Predictive Policing CRIME FORECASTING FOR THE FUTURE Its hot! PERF’s survey found that only 38 percent of responding police departments are currently using predictive policing, but 70 percent expect that they will implement this strategy within the next two to five years. This is not “Minority Report” “Minority Report is about predicting who will commit a crime before they commit it. This is about predicting where and when crime is most likely to occur, not who will commit it.”( Brantingham) Who is interested in this? Civil Rights Activists – Have concerns of these techniques intruding on the rights of citizens, especially the poor and minorities Practitioners (Analysts) – Have a professional interest on how this can make their work better / more useful Police Chiefs – Eager to find new techniques to reduce crime The U.S. Government – New forum for funding, research, literature, and evaluation The Private Sector – Sees potential for funding of research grants, consulting and software development Researchers-have background to design predictive models (Hollywood et. al) Predictive Policing This is the application of analytical techniques to identify likely targets for police intervention Borrowed from Business Walmart found that large weather events resulted in increased sales of water, duct tape and pop tarts (strawberry). They alter their supply chain to meet the demand before the storm hits. Goal Move from a reactive posture to a proactive posture. Why? ◦ Be more efficient and effective with your resources. ◦ Opportunities to prevent, deter, thwart, mitigate, and respond to crime more effectively (Beck and McCue 2015) Proactive Much of traditional policing in reactive in nature. ◦ After the crime Much of Predictive policing is more proactive ◦ Stopping something before it happens Example, Richmond VA Risk-based deployment was tested on New Year’s Eve 2003. Using a risk-based deployment strategy, police identified locations and times expected to be associated with increased complaints of random gunfire and proactively deployed police resources to those locations to prevent or deter crime or respond more rapidly. The results demonstrated increased public safety associated with the predictive-policing strategy. Random gunfire complaints were decreased by 47 percent, highlighting the deterrent effect associated with information-based deployment, while the number of weapons recovered went up 246 percent, underscoring the rapid response possible with effective prepositioning of resources. In addition, these marked increases in public safety were associated with a reduction in police resources required, resulting in a $15,000 savings in personnel costs alone during the eight-hour initiative. The ability to anticipate the time, the location, and the nature of crime supports the police manager’s ability to proactively allocate resources—preventing or deterring crime through targeted police presence and enabling rapid response by pre-positioning police assets when and where they are likely to be needed (Beck and McCue 2015). What’s in a name? The term “predictive” can be misleading. Others terms that can help understand the idea are: ◦ “proactive policing,” “preventative policing,” “adaptive policing,” “evidence-based policing” and “datadriven policing.” Why it makes sense Much crime is predictable ◦ Criminals tend to operate in their comfort zones ◦ Do what was successful in the past and generally close to the same time as place as the last crime (not all). ◦ Criminals and victims follow common life patterns; where those patterns overlap can lead to crimes – Geographic and temporal features influence the where and when of those patterns ◦ Criminals make ―rational‖ decisions using factors such as area & target suitability, risk of getting caught, etc. ◦ Can ID many of these patterns and factors; can steer criminals’ decisions through interventions (Hollywood et. al) What types of crimes Stranger offenses ◦ Robberies, burglaries, and thefts. Not as much for vice or some interpersonal violence ◦ DV perhaps Some crime will always be stochastic or random Another Example Currently, LAPD’s predictive policing program is being applied to a 50-squaremile area. The program breaks up the larger area into a grid composed of 500- foot squares, or “boxes.” Each forecast assigns a crime probability score to each box. Patrol officers are informed of the highest probability boxes and are directed to use any available time to focus on those boxes. The department encourages officers to proactively use their knowledge, skills, and experience to identify reasons why the box has a high crime risk and then actively work to address those issues (COPS 2014). https://www.youtube.com/watch?v=U0gX_z0V0nE Necessary Elements/Ideas Integrated information and operations: ◦ There are no silos, effectively integrate information and operations. Seeing the big picture: ◦ Prevention is as important as response, and every incident is an information-gathering opportunity. Cutting-edge analysis and technology: ◦ There is a wealth of tools and technology already available and it is imperative that departments learn how to use them. Linkage to performance: ◦ Understand what your impact was. Adaptability to changing conditions: ◦ This concept highlights the need for flat-networked organizations, training in how to adapt to strategies based on information and high professional standards. Business Model PP is an overall organizational posture. Prediction is only part of this model. 1. Collecting the data 2. Analyzing the data 3. Design police operations (use predictions and police experience) 4. Assessment of Criminal Response Business Process Four Main areas 1.Methods of predicting crimes ◦ Forecast places and times with an increased risk of crime. Hot spots Risk Terrain analysis ◦ Not based on past crime but on risk factors of the environment. ◦ ID geographic features that contribute to crime risks (bars, liquor stores, major roads) and make predictions about crime risk based on how close a given location is to the risk inducing feature. Four Main areas 2. Methods for predicting offenders Identify individuals at risk of offending in the future. ◦ Level of Service Inventor-Revised (LSI-R) used by corrections. ◦ Static and dynamic factors Four Main areas 3. Methods of predicting perpetrators’ identities ◦ Create profiles that accurately match likely offenders with specific past crimes. ◦ Find suspects using sensor data (GPS, LPNs) ◦ Anchor point (estimating an offenders place of work or frequent activity) based on the location of criminal activity. (assumes linked crimes) ◦ Offenders commit crimes close to their anchor points. ◦ The frequency of offending decreases as the distance from their anchor point increases. ◦ Suspect prioritization, patrol saturation, neighborhood canvasses Four Main areas 4. Methods for predicting victims of crime ◦ Identify groups or individuals likely to become victims of crime. ◦ Identify people at risk for victimization (engaged in high risk criminal behavior) ◦ Data mining to learn repeat offenders at risk ◦ Looking at network of people arrested together in Chicago. ◦ For example-living in certain places increases your homicide risk- simply being arrested during this period increases the aggregate homicide rate by nearly 50%, but being in a network component with a homicide victim increases the homicide rate by a staggering 900% (from 55.2 to 554.1). ◦ The further away you are socially from a homicide victim the lower your chances of being a victim (Papachristos and Wilderman 2013) ◦ Identify people at risk for domestic violence ◦ Data mining to find DV and other disturbances involving local residents when in other jurisdictions An example 1.Methods of predicting crimes ◦ Forecast places and times with an increased risk of crime. Hotspots identification-based on historical crime data Very small areas seem to have a large amount of crime ◦ One street block, one apartment building 4-6% account for about 50% of crime Hot spots are of consistent over time Treatment-13-15 minutes of a stationary patrol unit randomly every two hours. Hot spot map Predictions are only half Predictions are half the battle ◦ The other half is carrying out interventions based in predictions that lead to reductions in crime. Dangers Police stop following the predictions ◦ Go back to the good old ways Police given specific zones at risk treat all those in those zones as criminals. ◦ “The tyranny of the algorithm” (Marx) ◦ Harms legitimacy Hurried adoption Cost No effort to deal with underlying causes of crime ◦ There is something about those zones that make them hot. Model not well explained to officers ◦ No “Buy-in” Officers view it as micro-management ◦ Shut down More Dangers Officers and police managers trust their own decision making and don’t want competition from computers. ◦ From civilians Analysts fall in love with one method ◦ The methods of analysis have to vary with the problem. Let the analysts do their work ◦ Let the analysts worry about how the model works-you worry about how accurate it is. Find the balance in methods and implementation. ◦ A decent transparent model that is actually used will outperform a sophisticated system that predicts better but sits on a shelf. Some predictions will be wrong-focus on overall organizational efficiency. Data Must collect and analyzing data. ◦ Land use data-city planning department ◦ Incident data-RMS/CAD ◦ Time spent data-time on a call Quality and quantity of the data shape usefulness. ◦ Garbage in, garbage out. Data often difficult: ◦ Holes in data due to forms ◦ Holes in the data due to practice ◦ Sources are diverse which requires fusion-often difficult Interventions Have to have the resources to execute the interventions ◦ Money Interventions can be complex ◦ Trained them on what you want them to do Must provide information to officers that is specific enough to carry out the intervention. ◦ Communication Have to monitor the process of implementation of the intervention ◦ Process evaluation ◦ Skills to do this Myths of Predictive Policing The computer knows the future ◦ They only tell you risks not certainties ◦ “It's tough to make predictions, especially about the future.” The computer does it all for you ◦ No you have get the data, review and interpret the results, make recommendations and take action Accurate predictions automatically lead to crime reductions ◦ You have to take action to reduce crime-predictions are just directions. What you need to do it Substantial top level support. Resources dedicated to the model Enthusiastic and interested personnel Foster a good working relationship between analysts and officers Challenges Limited resources ◦ Can impact ability to collect, analyze, implement and analyze Political pressure ◦ Pressure for quick results not conducive to methodology Information management ◦ Share good data at higher level than every before Data overload ◦ Can feel overwhelmed with the amount of data that must be collected Data quality ◦ Timely, accurate and reliable Adaptation ◦ Initiatives must be adaptable to changes in operational environment Tool Box-First Steps ID data from other agencies that may be helpful in PP Create PP task force Create MOUs Develop software for PP analysis Create guidelines on how to make use of the predictive outcomes Questions?