:00:00. > :00:00.That is it from me. Kate will be here at 10.00, but first of all it
:00:00. > :00:24.is time for Click. We are now more surveilled
:00:25. > :00:28.than we have ever been. Authorities are gathering
:00:29. > :00:31.data on its citizens. It would be all too easy
:00:32. > :00:34.to confuse the real world Mr Marks, my mandate of the District
:00:35. > :00:41.of Columbia Pre-Crime Division. I'm placing you under arrest
:00:42. > :00:43.for the future murder of Sarah Marks and Donald Dubin,
:00:44. > :00:46.that was due to take place today, April 22,
:00:47. > :00:48.at 0800 hrs and four minutes. In the movie Minority Report,
:00:49. > :00:54.the Pre-crimes Unit race to arrest would-be offenders before
:00:55. > :00:56.they have a chance to Now, they use psychics but it turns
:00:57. > :00:59.out, something similar In Chicago, where the violent
:01:00. > :01:03.crime rate has exploded, law enforcement has been forced
:01:04. > :01:05.to try out unconventional Authorities are attempting
:01:06. > :01:17.to combine various technologies in an effort to predict
:01:18. > :01:19.where and when violent Marc Cieslak went to
:01:20. > :01:23.Chicago to find out more. Violent crime in Chicago has seen
:01:24. > :01:40.a dramatic increase. RADIO: A 15-year-old
:01:41. > :01:42.male, shot in the neck. We need a wagon
:01:43. > :01:47.with a body bag also. The drug industry is what helps
:01:48. > :01:50.them fuel the violence, by being able to pay
:01:51. > :01:53.for their activity. In 2016, 726 murders were committed
:01:54. > :01:56.in the city, a 19-year high. That's more than the number
:01:57. > :01:59.of murders committed in New York Chicago is a city most famously
:02:00. > :02:03.known as the Windy City. More recently, it has earned
:02:04. > :02:05.a nickname that few residents That's because gun crime
:02:06. > :02:19.is so extreme in some neighbourhoods, they are comparing
:02:20. > :02:21.them to war zones. The issue has received increasingly
:02:22. > :02:23.negative attention in the US, with President Trump tweeting,
:02:24. > :02:25."If Chicago doesn't fix the horrible carnage going on,
:02:26. > :02:41.I will send in the Feds". The response from Chicago's Police
:02:42. > :02:43.Department is a new initiative, driven by technology,
:02:44. > :02:46.which aims to predict where crimes The University of Chicago's Urban
:02:47. > :02:49.Labs are assisting the police in its efforts to integrate this
:02:50. > :03:05.technology into its operations. We have a lot of expertise
:03:06. > :03:08.in analysing crime patterns and trends in the city,
:03:09. > :03:10.from years of working with data And so we are leveraging that
:03:11. > :03:14.expertise to really help the Police Department think
:03:15. > :03:16.about where it should be allocating its resources
:03:17. > :03:23.to be most effective. So what kind of data or information
:03:24. > :03:27.is it that the police are providing We have a number of datasets
:03:28. > :03:31.that we work with from them, including data on crime patterns,
:03:32. > :03:33.actual crime incidents, A number of different methods
:03:34. > :03:36.of analysis are used, including machine learning
:03:37. > :03:38.and predictive analytics. This is software which takes large
:03:39. > :03:41.volumes of data and tries These trends can then help predict
:03:42. > :03:56.where a crime might occur next. This is a heat map of
:03:57. > :03:58.homicides in District 7. And we are looking at this year over
:03:59. > :04:01.year, from 2011 to 2016. And basically, what you see
:04:02. > :04:05.on the map is the darker the red, the more concentrated homicides
:04:06. > :04:09.were in a given area. What sort of factors are you finding
:04:10. > :04:12.are influencing crime in these Yeah, so, most of the prediction
:04:13. > :04:15.that we're doing is space-based. So, yeah, it's locations
:04:16. > :04:17.that are nearby that are high-risk locations,
:04:18. > :04:19.like a 24-hour liquor store, a gas station,
:04:20. > :04:21.where people tend to congregate. The weather seems to be playing
:04:22. > :04:24.a very big role in the data. You know, we've just had a beautiful
:04:25. > :04:27.weekend and we just had significantly worse amount
:04:28. > :04:29.of shootings than we had The police are using these
:04:30. > :04:40.predictive tools to inform the deployment of officers
:04:41. > :04:42.and resources to areas where they think crimes
:04:43. > :04:44.are likely to occur. Neighbourhoods in Chicago's West
:04:45. > :04:46.and South Side are some It is these neighbourhoods
:04:47. > :04:50.which have been chosen to test We are just driving
:04:51. > :05:05.through Chicago's South Side now. Now, this is one of the areas
:05:06. > :05:08.which has experienced the highest incidence of violent crime,
:05:09. > :05:10.mainly gun and drug related. To see how all of this
:05:11. > :05:13.different kit works, I'm on my way to a police station
:05:14. > :05:16.which acts as a command centre, bringing all
:05:17. > :05:20.of the technologies together. Heading up the project
:05:21. > :05:22.is Deputy Chief Jonathan Lewen So this is our Strategic
:05:23. > :05:28.Decision Support Center. So this is where you bring
:05:29. > :05:30.all of your different This is the first time
:05:31. > :05:35.that this level of technology integration has been done,
:05:36. > :05:37.not only here, I think, So what can we see on the screens
:05:38. > :05:41.we have got around us? So, all around us are various sensor
:05:42. > :05:44.inputs, cameras, gunshot detection. The screen behind you is
:05:45. > :05:50.something called Hunch Lab, which is a geographic prediction
:05:51. > :05:53.tool that brings a lot of data into a model to predict risk
:05:54. > :05:56.for future violence. So what you are seeing on these
:05:57. > :05:59.little boxes here are areas where the model is recommending
:06:00. > :06:01.that we deploy resources and implement strategies to fight
:06:02. > :06:03.some of the violence And then it is telling us
:06:04. > :06:07.that we should deploy resources, visit businesses, do foot
:06:08. > :06:09.patrol, various tactics. Shot Spotter just very quickly
:06:10. > :06:11.triangulates possible gunshot events using acoustic sensors that
:06:12. > :06:13.are located throughout the district, and it shows the officer exactly
:06:14. > :06:16.where, accurate to within 25 yards, And you can actually play the audio
:06:17. > :06:29.of the gunshot event, So here's an event
:06:30. > :06:35.with nine rounds fired. And in this case, you can see
:06:36. > :06:39.the location is actually the back yard of a house,
:06:40. > :06:43.so that's going to be very accurate. So this is the decision support
:06:44. > :06:46.system, and this is where everything It will soon be available in the
:06:47. > :06:50.hands of officers on smartphones. So in this case, we are looking
:06:51. > :06:54.at a 911 call of a robbery that just There are four cameras within a 300
:06:55. > :07:00.foot radius of that call. Here is the real-time
:07:01. > :07:03.video from those cameras. These guys here, these
:07:04. > :07:05.are possible suspects, or... These are people that might
:07:06. > :07:07.possibly be involved? How do we know that this
:07:08. > :07:11.is identifying the right people? We find when we test and measure
:07:12. > :07:15.them, that the model's recommendations, because we can
:07:16. > :07:18.backdate it, we can look at a known outcome period
:07:19. > :07:20.and see how it performs. And we know that it's picking
:07:21. > :07:23.the right people because we know But some of this technology
:07:24. > :07:30.is proving to be controversial, It's called the Strategic
:07:31. > :07:32.Subjects List. and locations, this list
:07:33. > :07:36.is concerned with predicting crimes Just like Hunch Lab is a place-based
:07:37. > :07:50.risk model, this is a person-based risk model that is looking
:07:51. > :07:53.at variables such as arrest activity, so have you been arrested
:07:54. > :07:56.for a gun offence in the past? So it's using some
:07:57. > :08:01.crime victim data. Is your trend line in criminal
:08:02. > :08:03.activity increasing or decreasing? What was your age at the time
:08:04. > :08:06.you were last arrested? Nothing about race,
:08:07. > :08:08.nothing about gender, It is using objective
:08:09. > :08:15.measures to determine risk It's basically telling us that this
:08:16. > :08:19.person is 500 times more likely than a member of the general
:08:20. > :08:21.population to be involved in a shooting, either
:08:22. > :08:27.as a victim or an offender. So in here, we can see his
:08:28. > :08:29.affiliations, his gang affiliations. We can see also his,
:08:30. > :08:33.is this his arrest record You can see that he has
:08:34. > :08:37.a weapons arrest. He was arrested here
:08:38. > :08:39.for aggravated battery. So here's a first-degree
:08:40. > :08:41.murder charge. Here's another arrest,
:08:42. > :08:44.this is a narcotics arrest. So the score estimates how much more
:08:45. > :08:48.likely an individual is to be the victim or the perpetrator
:08:49. > :08:52.of a violent crime. The police use this score
:08:53. > :08:55.to inform what they call This is not designed
:08:56. > :09:03.to be a punitive tool. This is used to drive what we call
:09:04. > :09:06.a custom notification process, which is literally a site visit
:09:07. > :09:08.to this subject, to say, "You've come to our attention
:09:09. > :09:10.for these reasons. We want to get you out
:09:11. > :09:13.of the cycle of violence. We can offer you the
:09:14. > :09:15.following social services". Maybe if they have children at home,
:09:16. > :09:19.it would be childcare services. "But also, if you don't leave
:09:20. > :09:21.the cycle of violence and you keep committing crimes,
:09:22. > :09:23.you're going to be subject to enhanced criminal penalties",
:09:24. > :09:26.because you're a repeat gun And can you see why,
:09:27. > :09:29.if police officers go and visit somebody out of the blue,
:09:30. > :09:32.it might seem like they are being Everybody who has a risk score has
:09:33. > :09:37.committed a crime in the past. Otherwise they wouldn't
:09:38. > :09:39.even be in the model. Groups like the American Civil
:09:40. > :09:41.Liberties Union, though, disagree. They aren't happy about the use
:09:42. > :09:44.of some of these technologies. The police showed us a database
:09:45. > :09:47.of people who have been involved in violent crime in the past,
:09:48. > :09:49.and an algorithm which suggests if and when they might again be
:09:50. > :09:52.involved in a violent crime. Oftentimes in large numbers,
:09:53. > :09:59.along with a number But what they won't say is
:10:00. > :10:02.what social services are offering. Is it just them or is it
:10:03. > :10:05.their entire family? What is the success
:10:06. > :10:07.rate once that occurs? The fact is, is that most
:10:08. > :10:18.of the people who are charged for... You know, if you take two
:10:19. > :10:21.people who are arrested for a simple drug possession,
:10:22. > :10:23.if one is white and one is African-American,
:10:24. > :10:25.the African-American is far more likely to be charged,
:10:26. > :10:27.maybe even convicted. We have seen that there has been,
:10:28. > :10:30.you know, in essence, a "once convicted, always guilty"
:10:31. > :10:32.sort of theme that While there might be disagreements
:10:33. > :10:45.about the use of this technology, everybody I spoke to had similar
:10:46. > :10:47.ideas about an ultimate solution to tackling
:10:48. > :10:52.violent crime in Chicago. It's got to be every,
:10:53. > :10:54.everybody that's a stakeholder in this coming together
:10:55. > :10:57.to solve the problem. What is really needed across this
:10:58. > :11:01.city is a commitment I think a lot of it has to do
:11:02. > :11:06.with preventing, with healing, and creating a space
:11:07. > :11:08.where individuals can civically And that's it for the short cut
:11:09. > :11:19.of this week's Click. The full-length version has
:11:20. > :11:21.a really fascinating story about a bunch of geeks trekking
:11:22. > :11:23.across the Arctic for charity. If you'd like to watch that, check
:11:24. > :11:27.out Click on the iPlayer right now. Follow us on Twitter at BBC Click
:11:28. > :11:29.throughout the week. Thanks for watching
:11:30. > :11:40.and we'll see you soon.