Sarah Brayne is an Assistant Professor of Sociology at The University of Texas at Austin. In her research, Brayne uses qualitative and quantitative methods to examine the social consequences of data-intensive surveillance practices. Her forthcoming book, Predict and Surveil: Data, Discretion, and the Future of Policing, draws on years of ethnographic research of the Los Angeles Police Department to understand how law enforcement uses predictive analytics and new surveillance technologies. Prior to joining the faulty at UT-Austin, Brayne was a Postdoctoral Researcher at Microsoft Research. She received her Ph.D. in Sociology and Social Policy from Princeton University. Brayne has volunteer-taught college-credit sociology classes in prisons since 2012. In 2017, she founded the Texas Prison Education Initiative.
Across the country, protesters are taking to the streets to fight against police brutality and systemic racism. The use of force by police departments as well as the seeming militarization of many has been a concern of activists for some time. Another concern has been the use of big data in the use of surveillance technologies by departments to conduct predictive policing. Advocates for the approach say it helps police better marshal resources as the data is used to identify where hotspots of criminal activity might be. Opponents suggest the approach can just reproduce long-standing biases in the criminal justice system compounding systemic inequality. The intersection of big data and policing is the focus of this episode of Stats and Stories, with guest Sarah Brayne.
Timestamps: How did you get interested in this topic (1:31), Tell use about your study, (2:44,)How were you received in the LAPD (4:22), How did you get trained in this? (6:43), How did you get the data you’ve collected (8:01), Misuse of Data (10:24) Define Predictive Policing (12:05), Most surprising part of your research (16:40, What insights did you get from ride-a-longs (18:20), Differences between Canada (20:15), What’s next for Sarah Brayne (22:20), Texas Prison Education Initiative 26:45
+Full Transcript
Rosemary Pennington: Across the country, protesters are taking to the streets to fight against police brutality and systemic racism. The use of force by police departments as well as the seeming militarization of many has been a concern of activists for some time. Another concern has been the use of big data in the use of surveillance technologies by departments to conduct predictive policing. Advocates for the approach say it helps police better marshal resources as the data is used to identify where hotspots of criminal activity might be. Opponents suggest the approach can just reproduce long-standing biases in the criminal justice system compounding systemic inequality. The intersection of big data and policing is the focus of this episode of Stats and Stories, where we explore the statistics behind the statistics and the statistics behind the stories. I’m Rosemary Pennington. Stats and Stories is a production of Miami University’s Departments of Statistics and Media, Journalism and Film, as well as the American Statistical Association. Joining me are regular panelists John Bailer, Chair of Miami’s Statistics Department and Richard Campbell, former Chair of Media, Journalism and Film. Our guest today is Sarah Brayne. Brayne is an Assistant Professor of Sociology at the University of Texas at Austin who uses qualitative and quantitative methodologies to study the social consequences of data-intensive surveillance practices. She’s the author of a book out this fall, Predict and Surveil Data Discretion and the Future of Policing, which explores the issue of big data in policing. Sarah, thank you so much for being here.
Sarah Brayne: Thanks for having me.
Pennington: Could you tell us how you got interested in this particular topic?
Brayne: Sure, so I have kind of a long-standing interest in the criminal justice systems, specifically in the United States’ criminal justice system. I’m actually Canadian and so way back to when I was in college I was sort of interested in the similarities and differences between the countries, U.S. and Canada, and one of the really dramatic differences in the scope of the U.S. criminal justice system and associated surveillance. So I had an interest in the criminal justice system and was focusing on my work in incarceration, but I became increasingly interested in sort of the front end of the criminal justice system; how people came to be involved in the criminal justice system in the first place, which is through police contact. And sort of as my interest in policing was increasing, this was in about 2011-2012, there was all of this hype about big data that was starting to happen where, you know, everybody was like big data is transforming everything from sports to finance to education to marketing, and there wasn’t a lot of work being done on the intersection of big data and the criminal justice system, or policing specifically. And so, I decided to start to look into that for my research.
John Bailer: So, you did this giant study where you were embedded for two and a half years with the police department, could you tell us a little bit about what this kind of ethnographic study design entails? How did you identify L.A. as the host of this and how did this happen?
Brayne: Yeah, so I was interested in how the police use big data and the first stage of that was figuring out sort of what police department would be best to spend time with. So, the L.A.P.D. is by no means a representative department. It is a very large department; it is one of the departments that covers the most area, most people, have the most officers. And so I selected it specifically because I figured it might be kind of on the front lines on the police use of big data and my like forecast trends of how future departments might start to use data in the coming years. So, it definitely isn’t representative, but it is sort of on the cutting edge of police use of big data. And so, I set out basically to get access to – well at that point I was open to the L.A.P.D., Chicago P.D., or N.Y.P.D. after doing a bit of exploratory research. One of the very first things I found is that a lot of police departments that at the time said they were using advanced analytics and such really weren’t, so it was- I narrowed it down, perhaps unsurprisingly, to those three biggest departments, and then tried to get access to all three. And when I got a degree of access that I thought was good enough to really dive in and get an on the ground understanding of what was going on with L.A. I moved out there, back in 2013 and the rest is history basically.
Richard Campbell: So how are you treated? How are you received?
Brayne: You know, overall, very well. The way that I framed this work was initially, largely descriptive, saying the L.A.P.D. is one of the most technologically advanced police departments in the country, there are a lot of people, both civilian employees and sworn officers within the department that were very proud of what they were doing in terms of new cutting edge technologies and uses of data, and so I just asked, you know, can you show me what you do on the ground? And also, there was like this openness to sharing some of the challenges that they face as well. So, you know, crime analysts that felt underappreciated could kind of talk to me about that for example, or officers that felt that the public didn’t understand their job, they would want to talk to me about that kind of thing. Now, all that said, there is this blue of silence right? Like the police are notoriously difficult to gain access to whether you’re a researcher or a journalist, and constantly was sort of dealing with this balancing act of you know wanting to be very open and receptive, but also being able to say when I didn’t agree with something occurring. You know being able to be honest in my writing, I guess, is what I’m trying to say. And so, through the process, I have actually shared a lot of what I’ve written with a lot of the people that I did observations and interviews with, and a lot of the time they agree with what I write. Sometimes the don’t agree with that portrayal, in which case we have a back and forth, you know I say well, what do you not agree with and they will give their standpoint, and so- because it’s very important to me that what I write is factually accurate because I think a lot of the media portrayals of the police, you know the police are so- I mean you know this better than me, but they’re very resistant to talking to journalists a lot of the time because they feel like there are these hit pieces and whatnot. So, the ethnographic method really permitted me to have like this extended contact with people, not just one-off interactions, and sort of have a back and forth where it can be this iterative process of research.
Campbell: Could you talk a little bit about your training too and the combination of ethnographic and quantitative research? That’s so interesting.
Brayne: Yeah, so I am a sociologist by training. my Ph.D. is in sociology and social policy and so I’ve taken classes, you know, in everything from like survey methods to ethnographic methods and in this particular project- even though ironically it’s on big data, it is entirely qualitative, even though I’ve done quantitative work before because really like there isn’t just that good- there isn’t strong data on police usage of data. And so, this ethnographic approach was sort of the first step in mapping how different surveillance technology and sources of data are used. And so, taking an ethnographic approach to these questions of police use of data was really just what was required of the question at hand initially. And so, I think moving forward a more mixed methods approach could be great, but it really was because there wasn’t existing data on police use of data that I could leverage at that point.
Bailer: I love the fact that you describe this that there wasn’t data on the use of data. [Laughter] This is a real meta-level. So, can you talk a little bit about what does data look like that you collected? And given the data that you collected, what does an analysis look like from such data?
Brayne: Absolutely. So, there are two main sources of data basically that I collected in this project. The first is interview data, so I would conduct in-depth interviews with officer’s crime analysts that kind of thing when they would permit me to I would audio record them and then they would just be professionally transcribed. But in about half of the instances, the interviewees did not want me to audio record them, in which case I would basically frantically write interview notes. There are different ways of, you know, kind of writing out interview notes where you can get some direct quotes etcetera with your timing in the margins. And then the second really critical source of data is field notes; observational field notes, so for example when I would go on ride-along in cop cars I initially started taking notes by hand and then I realized why am I doing this? I spend all day on the computer, I’m way faster at typing, and also it’s kind of unnerving to people when you are taking physical notes about them when you are talking to them, and so I started using my phone. My phone was my main research tool because people were on their phones all of the time anyway, and so I would just use different functions on my phone and take notes that way and then clean them up at the end of the day or the next day. That was kind of the field notes, but basically just writing a ton of field notes. And then there are different approaches to analyzing that kind of largely textual data; I did take some photos as well, etcetera, but I didn’t analyze the photos in any systematic way, that was just sort of additional data source. But yeah there are different approaches and mine was basically an iterative coding- one way of talking about it is it was a grounded approach, basically, where I looked at the different kind of emergent themes coming out, and because it was such a long term research project I then- when I would do follow up interviews with the same people or interview a year later, I would focus more on certain themes rather than okay, let’s go through this huge run-through of the in-depth interview. So, I’d sort of refine the interview protocol, if you will, or what I would want to observe down the line based on the different findings that were starting to come out of the data.
Campbell: Could you talk about some of the misuse of data that you saw in your observations? I thought that was really interesting.
Brayne: Sure, you mean the police misuse of data?
Campbell: Yes.
Brayne: Well I mean there- so there also is a bunch of off the record stuff that I observed as well.
Campbell: Journalists know about that.
Brayne: Yeah, right right. The main issue here in terms of misuse is that surveillant technology is- the things that you can do with them so quickly outpace the rules and regulations and laws governing their use. So, there’s this massive gray area basically where it’s like, is this allowed? Can we even collect this data on this person? Can we search them in this database? Will anyone even know if we search them in this database without legal grounds etcetera? And I would say that actually, the majority of use of data in that sense occurs in that very gray space. And so that is where I saw so much activity and in that sense, you know, police are human beings trying to do their jobs. So if a detective, say, is trying to solve a homicide and there might be information about an individual that they think is involved and they don’t know whether or not it would be admissible in court, you know? It’s very tempting to get as much information as you possibly can. So yeah, I think that the existence of this massive gray area leads to a lot of questions about what’s acceptable and what’s not acceptable use.
Pennington: We’ve had people on the show before talking about issues related and one guest, in particular, wouldn’t land on a definition of predictive policing, so I’m wondering if you could talk a bit about how you think about or define this concept of predictive policing in your work?
Brayne: Yeah, so I think of predictive policing in very kind of spoken general terms and it’s simply the idea of using historical data in order to predict where and whom are at higher risk of being involved in future criminal activity. Then how those predictions are used or deployed by law enforcement, that can vary a lot and I don’t really necessarily think that that has to be involved in the definition, you know? It may deploy certain resources, you know? You send your cop cars to certain areas and not others or it may be, you know, go knock on these individuals’ doors, that kind of thing. But in its most general terms, I think that predictive policing is just the use of historical data to predict future risk of crime.
Bailer: I was intrigued when looking at one of your papers you had a nice figure that talked about the- how data has transformed policing.
Brayne: Yeah.
Bailer: And I thought that distinction of, you know, this issue of the integrated data versus disparate data being one of the most dramatic components. And also, the idea of a low inclusion threshold that’s just that some say in some sense what information is available and then how quickly you’re prompted to use it seems to be a big part of the story. Like the alert base and this transition from predictive to expletory or reacting to things seems- it just- I hadn’t really thought about how it changes practice so dramatically. Could you talk a little bit more about some of those ideas?
Brayne: Yeah well, that was really one of the benefits of, again, this like ethnographic approach, which is over the course of a few years in the different divisions within the L.A.P.D. adopted different predictive policing methods at different times. So, I was able to actually like do observations and interviews in places when they weren’t using the technology, and then they started, for example, or as they adopted a new form of predictive policing say. To actually be able to observe that kind of change over time because, I mean as statisticians and researchers you know that there is so much difficulty in knowing what actually you can causally attribute to an intervention say, or a policy change, versus a whole bunch of other forces. So all that is to say that yeah I think that the this idea of low inclusion thresholds, which is basically the idea that the police have always collected data on the individuals that they interact with and come into contact with but increasingly the police are gaining access to non-police data or data about individuals who don’t have to have any kind of police contact, automatic license plate readers being sort of the simplest example. Some automatic license plate readers are on cop cars or on cameras that are counted at intersections, but also a whole bunch of other folks have ALPRs, or automatic license plate readers, to like repossession agencies. And so, law enforcement, just like anybody else, can sometimes purchase or gain access to those kinds of data. And so, thinking about this like shift from query-based to alert based systems or exploratory or reactive policing as opposed to predictive policing. One of the things that this kind of mass surveillance can do is it can bring to law enforcement's attention people or incidents or things or places that they might have not noticed had they not been allocating their resources to them. However, I think that that would only fully transform policing in that way if we had near-perfect surveillance, which normally is a question that policy [inaudible] may not want because at the end of the day something like okay yeah something like an automatic license plate reader is a mass surveillance tool; they’re still deployed in certain neighborhoods more than others so not everybody has equal chance of being in a database. It’s a lot of the similar debates like with DNA databases, for example, you know conditional [inaudible] committing a crime, does everybody have equal chance of getting caught? What was really fascinating about this research is a lot of these questions that are typically framed in like a technical way, they’re actually very formative questions about what do we even think the police should be doing and how should we be allocating resources and such.
Pennington: You’re listening to Stats and Stories and today we’re talking with Sarah Brayne of the University of Texas Austin. Sarah, what was the most surprising thing that you as you were going back for your field notes and maybe sort of beginning to compile the book that sort of emerged for you as you were thinking through the work that you’ve been doing?
Brayne: Yeah, I think that the most surprising or unexpected thing which, and I mean this is the thing about findings is like in retrospect, of course, it makes perfect sense, but at the time I was surprised by it. It was actually on my very first ride-along so as I mentioned at the beginning of this episode I selected the L.A.P.D. because it’s so technologically advanced relative to a lot of other police departments. And on my very first ride-along you know we pulled up at a particular address, it was a house, kind of like an abandoned or an unoccupied house and the Sargent typed into his in-car computer you know that he was a particular code at the address and I was like is there not some automated way of knowing where the vehicles are? Like I picked the L.A.P.D. because I thought that so much stuff was automated and that it was really technologically advanced and he responded oh yeah, you know all the cars are equipped with AVLs, or these automatic vehicle locators that ping the location of the car every five seconds, but they’re not turned on because of the Police Officers’ Union. And so it was in that moment that I had this lightbulb of, you know, I need to be thinking of this situation from sort of a labor perspective and a work perspective you know? These officers are employees that resist managerial surveillance just like anybody else and that really shapes how policing plays out. And so, it was that very, like again, not really a technical story, but again like a very deeply sociological story of what was going on.
Bailer: What other things did you learn from the ride-along? I’m sort of picturing you in the helicopter with them or [inaudible] and what aspects of- what other insights did you glean from that?
Brayne: So, to be honest I don’t know that the helicopter remains in my research but like, I’m not going to say no to going on a ride-along in a helicopter, you know? Yeah well, the ride-along specifically I think this is particularly relevant to the conversations going on now around defunding or shrinking or abolishing the police is that the cops are called to so much stuff that I don’t think the cops need to be at. But if there’s a call for service they have to respond most of the time. And so, you know, so much of what we would respond to- now when you’re on a ride-along you’re never the first at a scene, like you have to have someone else- like I’m sure it’s liability. So, they don’t want some researcher getting killed or whatever, but they are responding to all of these things that probably would be way better dealt with by a much less sort of punitive or enforcing institution. And, in some ways, the cops’ hands are tied when they have to respond to something, you know, if someone calls them in. I think it speaks as much to how sort of enervated or emaciated our welfare state is in the United States. That folks in really desperate situations, you know I think I saw a lot of people on the worst day of their lives, basically. And I think that you know the cops often shouldn’t be there, and you know, if I think of the worst day of my life, I think in many cases the cops showing up might make it worse. And so, I think that was something that really struck me was a deep sort of sadness about when the cops are called sometimes; those people are in really desperate situations.
Campbell: That reminds me of I guess I was in [inaudible] I was listening to a researcher was pointing out that about 80% of all crime investigations involve misdemeanors. I think what you were alluding to there was about, you know, the cops don’t probably need to be at a lot of these places. Was there- and given your Canadian background, is there a- is that same kind of statistic true in Canada as well? And you mentioned some of the difference between Canada and the U.S. earlier…
Brayne: Yeah, that’s a good question. So, I haven’t actually like done comparative research on Canada, so this is entirely anecdotal- or as I heard someone recently call anecdotal stuff partisanal data.
Bailer: That’s pretty good.
[Laughter]
Brayne: Yeah so just thinking about my artisanal data.
Bailer: Just produced in small batches- I’m trying to picture this.
Brayne: Exactly mason jars, yeah exactly.
Campbell: That’s good.
Brayne: Yeah, so yeah, I mean I do think- I taught a class, for example, back at the University of British Columbia that was on urban poverty and we actually had guest speakers come in, including cops from the east side, which is where the majority of extreme poverty and drug use etcetera are. But crime rates are lower, way less people are armed and the interventions that are there are non punitive interventions. So, for example, there are safe injection sites, decriminalization of marijuana, these sorts of things, you know, which opens the door for more medical or social services professionals to come in. Now that said, I don’t want to say oh you know if we just reassign these quote-unquote social problems to medical or social service professionals everything will be fine. You know a lot of the time benevolent interventions can turn into this harder edge of social control as well. But yeah I do think that just the criminal justice system in Canada is so much smaller and the more I learn the more I think- much like with mass incarceration, the question is not necessarily how do we incarcerate better or how do we police better, but like how to we incarcerate less, or police less?
Bailer: So, what’s next for you? I saw that you may be looking at medical care and criminal justice systems, those two areas and their intersection, what might that look like? What are you exploring?
Brayne: Yeah, so actually in the early stages with this project with the L.A.P.D., I was sort of pursuing two different potential research projects because I don’t know if I would get the degree of access required with the LAPD that I thought I needed. And so, the other project was in an emergency room in Philadelphia basically looking at what are called custodial patients, like trauma vics that are brought into the emergency department in police custody. So, typically those would be folks with gunshot wounds or something like that and then looking into how information is shared across those institutional boundaries. So, like across medical and criminal justice institutions so for example there is some data that is reported mandatorily right? So, like when the assumption is that whenever there is a gunshot wound there is maybe some criminal activity involved so it’s mandatorily reported. But there is a lot again in this gray area that’s highly discretionary of what kind of information lawyers and cops are going to get on what happened. You know, there are these narrative parts of electronic medical records, for example, that would probably be very interesting to lawyers in a given case. So, I am interested in looking at how the same data can be used very differently across different institutional boundaries. So, doing some comparative work in the future and then I also have a smaller project following this stuff that I’ve observed in the policing context into the courts. So, into the future stages of criminal justice processing. So for example you know if the police have been using say there was a predictive policing algorithm and that’s why they were in that particular place or this individual had a high-risk score and that’s why they stopped this guy and then ended up arresting him down the line. Does that make it into the documentation that is, you know, part of the affidavit to get the arrest warrant? Or is it admitted as evidence? Is this in discovery? Or is all of that kind of erased? Like how following this big data policing further into the criminal justice system does it come up or is it invisible?
Campbell: Hey I’d like to- I think we’d all like to congratulate you on your book that’s coming out.
Brayne: Oh, thanks.
Campbell: And what I want to know, do you still get to change your introduction in light of all the stuff that’s happening?
Brayne: No, oh my gosh, I know, seriously, no like the page proofs are done.
Campbell: You’d think they’d want to do this just to promote the book better given how timely it is.
Brayne: Well, I will give you guys the presses information and you can tell them you’d like a second edition where I can write an updated intro, yeah. But I think though the thing is that is also important- this is such a historian's take, but like it’s important to recognize that even though all of this stuff is so urgent and so important right now, like this has happened before, you know? Police brutality, racist institutions, you know, these have been around for a really long time and actually the whole reason that the L.A.P.D. started using predictive analytics and data in the first place was because they got sued by the feds right? They were under this federal consent to [inaudible] whereby they need to comply with certain standards and then the exchange of the withdrawal of the criminal charge. And so, by sort of removing these all too human problems of bias, discretion, racism and instead doing data-driven decision making, that was how a lot of these technologies and analyses came to be in the first place. And so I think- in a sense I think it can kind of be a cautionary tale, right? Where now there are a lot of reformers suggesting data-driven policing as the solution to racist or biased policing, and of course while I think that data needs to play a role, it’s not the silver bullet, necessarily, that some folks may think it as.
Bailer: Yeah a bias that you have if it’s implemented in the model, it’s still biased.
Brayne: Yeah really it reproduces it, but at least it makes it look good.
Bailer: So I’m curious if your work and the study that you did was part of the inspiration for the Texas Prison Educational Initiative that you have been involved with, and if you can talk a little bit about the impact of that?
Brayne: Yeah absolutely, so the Texas Prison Education Initiative is just a group of volunteers at the University of Texas; it’s primarily graduate students and faculty members who volunteer to teach college credit classes in prisons in Texas. And so, it again sort of stems from my long-standing interest in the criminal justice system. I taught in prisons for a few years when I was in graduate school at Princeton, and when I moved to Texas I figured oh there’s probably some similar prison-ed programs here. God knows there are enough people in prison in Texas to have one, but there wasn’t, so I set that up and we started teaching in a juvenile facility and are also now in an adult facility. And it’s interesting, like while it very much is part and parcel of the same bundle of interest I have in how institutions in the United States are implicated in the reproduction of inequality, it is much more of a direct service kind of thing and I don’t currently have any research component to the program. I think, you know, own the line that may be something that we end up pursuing; there are so many more people that want to take our classes than we are able to enroll, and just because we sort of have limitations in terms of how many folks we can offer it to. So all that is to say that there is- we’re over-subscribed, so we could have some random assignment of different people and stuff but for me, it was really important in like starting the program to have it be just very simply direct service delivery. And you know education has really transformed my life, and therefore I just kind of wanted to afford that opportunity to folks that were having it denied to them, basically.
Bailer: Very good.
Pennington: Sarah, that’s all the time we have for this episode of Stats and Stories. Thank you so much for being here.
Brayne: Oh, thanks for talking to me.
Pennington: Stats and Stories is a partnership between Miami University’s departments of Statistics and Media, Journalism and Film, and the American Statistical Association. You can follow us on Twitter, Apple podcasts, or other places where you find podcasts. If you’d like to share your thoughts on the program send your email to statsandstories@miamioh.edu or check us out at statsandstories.net, and be sure to listen for future editions of Stats and Stories where we explore the statistics behind the stories behind the statistics.