Big Data and “PredPol”: STOP Crime before It Happens

predpol_hero

By: Nikola Rujkov, Mythili Manirajah, Finn Klebe, and Min Wu

Why did we choose “PredPol” and what does it have to do with Big Data?

In general, working with big data means to analyze th huge amount of data that is very hard to process when using traditional data mining systems and algorithms. “PredPol”, the company we chose for our project, works in the same way. Often the vast amount of data is collected in order to maximize profits, for example, by tailoring advertising to each and every customer individually. Moreover, the collected data is, in the most cases, highly personal and confidential, and possibly even collected without the customers knowing it. We, as a group,  had the aim to find other branches in which neither profit maximization, nor the collection of highly confidential data is on top of the agenda. In the case of “PredPol”(Predictive Policing), big data is used to assist the human beings. However, instead of maximizing profits, meeting the noble goal of  maximizing citizens’ security is intended to be facilitated.

When we, as a group, came across this company, we found it intriguing to see how big data broadens the opportunities of human beings. Without forestalling the summary of their service, the tool “PredPol” is used in the police branch, to be more precise in criminology. However, it is not used to replace police officers – usually being one of the main issues of advanced technology, but rather to assist and encourage them to use their workforce as effective as possible. We find the overall aim of the tool to make all of us feel safer in our own environment remarkable and likely to be able to constitute a role model of big data usage in a moral way. As a result, we consider the tool worthy enough, to dig deeper in order to find out how it works and if it is able to confirm our first thoughts and aspirations.

“I’m not going to get more money. I’m not going to get more cops. I have to be better at using what I have, and that’s what predictive policing is about… If this old street cop can change the way that he thinks about this stuff, then I know that my [officers] can do the same.”

-Los Angeles Police Chief Charlie Beck

Short summary of the service provided by Predpol

“PredPol” was founded in 2012, based in Santa Cruz, California. “PredPol” is a secure, cloud-based software-as-a-service, developed by a team of PhD mathematicians, criminologists, and social scientists. It is an invaluable added tool that allows our police to use their patrol time more efficiently and helps stop crime before it happens – this technology offers an excellent crime-fighting solution that will ultimately make our lives safer. The core technology includes prediction of drug crime, gang crime, anti-social behavior, and the recently released gun violence prediction tool. In other words, “PredPol” tracks various types of crime: burglary, robbery, vehicle theft, theft from vehicles and gun crime as well as traffic accidents.

kentonline

Only three criteria of data are used to make predictions – type of crime, place of crime, and time of crime. No personal data is utilized in making these predictions. This predictive policing gives officers a significantly better idea of when and where to be so that they can deter crime. There is a proven track record of crime reduction in communities that have deployed “PredPol”.

First obvious opportunities and threats

PredPol plays a vital role in ensuring public safety because it enables law enforcement agencies to take preemptive steps ahead of crimes and reduce the crime rate in regions that are prone to repeated attacks. “PredPol” saves time, money and energy (logistics) because law enforcement agencies are more focused and only have to deploy their resources when it is necessary. In other words, it ensures maximum efficiency and effective implementation of invaluable resources to prevent crime. In addition, “PredPol” is cost effective because it is based on pre-existing crime records that are safely transferred and analyzed to produce accurate crime predictions. It takes advantage of the available data and reproduces it for a different purpose at a minimal cost as the company only expends money to maintain the software. However, the company could be forced to work with old data if by chance the law enforcement crime database is not updated on a regular basis as the database is the foundation of “PredPol’s” predictions. This, in turn, could affect the company’s relationship with its customers as they could potentially switch to “PredPol’s” competitors whose services might not depend on the law enforcement crime database. In addition, “PredPol’s” predictions are somewhat unreliable in that they only pinpoint the time, day, place and very limited information on the criminal. Law enforcement officials could misjudge innocent individuals and arrest them solely based on the prediction. There are also concerns about the use of predictions to convict a criminal as such predictions are only mere speculations and don’t necessarily reflect the true intentions of the individual on trial. It is unclear whether the predictions are reliable and hence one may wonder whether it is ethically acceptable to carry out convictions based on technological predictions. Moreover, “PredPol” is less effective when considering crimes that less repetitive in nature or the new one for that matter because there is very minimal data on such crimes; hence, PredPol is required to wait until there’s enough data to produce reliable and accurate predictions on such crimes. This, in turn, could be too late in some cases and often times law enforcement officials act based on their intuitions. It is also why law enforcement agencies report a decrease in crimes such as theft and burglary (crimes that are repetitive in nature) whilst reviewing the effectiveness of “PredPol”.

Timeline

Name Distributed Sessions Meeting Time
All members Summary on Big Data and Social Media Oct 14th
Nikola Rujkov Key Partners, Key Activities & Key Resources Oct 14th
Min Wu Value Propositions & Channels Oct 28th
Mythili Manirajah Customer Relationships & Customer Segments Nov 11th
Finn Klebe Cost Structure & Revenue Streams Nov 25th

Further videos for more information 🙂

PredPol- The Science Behind Predicting Crime

Software predicts violent crime

Preliminary Bibliography

“Don’t Even Think about It.” The Economist. The Economist Newspaper, 20 July 2013. Web. 05 Oct. 2016.

http://www.economist.com/news/briefing/21582042-it-getting-easier-foresee-wrongdoing-and-spot-likely-wrongdoers-dont-even-think-about-it

Garriss, Kirstin. “Intelligence Based Software Helps Police Predict Crimes.” YOUR4STATE.

Nexstar Broadcasting, Inc, 04 July 2014. Web. 05 Oct. 2016.

http://www.your4state.com/news/news/intelligence-based-software-helps-police-predict-crimes

“Predictive Policing Pros and Cons.” PredPol RSS. Web. 05 Oct. 2016.

http://www.predpol.com/predictive-policing-pros-and-cons/

“Proven Results of Our Predictive Policing Software | PredPol.” PredPol RSS. Web. 05 Oct. 2016.

http://www.predpol.com/results/

Advertisements

2 thoughts on “Big Data and “PredPol”: STOP Crime before It Happens

  1. It is nice to see that there are actually alternative ways to make use of Big Data beyond commercial uses. I think that the idea behind predpoll is well-intentioned but to be honest I question the reliability of this service.
    As far as I understood it, it predicts where, when and what type of crime will happen based on a database. In your blog entry you mentioned that “Law enforcement officials could misjudge innocent individuals and arrest them solely based on the prediction.” I totally agree with this. This database will probably contain basic information about the criminal, including the migrational background, religion or race of the criminal. My concern is that this could be used as a legitimization for harsher treatment of e.g. African-Americans or citizens from the Middle East. There is already evidence for a bias when it comes to these two specific groups, even without this service being used in all American states.
    Another possible threat to this service is, if the database gets hacked so that criminal groups get access to the predictions. Even if it predicts the location and time of a crime, it would then serve as a warning system for criminals. They could then avoid those places, which would have the opposite effect of counteracting them being found and arrested.

    Like

    1. “I think that the idea behind predpoll is well-intentioned but to be honest I question the reliability of this service.”

      To be honest, what really concerns me, is that this system might be very reliable (and it looks like they are, although the known results might simply reflect and movement crimes to different places, as opposed to an actual reduction).

      “Law enforcement officials could misjudge innocent individuals and arrest them solely based on the prediction.”

      Well, at least current systems (the ones already running) are not actually used to decide whether suspects are to be arrested. And in any case, I doubt that the rate of false positives of police officers equipped with these systems would be worse that that of them without.

      However, in at least one case they do give us a good estimation on their reliability, so we would have to decide which percentage of false positives/false negatives we will be willing to accept, a question that does not arise in current more complicated (for instance humans) systems with a less predictable reliability.

      “My concern is that this could be used as a legitimization for harsher treatment of e.g. African-Americans or citizens from the Middle East. There is already evidence for a bias when it comes to these two specific groups, even without this service being used in all American states.”

      That is actually a very valid point. When considering that these systems deploy policemen, solely based on patterns and not at all based on causal expectations, you might want to conclude that the system is solely relying on discriminative data. So even though it might be effective, we will have to ask ourselves whether and when we really want its deployment.

      “Even if it predicts the location and time of a crime, it would then serve as a warning system for criminals. They could then avoid those places, which would have the opposite effect of counteracting them being found and arrested.”

      Actually, this wouldn’t work in practice. As soon as the clever criminals adjust to the predictions, the system (which is basically relying on correlations) would notice the new pattern and adjust accordingly. So if this would have any effect, it would help balance off the load of the crimes between different places (which might or might not be desirable).

      Still, if these systems should be developed even further and their deployed also in different usages (say judging suspects), their application will probably lead to the (innocent) people adjusting their behavior. At this point the system will adjust automatically increasing its sensitivity and forcing the people to restrict and sensor their behavior even further. This way people refusing to comply will run into troubles very soon. As these systems naturally look for deviant behavioral patterns, deviant people will get problems and the remaining ones will loose a lot of their previous freedom.
      As many of these processes happen on a subconscious level (for instance the feeling of being either anonymous or monitored changes peoples behavior also on subconscious levels). So in the end we might not even notice the steady loose of our liberties.

      Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s