just like all policing programs and technologies before it. Thinking through the ecology informs our resistance, and our journey toward abolition of all policing—data-driven and otherwise.
We ground the story of the Algorithmic Ecology in Skid Row, a community on the east side of downtown Los Angeles and home to the Stop LAPD Spying Coalition. A community where Broken Windows policing programs like the Safer Cities Initiative (now known as RESET) and city municipal codes like 56.11 and 41.18D regularly criminalize daily life. Under Broken Windows policing, activities like sitting, jaywalking, resting, sleeping, or tossing a cigarette are enough to earn a ticket which, if left unpaid, can lead to jail time. Central to our story is PredPol, an algorithmic, “predictive,” policing technology that was forged for the battle fronts of Iraq and Afghanistan and then unleashed on the streets of Los Angeles. The PredPol platform claims to be scalable across such disparate contexts (war zones and domestic neighborhoods in different cities) through its “neutral” and “efficient” data-driven approach. We go beyond the fog of these claims to outline the broader ecology of PredPol, paying attention to the ideologies that underpin their rationalist, racist logics, and the institutions that operationalize technologies like PredPol for domination and control. The Algorithmic Ecology is a call to action!
Beyond the Feedback Loop
On June 1, 2018 the Stop LAPD Spying Coalition filed a public records request for LAPD’s PredPol hotspot maps of its Central Division, a division that includes the community of Skid Row. This records request was part of a broader strategy to expose harm in the community and ultimately dismantle Predpol. Predpol purports to “predict” crime by using an algorithm to process historical crime data. The proprietary algorithm outputs 500×500 square foot boxes or “hotspots”—maps used to direct LAPD officers to patrol areas where crime is supposedly most likely to occur. Data-driven policing programs like PredPol claim to be more objective, less-biased forms of policing because they focus on place/property and not “people,” conveniently forgetting that location functions as a proxy for race and class in a place as violently segregated as Los Angeles.
The dominant critique of PredPol is that this form of data-driven policing creates harmful feedback loops. Biased crime data, collected from already hyper-policed communities, is fed into the algorithm, so the algorithm would necessarily determine that crime is occurring most in these same hyper-policed areas where the data was invented. On November 8th, 2018, LAPD released hotspot maps of Central Division from June 21st to December 31st, 2015, and another from March through June, 2018. The Coalition analyzed these records and created density maps to visualize where hotspots occurred over time.
PredPol hotspot maps from 2015 (left) and 2018 (right). The star on each map designates the location of the Stop LAPD Spying Coalition, which is located in Skid Row.
Given the prevailing notion that algorithmic policing would create “feedback loops,” our expectation was to find Skid Row—the area around the star marked on the map—to be laden with PredPol hotspots. But the hotspots were instead clustered at the periphery of the community. Rather than visualizing the hyper-policing that we know occurs in Skid Row, the PredPol hot spot maps appears to be drawing a digital border to contain, control, and criminalize Skid Row.
We sought to uncover this violent quarantining that PredPol was accelerating in our community with a framework that looked beyond just the algorithm and the data. As part of the strategy to dismantle PredPol, the Coalition partnered with Free Radicals, an activist collective that organizes at the intersections of science and social justice. This partnership coupled the Coalition’s ongoing analysis of PredPol and broader efforts to dismantle police surveillance with Free Radicals’ anticolonial lens on science and technology. Together with the community, we mapped out the connections, histories, and political and financial interests in our existing research on PredPol and policing in Los Angeles, and arrived at the notion of an Algorithmic Ecology, pictured below. The next section introduces how the Algorithmic Ecology functions as both a framework and a tool.
Algorithmic Ecology: Framework and Tool
The Algorithmic Ecology is a framework that tells a more complicated, but more complete story of the role of algorithms in our lives. It emphasizes that algorithms—PredPol or any other data-driven technology—cannot be the sole focus of critique because they are not developed or implemented in a vacuum. The algorithm is designed to operationalize the ideologies of the institutions of power to produce intended community impact. The community impact is part and parcel of these systems’ design, not an aberration that can be addressed at the technical or operational level. The ideologies of the institutions (at the top of the graphic) are inseparable from the harmful, violent effects on the community (at the bottom). Our framework separates components into four layers: community impact, operationalization, institutions, and ideologies. This structure shows how power moves at a macro and micro level, and how the operationalization of these programs mediates institutions and communities. The arrows represent the flows and relationships between the network nodes. At its core, this a relational framework—an ecology; while we’ve created a hierarchical diagram, we recognize that these flows and relationships are dynamic and ever-evolving.
The Algorithmic Ecology is a tool that maps, visualizes, and communicates the relationships of power that surround any algorithmic technology—from facial recognition to automated risk assessment, from social media use to hiring software. In the process of mapping, we not only visualize the power and intention to inflict harm, but also seek to highlight points where the community has opportunities to resist, dismantle, and abolish systems of oppression altogether. Each Algorithmic Ecology map will contain different nodes and flows from those in our PredPol example, revealing the central actors in your fight.
The Algorithmic Ecology of PredPol
In order to help readers (all of us!) use the Algorithmic Ecology as an organizing tool, this section will break down each component of the framework, using PredPol as an example.
First Layer: Community Impact
We start with the ways that algorithms impact communities: the harm and violence that they enact and perpetuate. The community impact layer represents the people who are most impacted by the algorithm. Some questions that guide our thinking about community impact include: Who is the community that is targeted by and most impacted by this algorithm? What kinds of harm does the algorithm create or intensify? Are we understanding the extent of this harm by engaging with, listening to, and learning from the community?
PredPol fortified a boundary that was already very familiar to the community of Skid Row. The Los Angeles Community Action Network, a community organizing center in Skid Row, had named it the Dirty Divide. This segregation and quarantining of space on the east side of Downtown LA meant residents could not access basic life necessities like toilets, trash cans, and water while the west side was sprouting hundreds of vacant luxury lofts. In its earlier literature, PredPol strongly encouraged Broken Windows policing. Broken Windows policing is a tactic that aggressively punishes people for low-level “crimes” like sitting on a sidewalk or jaywalking. These punitive measures bear the flawed assumption that preying on minor offenders / offenses will somehow prevent other forms of more violent “crime.”These racist, violent tactics were already very familiar to the community of Skid Row. Returning to our density maps above, we note that the PredPol hotspots show up only on redeveloping land near these westside lofts. This is better understood in the context of PredPol’s simultaneous promotion of Broken Windows policing. By targeting, ticketing, arresting, murdering, and displacing Black, Brown, poor and unhoused community members who exist in gentrifiable land, these policing logics and their related technologies and programs create the condition of attrition through enforcement while simultaneously digitally redlining a community. Here, PredPol lends a data-driven, “objective” authority to advance these violent practices.
Second Layer: Operational
The next layer addresses the way that this harm is enacted, or operationalized. Represented in this operational layer are the technical components of the PredPol platform and the actors who legitimize and implement these technologies. This layer lends a veil of objective authority mentioned above to obscure the aims and assumptions inherent in any technology. Guiding questions include: What is the function of the algorithm? How is the data collected? Who interacts with the algorithm? What do they do with it? And why?
LAPD touts PredPol as a purely objective science that can solve decades of racist policing with math. Instead, the algorithm enhances and accelerates the inherently violent underpinnings of policing. As illustrated in the Coalition’s report Before the Bullet Hits the Body, crime, and hence crime data, is a racist social construct that has been developed over time to target and control marginalized communities. When PredPol labels communities as ‘criminal,’ officers are primed to find something suspicious in these hotspot areas, and many neighborhoods are already viewed by police as inherently suspicious. The critique of the feedback loop, and the adage “dirty data in, dirty data out” speaks to this bias (bias in the data, the data collectors, and the data users) but doesn’t go beyond it. The centering of the technology, data bias, and dirty practices does not provide a broad enough lens to understand the motivations and violences of algorithmic policing. For example, it leaves out entirely the political processes of land development that work in concert with policing and the specific consequences to the unhoused community of Skid Row and elsewhere.
Additionally, the operationalization of PredPol’s hotspots involves many stages of subjective input. For instance, crime analysts process data at ‘Community Safety Operations Centers’ (CSOCs) to produce patrol routes for officers. Crime analysts identify which hotspots to highlight, and officers determine how and when to patrol the hotspots; it is completely up to individual discretion, and therefore subject to human bias.
Third Layer: Institutional
Before even the first line of code of PredPol was ever written, an entire infrastructure of government agencies, academia, and private interests converged to set PredPol in motion. These institutions uphold harmful ideologies and fund, create, or otherwise influence the development and implementation of technology. Guiding questions include: Who funds the program? Who benefits? What interests came together to create the program?
The person most intimately involved in the development of PredPol is Jeff Brantingham, a UCLA Professor of Anthropology. Brantingham’s research into predictive policing began with funding from the Department of Defense for a counterinsurgency project that would predict where terrorist attacks might occur in the battlefields of Iraq and Afghanistan. In several training materials, Brantingham compares insurgents in the Middle East to young Latino Youth. While PredPol was originally publicly funded as a war technology destined for the frontlines of Iraq and Afghanistan, Brantingham simultaneously imagined its profitable application in our own backyards.
Brantingham and the LAPD worked out a collaboration: in exchange for free access to PredPol, Brantingham could have access to LAPD’s crime data. Los Angeles thus became a ‘testbed’ for PredPol’s development, with an initial experiment in LAPD Foothill Division and then a citywide experiment that made Los Angeles a lab and its residents experimental subjects. Although the technology was developed with public money, Brantingham went on to patent the algorithm and founded a for-profit corporation called PredPol, Inc., which now sells the PredPol platform to police departments all over the country and beyond.
In his 2018 State of the City address, Mayor Garcetti praised predictive policing as the wave of the future. He claimed this program would protect communities, and especially property. However, Garcetti’s enthusiasm for and entanglement with Predpol became more clear in the summer of 2019. The Mayor’s office is responsible for appointing the Los Angeles Board of Police Commissioners (LAPC). The LAPC is the civilian oversight board of LAPD in charge of overseeing all policies and procedures of the police department. In August of 2019 it was discovered that one of the Police Commissioners, Sandra Figueroa-Villa, had taken money from PredPol under the direction of Mayor Garcetti for her non-profit organization El Centro Del Pueblo. Thus, a medley of individuals and institutions—from non-profit, private, government, and academic sectors—were upholding and profiting off of LAPD’s PredPol experiment.
Stoking public fears is a way to manufacture the consent needed to use vast cuts of our civic budgets for policing projects. The PredPol hotspot map visualizes the ‘Dirty Divide’ that segregates those whose property is deemed valuable (“New Downtown”), from the ‘others’ who are deemed property-less (Skid Row) and who often have their property stolen by the City Sanitation Dept. and the LAPD. Academic institutions like UCLA are not only complicit in this narrative, but also advance it through research and tech development. Police oversight bodies reveal themselves to be at best complicit, and at worst as active stakeholders in harmful policing regimes. LAPC members themselves profit from PredPol, so how can they serve as an impartial oversight body for the people? At the same time, the Mayor and members of the City Council benefit directly from the downtown development projects that PredPol protects, again serving profit over people.
Fourth Layer: Ideological
This final layer, ideologies, explores the values that undergird and are perpetuated by the Algorithmic Ecology. Guiding questions include: What ideas motivate the actors within this system? What belief systems are upheld, intentionally or unintentionally, by actors and institutions in the Algorithmic Ecology?
It is clear that the violence directed toward the community of Skid Row is racist and anti-Black; Skid Row is a community that is predominantly Black. Powerful institutions like the DOD and even our local government have long upheld ideologies of “dangerous people” and imperiled property to justify the creation and use of weaponry like PredPol. Data-driven policing becomes a new way to justify racial disparities in policing. But this attempt to banish Black, Brown, and poor people from the valuable land of Downtown Los Angeles has its roots in white supremacy, anti-blackness, settler colonialism, capitalism, and patriarchy. The story of Skid Row demonstrates the intersections of all these harmful ideologies at work. Throughout the 1980s, the Skid Row community extended all the way to Pershing Square. Over time, the boundary has been pushed several blocks east of Pershing Square, now beginning on Los Angeles Street and reaching east to Central Street. In October 2016, the LA’s Dept of City Planning proposed to rezone parts of Skid Row to include market rate housing as part of the Downtown community plan update, or DTLA 2040. Just as for the plantation owners, the overseers, and slave patrols of the past, it comes to no surprise that the Mayor, LAPD, and all other private interests have a vested interest in PredPol, a program that ultimately serves broader interest in gentrification plans with a veneer of data-driven “scientific objectivity.”
Relationships of Power: Arrows
Building out each layer tells a more comprehensive story, but that story is not complete without the interrelationships which we show via arrows. The arrows between different actors in the Ecology represent the many ways that power is exerted between these actors, whether through direct relationships or other forms of influence. These arrows also show that connections are not just between entities within the same layer but entities in different layers as well. Algorithms are not operationalized without funding, influence, and directives from institutions, just as the community is not impacted without direct and indirect actions of operationalizers, institutions, and their ideologies. Guiding questions for placing the arrows include: Where does money change hands? Who has decision-making power? Where are softer or less explicit forms of influence or motivation apparent?
These arrows help us identify how ideologies which may seem abstract and institutions which may seem distant are directly connected to community impacts. In the case of PredPol, settler colonialism is an inherent ideology in processes of gentrification that benefits Los Angeles city officials, who commission the use of technologies like PredPol by the LAPD to quarantine the Skid Row community. This is one concrete way that the ideology of settler colonialism manifests itself today.
- Community Impact: people who are most impacted by the algorithm, and who hold the power to resist it.
- Operationalization of Power: the technical components of the algorithm and the people who directly use those technical components.
- Institutions of Power: the institutions and entities that uphold the ideologies of power and fund, create, or otherwise influence the development and implementation of technology.
- Ideologies of Power: the values and ideologies that undergird and are perpetuated by the Algorithmic Ecology.
- Relationships of Power: the ways that power is exerted between entities.
How do you disrupt an Algorithmic Ecology? Abolition, not reform!
Many reform efforts have been proposed to cushion the failings of algorithmic systems, such as transparency, training, oversight, and addressing bias in data and in the algorithm. However, none of these interventions meaningfully alter the ecology as a whole. The Algorithmic Ecology illustrates that inserting additional actors only strengthens the ecology as whole by adding more stakeholders without addressing the institutions and ideologies of power that benefit from its existence. This was recently demonstrated when the Police Commissioners themselves shut down a public Police Commission meeting, preventing dozens of people from giving public comment on PredPol’s harms, and then voted to continue the use of PredPol behind closed doors. The only way to dismantle harm is to abolish these systems of dominance and oppression.
Stop LAPD Spying Coalition and Free Radicals targeting various points of the PredPol Algorithmic Ecology. From L to R, top: graphic encouraging community members to join us at the October 15th, 2019 LAPC meeting where LAPD responded to the OIG audit, members at the July 24, 2018. Public Hearing on Data Driven Evidence-Based Policing at LA City Hall. From L to R, bottom: Community gathered for People’s Press Conference before the Oct 15th LAPC meeting, flyer for an event on March 4th, 2020 at UCLA to end algorithmic policing.
The Algorithmic Ecology is not just an analytical tool, but a tool for resistance. It is a tool that shows that abolition is both necessary and possible. It demonstrates why we’ve been bringing our fight to the LAPD, LAPC, City Hall, as well as the halls of UCLA and to those technologists that herald algorithms as the foregone future. In sharing this framework, we aim to arm campaigns that are fighting against technologies of oppression and the broader national security police state that seeks to track, categorize, and control us.
The Algorithmic Ecology is about much more than PredPol. It is about the current and future residents of Skid Row, and their right to live without fear of criminalization, harassment, and violence. It’s about the thousands of algorithmic ecologies that impact Black, Brown, and poor communities across the globe – the algorithms that dictate who is deemed worthy of getting a loan or a job, who is classified as a gang member, and any algorithm that is designed to coerce and control communities under the false guise of objectivity or efficiency. Now, it’s your turn. Take the template below and make it your own. Map your fight, and your resistance. Join our movement to dismantle these ecologies of harm, and to build a future where every person is deemed worthy of dignity, power, and liberation.