Big Data Policing, Supreme Carpenter –– Talking Tech w/ Andrew Ferguson & Lisa Hayes
Host Brian Wesolowski tackles a supremely interesting topic, sitting down with Professor (and previous DC Public Defender) Andrew G. Ferguson. His new book," The Rise of Big Data Policing," takes a look at the latest data fueled technologies being used by police departments and examines the impact these technologies are having on society. Would you be comfortable having a score attached to you about your potential to commit a crime? <br><br>Then Brian brings in Lisa Hayes, CDT's Vice President of Strategy & General Counsel, to talk about a Supreme Court case that deals with law enforcement access to cell phone location data. As that data becomes more precise, should law enforcement be able to access information about our cell phone location without a warrant, or should that information be more fully protected under the Fourth Amendment?<br><br>More on Professor Ferguson: http://bit.ly/2yzEU1A || https://twitter.com/profferguson<br>More on his book: http://amzn.to/2zygkBP<br>More on Lisa: bit.ly/FOSOpl<br>More on the Carpenter case: http://bit.ly/2zUQXem<br>More on our host, Brian: bit.ly/cdtbrian<br><br>Attribution: sounds used from Psykophobia, Taira Komori, BenKoning, Zabuhailo, bloomypetal, guitarguy1985, bmusic92, and offthesky of freesound.org.
Top Keywords
- crime 0.017
- policing 0.016
- data 0.015
- police 0.012
- court 0.009
- score 0.008
- predictive 0.008
- warrant 0.007
- cell phone 0.007
- chicago 0.007
- cell 0.007
- information 0.006
Transcript
Welcome to Tech Talk. Bye.
CT. Tea.
Welcome to CDT's tech talk where we dish on tech and Internet policy while also explaining what these policies mean to our daily lives. I'm Brian Wasilowski and it's time to talk tech. Our personal data is everywhere but how should law enforcement be using that data And what rules should they follow to gain access? In this episode, we'll be talking about policing in the age of big data and what that means for society. First, we'll hear about how police departments worldwide are embracing big data in a variety of ways, including using it to predict whether someone might commit a crime. Would you be comfortable having a score attached to you about your potential to commit a crime? I would not be. Next, we'll talk about a supreme court case that deals with law enforcement access to cell phone location data. As that data becomes more precise, should law enforcement be able to access information about our cell phone location without a warrant? Or should that information be more fully protected under the Fourth Amendment? Police departments worldwide are embracing technology and exploring ways to leverage big data to help with law enforcement. Data driven policing often comes with the promise of increased efficiency, reduced bias, and sometimes even the ability to predict future crimes. And while the potential sounds great, what are the costs of data driven policing? In his new book, The Rise of Big Data Policing, today's guest, Andrew Guthrie Ferguson, takes a look at the latest data fuel technologies being used by police departments and examines the impact these technologies are having on society. Welcome to Tech Talk, Andrew. Congratulations Thank you so much for having me. We have a beautiful copy here, which is available. First, tell tell everyone where they should be getting this book because they're gonna be very excited after our podcast. Sure. Sure. You should support your local bookstores. I love that. Your help. You You should support independent
university presses. It's University Press Week today. And if you're a fan of big data, well, guess what? You can get it on Amazon and be tracked for the rest of your life. That's great. I hear it's at Politics and Prose, which is actually
my second favorite bookstore in Washington, DC, only behind Kramer Books just because it's closer to me. So it's easier to walk there, but they're both fantastic bookstores. So, first, but let's just talk about it. What are the most common ways the police departments are using big data right now? What are ones that people may not even know about? So right now in Chicago, Illinois,
police are using an algorithm to predict the people most at risk of violence, be they a perpetrator or a victim of a crime. A score, a threat score, is given to individuals. Anyone who's been arrested for the last four years in Chicago has a threat score by Chicago Police Department from one to 500. That score pops up on a dashboard, so a police officer stops you on the street. They look you up. Literally, a score is next to your name, an SSL score, strategic subjects list score, and that score tells the police a bit about you and obviously shapes how they might interact with you, how they might, treat you. Maybe they might be more, deferential or more defensive, and could actually lead to issues of use force. It could affect the fourth amendment. There are predictive policing, programs in 60, jurisdictions in America that focus on, places of crime. So they are predicting their forecasting areas of crime, perhaps a heightened burglary area or a heightened car theft area. And that information comes from, past crime data added in with, like, maybe the weather. If it rains, you know, bad guys don't like getting wet. If it, you know, if it's payday on Friday, there might be an uptick in robbery. So they're taking it. They're crunching all this information. So even, like, you know, big football games on Sundays can actually impact crime patterns in a particular city. You have surveillance technologies that are, across America. Right? You have, domain awareness systems, which is 9,000 linked cameras in Downtown Manhattan that's able to track you as you go from block to block, place to place. There are automated license plate readers that track every car that goes into that area. There are even automated prompts. So if someone drops a bag and walks away, the cameras will alert to it to determine if it's, like, a forgetful tourist or, you know, an ill tempered or ill intentioned terrorist. And so you have these technologies of surveillance, predictive analytics, and, of course, we have data mining, the ability to the ability to search and find crimes and patterns of data that we leave behind are changing how police do their jobs, who they're targeting, where they're going, and really, in many ways,
changes the role of police in America. Yeah. That's a lot of data out there about us now. So, obviously, all you just all you described requires data. And a lot of that data generated, processed by law enforcement. One of the things that you did a really great job, I thought of reminding readers of your book, is that data is really people.
Why does that matter so much in a law enforcement context? Because if you're one of those people in Chicago who has a heightened score, you might actually knock on the door with a detective behind that door, maybe with a social worker and someone from the community, who's there to tell you that you've made this list. In fact, they have a letter, a custom notification letter that explains your past crime history and why they think you're there. On the one hand, maybe you say that's a wonderful sort of public health approach to violence. On the other hand, it's a real form of social control. It's a form of surveillance to say, look, usually it's a young man. Look, young man, you are, on the wrong path. And we know it. We're the police and we're watching you. And so in one sense, it can be this kind of public health approach, but in many ways, it becomes sort of a virtual must wanted list. And so there are people behind the data. There are people behind the, areas of predicted crime. Right? If you happen to be in an area of forecast crime and police are supposed to be in that area looking for burglars or car thieves or whatever, it is both natural and the consequence that officers will then see the area as a place of crime. And if you happen to be walking through that, it's gonna impact how they treat you, it's gonna impact the suspicion the streets, and it's gonna impact whether or not, maybe they stop you, frisk you, and all the human dynamics
and and dimensions that go on. Yeah. Not to mention the privacy issues. Right? If you're being surveilled as you walk down Lower Manhattan, like, that's changing, like, who you're gonna associate with, it's changing what you might be doing. It's gonna impact First Amendment freedoms, and it really can impact how we live our lives. Yeah. And a lot of what you just described can be things very much out of that person's control. You know? So there's certain ones which you had said if you had committed a crime before, some people might be, like, okay. That seems reasonable. But the neighborhood you live in, I mean, these are things that aren't always in your control. Does some of this data translate across jurisdictions? So say you're in you're you live in New York and you're visiting Chicago or a different city. Do law enforcement,
police departments communicate this data between each other? So, usually, the law the local law enforcement keeps the data relatively internal to itself. But, of course, overlaying all of this are federal fusion centers, which were arose out of post-nineeleven, to sort of coordinate some of these data collection policies. We don't know much about the data in the fusion centers, except that they are collecting a lot of data and trying to do this sort of trans jurisdictional information sharing, because that was one of the concerns post-nineeleven. 11. Mhmm. Generally speaking, local policing is sticking to local police data, and it and there's so much data that they are they are collecting that's hard to share, and the systems don't really talk. But, you know, if you're talking about, like, a state like California, the jurisdictions are starting to share it because they're connecting with private companies like Palantir to start creating these sort of social network analyses of large scale criminal syndicates. Because, of course, people involved in criminal activity don't stay necessarily in one neighborhood. They move. Right? And so one of the useful pieces of, sort of this big data policing world is that, in states like California where it's large enough that you can sort of track from city to city, they're starting to collect in different areas so they can start creating a shared database for these, you know, larger crime groups and gangs.
Interesting. So let's we touched on this a little bit. Talk to me a bit about the social biases that may be baked into some of this data. You know, we've talked about how data is really the people, but how are some of our social biases baked in? Because this is clearly where a lot of civil rights groups or, civil liberties groups really, really are concerned.
And rightfully so. Right. So go out to Chicago. Right? The way you make it onto this heat list is the following, inputs. Whether you've been arrested for a crime of violence, a weapons offense, or a narcotics offense, whether you've been the victim of a crime, either a shooting or a violent offense, your age at your last arrest, the lower the age, the higher the risk. If you're younger, the higher risk. And then sort of the trend line, is this going up or down? Are you, like, more involved in crime or less? Now some of those inputs are arrests. And so if you're in a city like Chicago, which in 2017 had the civil rights department, the Department of Justice Civil Rights Division go do an in-depth study of Chicago police practicing and found racial problems, you know, endemic, systemic, structural racial problems, racial bias all the way through, and you know that arrests are discretionary, some of those inputs are going to be biased because the people behind it are going to have implicit or explicit bias, however you wanna describe it. And so if your inputs are gonna be biased, your outputs are gonna be biased. And so we have to be aware of it. That doesn't necessarily de legitimize some use of predictive policing, because, you know, the people getting shot in Chicago tend to be young people and young people of color. So there's there is, you know, it's a complex issue, but we need to be aware of it. Because one of the real dangers of big data policing was that people use it to pretend that the bias doesn't exist. Like, don't worry. This is data driven. This is objective. Certainly. We are following the data. We're not, you know, all that all too My algorithm is neutral. Right? That all too human policing concern you're worried about that caused, you know, riots in Ferguson and Baltimore, don't worry. We've turned the page because we're now doing data driven policing. And the problem is if you don't analyze it carefully, you can allow some of this bias to seep in and then become sort of reified and justified as part of the suspicion calculus that you have created. And that's kinda why I wrote the book is because I said think all big data policing has a black data problem. Black in that it's a black box problem. We can't see through it. It's not transparent. We don't know. Black is that it's racially encoded in the sense of a lot of these these, inputs do have, you know, old fashioned traditional biases that we've always had in policing, certain communities, and black because it's distorting. Like, the law that we had came of age in, like, a small data world. Like, the Supreme Court was deciding small data cases with individuals and what they could see. They didn't know, and they weren't contemplating what happens when you overlay that with a, you know, big data world. So for example, if you're that cop in Chicago and you see someone on the street who has a 500 score, which is, like, the highest score, shouldn't that affect your suspicion? I mean, some computer just told you objectively in some non objective way that this guy is more dangerous or more at risk than someone else. So doesn't that impact how you treat that person? Doesn't it impact how you shape the suspicion in your head? And we don't have a law that has really recognized yet Right. That yet, and yet we're applying that law every day in courts and everywhere else. Yeah. I mean, that's just human nature. If you saw those scores, of course, you're gonna respond to them in a certain way, and it's going to impact how you interact with that. So a lot of this, when you talk about predictive policing, it's based on kind of old crime theory, you know, that you can predict where things are. So
are these theories accurate? You know, I'm sure you've done some research on that. Is that true that you can kind of predict exactly where things are going to happen, how they're going to happen, and who's going to do it? So
if you take place based predictive policing, right, you can predict forecast a particular area that there will be an uptick in crime. The original, like, social science theory and research behind that was actually pretty strong when you studied certain kinds of crimes, burglaries, car thefts, thefts from auto. And the reason being is those kinds of crimes are sort of contagious. If there is a burglary in one neighborhood, there tends to be burglaries right around that area, in part because maybe the same burglar just came back and realized there's some environmental vulnerability here that I can exploit, or maybe he told his buddies about it, and so other people went back. Mhmm. But it's pretty common to show across jurisdictions that certain crime is kind of viral, in part because there's some environmental, like, weakness or vulnerability that criminals are exploiting. And that was really the theory behind the algorithm, which is the the original PredPol algorithm was a seismic aftershock algorithm, essentially, saying, look. Okay. If you have a a particular crime here, there will be these ripple effects, and you can trace them. And that tends to work and there is social theory behind it. As you expand to other crimes, as you expand to other, types of offenses, we don't have the data behind it. There's also social science data behind, the idea that you can predict people who are more at risk of being the victims of crime or being involved in crime. You know, in Chicago, the theory behind focused deterrence, which is kind of the underlying, idea behind some of the predictive policing technology is that there are certain people who are shooting other people because they are in sort of adversarial relations. They are either in gangs that are rivals or in drug dealers that are rivals. And the reality is that if you shoot my buddy, I shoot you. So you can almost predict that there will be a shooting because you know there has to be some retaliatory action. And the theory goes, if you can, you know, intervene then, if you can stop that sort of reaction back and forth, you can reduce crime. And that makes sense. Right? It kinda makes sense. And so the theory was, well, if we can identify inputs that sort of show who these people are and who are in these circles, we can reduce crime. The proof in the actual empirical reality of, you know, the cities and whether shootings are going up or down, it's pretty inconclusive. You know, there are times where it goes down. There are times it goes up. It's hard to you know, crime has gone down nationally in most cities. It's gone up in Chicago and a few other cities.
But, you know, in recent months, they say it's been going down, and they credit predictive policing for that. So Wow. You know, the jury's still up. Okay. So, I mean, obviously, let's go back to the scores a little bit. If I'm someone with a score, you know, whether I'm someone who is either being predicted to commit a crime or, you know, someone who predicted to have a crime committed against, shouldn't you have some sort of recourse as a citizen? Do you think what what do you think there should be? If if police departments are starting to use more data and even scoring citizens,
What sort you know, should you know that score in advance? Should you have the ability to look at that score? What are your thoughts on kind of that world of big data? I think we need a conversation. I think that Okay. We have done a really poor job of engaging citizens in the fact that any of this is happening. The part of writing the book was trying to get people engaged to see, like, maybe you should figure out and ask the chief, why are you doing this? What's the justification? Is it accurate? Is it audited? Are you sure? And if any of you have ever checked your credit scores, you know, it can be wrong on so many occasions. How are you gonna correct for, the errors here? Because we know every time we collect data, there are mistakes. There's errors. And if those errors have real liberty impacts, we really need to be careful. So let's figure out a process. It's not that we still use credit scores even though we know they're flawed, but we have a system where we can try to have some recourse, to make sure they're accurate. And some people, you know, check-in in certain ways. So maybe that's a bad example because honestly, who knows if your credit score is great or not. But we're not having that conversation now. Right? Law enforcement is changing. Policing is changing because of the impact of these big data technologies, and citizens are not, really engaged. I I go around and I ask people, here are two here's a quiz. Two things. Things. One, do you know the surveillance technologies are being used in your hometown right now? Answer is no. Even the the the experts don't really know all the technology. Right. And b, if you don't know, where would you go to find out? And that's another big gap. There's not even a place that you would go. And so in the book, I kind of propose these surveillance summits that every year, there needs to be the one at least one accountability mechanism where the chief of police and the mayor and the city council get up there and say, look, this is what we're paying, you know, this vendor for. This is why we're spending your tax dollars. Instead of building more libraries or after school programs or giving officers raises, right, we are actually using this technology, and here's why we think it it's right. And they may well be able to convince us that they're right, but right now, we're not having that conversation. We're not engaging to say, hey. As a democracy, we should be doing it. You know, there have been organizations, including this wonderful organization, the ACLU and EFF, that have begun some of the democratizing process. Right? Saying, like, maybe we need to get involved. Maybe we need to have these city council meetings where the discussion of technology and surveillance and democracy is all at the forefront, because the way to engage citizens citizens in the fact that their relationship with government and policing is changing, and they should have a say in that. That. I would agree with that a 100%.
So let's now pivot a little bit to, I thought one of the more interesting insights in your book was how there is a potential positive here. A lot of us have concerns, you know, and, sir, we've talked on the broader positives for policing. But in terms of police accountability,
how can data help with police accountability and kind of, as you put it, turn the tables a little bit? Right. So there's a chapter in the book about blue data. And I I posit that what if you turn this architecture surveillance we've built, for citizens on the police. Right? The same body cameras that are pointing out to the world are also monitoring what police do. The same domain awareness system that's monitoring citizens is actually watching the police in the same camera angles. Right? Right. The same predictive analytics that can predict people who are involved or would be involved in crime can predict officers who might be involved in incidents of violence or or misconduct. And we can use those same technologies to improve police accountability and police training. And some of this has been done. There are a bunch of data scientists out of Chicago who went to Charlotte Mecklenburg's police department, and they're actually given wonderful access to, all the different variables. They weren't sure what they were gonna find. They were trying to look to see, could they find, inputs that would that created sort of situations of conflict, like, be it a a violent action or things that police might get disciplined for? And one of the things they found was that officers who responded to a traumatic scene, maybe it was, like, a child's death or a suicide, in the next shift had a much higher elevated likelihood of being involved in a violent crime or a violent incident. And the reason why being, of course, is they're human beings, and they just process this trauma that they saw, and they're not given great services to process overreact. It's it's very typical PTSD. A lot of officers have undiagnosed PTSD, post traumatic stress disorder. And the idea was, you know, maybe we don't send that guy to the next scene. Right? Right. They also learned with domestic violence. If they only sent two officers to a domestic violence call, it tended to sort of be an explosion of additional rage. But if they sent, like, six officers or more, there really wasn't a conflict because the usually, the man who was upset about whatever was going on didn't put up a fuss with that kind of show of force. So they reduced violence by sort of, like, data mining their own,
information. That is interesting. Alright. So I'm gonna I'm gonna let you go now, but this is a great book,
The Rise Rise of Big Data Policing. Any final thoughts you wanna share with our listeners? Yeah. I mean, so there's one last piece that I think people need to to pay attention. Like, big data policing is based on predictive analytics that, identify risk, be it places or people. But the solution doesn't actually have to be a policing remedy. You can actually divorce the value of the predictive analytics to identify a particular block that might have crime, or a particular person who might be involved in crime. But the answer doesn't mean, like, put a police car in that block or send a police officer to their door. It might be, like, fix up that block. Let's get the city services in there to do that, or let's give give the social services, educational services, the job services to that individual. It doesn't have to require a policing remedy. So we have, because of the way we fund it, the way the money has come through, the vendors and the rest, we've kind of connected predictive policing with policing, but maybe predictive policing is just as good without the policing part of it if you can take the predictive analytics and use it to solve the underlying social problems we know exist in the world, and that can be fixed and identified with these kinds of technologies.
I think that's a great thought to leave it on. Thank you so much. It's a wonderful book. Pick it up at your local bookstore or on Amazon. Thank you so much for joining us on CDT's Tech Talk, Andrew. Thank you so much. On November 29, the Supreme Court will hear arguments in Carpenter v US, one of the most important technology policy cases pending at the court this year. The justices are expected to decide whether the Fourth Amendment permits the compelled, warrantless disclosure of increasingly precise and revealing cell phone location information. Think about it. No warrant for information about everywhere you go with your cell phone. That's, well, probably everywhere. CDT filed a brief in the case and our always charming general counsel, Lisa Hayes, she has many other titles as well. She's here to talk about that brief and why this case matters. Welcome, Lisa. Thanks, Brian. It's good to be here. So first, can you give us a background on this case?
Sure. So Timothy Carpenter was identified as a suspect for organizing a series of armed robberies in the Detroit area, and the FBI got one hundred and twenty seven days of his historical cell phone location records. Yeah. That's about That's a lot of days. Four months of records without a warrant. So thanks to all of the call detail records, they were able to get his cell phone within a half mile to two miles of all of the robberies at the time that the robberies were occurring. And to make a very long story short, he was convicted and sentenced to a hundred and sixteen years in prison. And the question is whether or not if they had probable cause to believe that Timothy Carpenter had committed these robberies, they should have had to get a warrant before getting his cell phone information.
So that's the crux of what the Supreme Court then will decide, whether a warrant is required for that sort of information. Correct?
That is one of what I would argue are two cruxes. Alright. Give me the other crux. So so one argument is is before the court is whether or not you need a warrant. Yep. The other issue is a frankly antiquated legal theory called the third party doctrine and whether the third party doctrine is really relevant in today's digital age. And that's where a lot of CDT's brief focused our energy. So what the heck does that mean? Okay. So back in 1976, the court said it was okay to get bank records without a subpoena. And the idea was that you were entrusting this information to the bank, who clearly was a third party. You were not sticking the money under your bed. You were not burying it in your backyard. And so you were able to get access to those records without a warrant. And then it was extended to telephone records a couple years later. And the idea being if you entrust your information to a third party, it can't really be that private. The key is in the nineteen seventies, if you had personal information, odds are good you were storing that in your house. If you had photographs, what books and magazines you were reading, books and magazines you were reading, what your health records were. All that information would be contained offline Right. In paper format somewhere in your home, and the police would need to go get a warrant to get that information. What has happened over time is the third party doctrine has been stretched to encompass everything that's held by a third party. So this is a case where the court is going to look at the third party doctrine and see if it really should apply to things that you're trusting, say, to the cloud.
Interesting. So cell phone location data. How precise is it? Because that I mean, that's like a virtual in my world, it seems like a virtual tracking beacon if someone can just, like, identify, you said, within
half mile to few blocks? Oh. It can get quite a bit closer than that these days. So and the interesting part of it is that this information is automatically conveyed. Your cell phone is constantly pinging the towers. It is not similar to to the case back in the nineteen seventies where there was actually telephone operators who were helping to make all of these exchanges. But right now, you can get remarkably precise information about a person's wear being based on their cell phone through a variety of technologies. And so I hope that the court doesn't focus solely on the cell phone towers in this case because new technologies are coming all the time with micro cell sites, and and we have a lengthy explanation on this in in the brief. But let's build a rule and a precedent that will last no matter where the technology leads us in the years ahead because the technology we have today is pretty different from the technology we'll have in five years. But, yes, very precise.
So what exactly do you hope the court does then?
We hope that the court says that you must get a warrant to get information, either from cell phone records.
And we're hoping that it's an opportunity for the court to reconsider the third party doctrine in this case. Okay. That makes sense. Is there any you know, obviously, the case has not been argued. That happens on November 29. But are there any prior decisions you've seen from this court that might suggest, you know, how they might go? Predict a little?
It's important to remember that it's a new court. Last year was a little bit of a slow year quite candidly because the court was down one justice, after justice Scalia's death. Interestingly, the case that I think has a lot of precedential value, for the Carpenter decision is US v Jones. Ironically, justice Scalia was in the majority in that case and wrote the opinion, and that's a case where the court recognized that the new types of information that we contain in our cell phone are private, materials that in historic times would have been stored under your bed at home, but now you're keeping on the phone with you everywhere you go. Justice Scalia has been replaced by justice Gorsuch, and that will definitely have an impact in how the court breaks down on this case.
And I don't know, how that will play out yet. A lot to learn about the Supreme Court, definitely. And Lisa is, of course, a Supreme Court, nerd slash expert, however you wanna term it, but, aficionado perhaps. Fan girl. Fan girl. There you go. So regardless of the outcome, there's a lot of folks who are saying maybe Congress should act on this. Is that something that you and CDT think? I know that Senator Wyden has a bill out there. I think it's called the GPS
Act or something like that. Close? In fact, Congress has been invited to act by some of the appellate courts who've heard this case. And they've said, you know, if if you think you should need a warrant for this information instead of relying on the third party doctrine, excellent. Have Congress pass a law that says you need to have a warrant for this type of information. CDT thinks that's a very reasonable standard. We're certainly not pro crime. We are pro probable cause. Sure. Rather than having unlimited government surveillance, we think if you're suspected of a crime and there's information supporting that, you simply obtain a warrant using the probable cause standard and, and you're able to obtain the information.
That makes sense to me. So will you be one of the folks in line on the twenty ninth to hear the case live at the Supreme Court? Only if I find a babysitter. So watch this space. Well, thank you so much for joining Tech Talk, Lisa. Thank you, Brian. And for those who want to know more about this case, CDT is hosting an event that will go more in-depth on it on Tuesday, November 28. That's the day before, the Supreme Court will hear arguments and senator Ron Wyden will actually be our featured speaker there. If you're interested, check out the event details at cdt.org under events. That's it for this episode of Tech Talk. Again, if you'd like to hear more about US v Carpenter, check out CDT's event page at cdt.org for all the details on our event with senator Wyden. I'm Brian Wasilowski. Thanks so much for listening.
