LAPD
ALGORITMS

In Los Angeles, police helicopters are not reserved for emergencies, but are a routine part of police surveillance. 

Los Angeles has the largest municipal police helicopter operation in the world.

$1056 per hour, when factoring in personnel, fuel, and maintenance.

In 2020, the cost of helicopter fleet maintenance alone was $23 million.

Disproportionately target neighborhoods with large minority populations.

The LAPD's interest in predictive policing- a science that uses algorithms to make predictions about future crimes- began around 2010. For nearly a decade they relied on two predictive policing programs: PredPol and operation LASER (Los Angeles Strategic Extraction and Restoration).

PredPol used the time and location of historical crime data and created 500 by 500 square foot areas, called "Hot Spots".

LASER analyzed the locations of reported crime and arrest data and used it to create LASER Zones, which were subjected to increased police surveillance for 9 to 12 months.

Psychological toll from constant surveillance takes on citizens.

PredPol uses an algorithm based earthquake prediction to "predict crime".

Another form of systematic racism.

PredPol targets people of color and puts them at disproportionate risk of harm.

Academics confirm major predictive policing algorithm is fundamentally flawed (simplistic and harmful).

Software "based on nearly seven years of detailed academic research into the causes of crime pattern formation... The mathematics look complicated - and it is complicated for normal mortal humans - but the behaviors upon which the math is based are very understandable."

PredPol looks forward and projects when and where crime will most likely occur with a seismology algorithm used for predicting earthquakes and their aftershocks.

The algorithm models evidenced based research of offender behavior, so knowing where and when past crime has occurred, PredPol generates probabilities of where and when future crime will occur.

"Repeat victimization" of an address

"Near-repeat victimization" (The proximity of other addresses to previously reported crimes)

"Local search" (Criminals are likely to commit crimes near their homes or near other crimes they've committed)

The academic foundation for PredPol's software takes a statistical modeling used to predict earthquakes and apply it to crime. Much like how earthquakes are likely to appear in similar places, the papers argue, crimes are also likely to occur in similar places.

"I would say in our mind, the key difference is that in earthquakes models, you have seismographs everywhere -  wherever an earthquake happens you'll find it. The crux of the issue really is that to what extent are you able to get data about what you're observing that is not also totally on the model itself." - Suresh Venkatasubramanian

"Because this data is collected as a by-product of police activity, predictions made on the basis of patterns learned from this data do not pertain to future instances of crime on the whole. In this sense, predictive policing is aptly named: it is predicting future policing, not future crime." - Venkatasubramanian

Unless every single crime is reported, and unless the police pursue all types of crimes committed by all people equally, it is impossible to have a reinforcement learning system that predicts crime itself. Instead, you'll just be creating a self-fulfilling prophecy where police find crimes in the same places they've been told to look for them, rather than everywhere.