In the late 1800's, Cesare Lombroso published his work titled "L' homme criminel" - which directly translates to Criminal Man. This work was Lombroso's attempt at tying biological measures to measures of crime and deviance. In Criminal Man, Lombroso proposed that criminal proclivity was directly inherited and could be empirically studied by examining physical differences in facial structure (amongst other factors).
Lombroso was a product of his time - from the late 1800's to the early 1900's, Charles Darwin's influence on biological studies had spurred a huge movement in the social studies towards biological determinism. Biological determinism, in short, argues that our social behavior is hinged upon our genetic makeup.
Today, Lombroso is largely characterized as the father of criminology -- despite the fact that his work was rooted in racist notions of social behavior. Through the years, criminologists have followed in the footsteps of Lombroso, constantly seeking biological elements which are tied to criminal and deviant behavior. For an example of this contemporary work -- see MAOA gene research.
So what does all this mean for the topic of predictive policing?
With the rise of artificial intelligence (AI), police in the U.S. have collaborated with criminologists, anthropologists, and computer scientists in pursuit of creating systems that predict criminal behavior. This work can be viewed as an extension of Lombroso's work given the predictive aim of predictive policing systems.
Predictive policing, as an AI concept, was birthed in 2008 in Los Angeles, California. LAPD stopped using their predictive policing system in 2021. Other police departments have chose to continue using predictive policing applications. Here is a non-exhaustive list of some of the departments that still use predictive policing to this day:
1) Miami Police Department
2) Miami-Dade Police Department
3) New York Police Department
4) Albuquerque Police Department
5) Detroit Police Department
Earlier this year, predictive policing applications were banned in the European Union (EU) as a part of their AI Act. Predictive policing gets a number of things wrong:
1) The algorithms rest on the assumption that crime is objective and measurable. Crime and deviance, as we know them to be, are highly subjective phenomena that are constantly being created, defined, and destroyed. Criminal law is a living document that is open to change. Notions of criminal behavior, and what constitutes a criminal action, have waxed and waned through the years. Crime is not objective, it is a highly subjective phenomena that is open to interpretation.
2) Predictive policing algorithms utilize past cross-sectional data that inaccurately reflects the current reality of criminal behavior. Using past data is great for telling us about the past, but it does nothing to predict or explain the future.
3) Predictive policing algorithms perpetuate harmful racial biases by making associations between racial characteristics (indirectly) and criminal proclivity. This is the fallacy of American racism - the belief that some racial groupings of peoples are more likely to commit crime is a truly American belief.
4) Predictive policing assumes that crime is empirically detached from a person's lived experience. The idea of crime as being an objective feature of individual behavior is flawed because crime is an interpretive phenomena that is based in human interaction.
5) Predictive policing reflects human biases that are especially harmful to people of color in the U.S. -- it leads to over-policing of geographic areas that are concentrated with Black and Brown people. It leads to wrongful convictions. It destroys families and communities.
Predictive policing is a human expression of technology in its most abstracted form. We have the power to restrict the utilization of these technologies -- and we most certainly should. Work is currently under way to develop ethical safeguards and frameworks related to the use of AI in the criminal legal system.
By denying the mainstream assumption in criminology that crime can be predicted based on empirical features, critical criminologists can begin to demonstrate how crime is a social construct.
-M