 Good afternoon data nerds and welcome back to Stanford University. We have the pleasure today of live broadcasting all day on theCUBE at the Women in Data Science Worldwide Annual Event. My name's Savannah Peterson, delighted to be here with our fantastic lineup of guests. Starting off our final afternoon segment today with Kim, Kim is an absolute inspiration, just talking to you in the last four seconds before we were getting set up. I already feel inspired. You are one of the founders of the Stanford Human Trafficking Data Lab. Tell us about the lab, what does that mean? Well thank you so much for having me and thank you for your interest in the lab. We formed the lab four years, five years ago at this point, after a pretty inspirational talk here at Stanford in which we connected with a really inspirational partner who was really invested in bringing evidence-based interventions for human trafficking into Brazil's very robust anti-trafficking community. So the lab grew out of that partnership and we formed the lab with a couple of PIs here at Stanford and the Center for Human Rights and it's been a wild ride for the last five years developing a huge portfolio of projects that we all contribute to and it's been really exciting and such a rewarding experience as an academic to work in this kind of context. Yeah, absolutely and this collaboration and partnership spawned from meeting at an event, right? Absolutely. Two minds intersecting, a little bit of Kismet synergy and five years later you're crushing it. Kismet is just the right word for it, yep. Yeah, so you mentioned something in that intro of evidence-based data in application here and anti-trafficking. Slaves are a hard demographic to reach and get data and evidence on. That's right. Victims of human trafficking aren't exactly gonna pick up the phone when you do a survey. That's right. Where are you finding this evidence? Yeah, that's a really great question and I'll say that when I first started working on human trafficking, you know, we did not have any data whatsoever. We had like anecdotal data, we had case studies, we had sort of qualitative ethnographic kinds of work on human trafficking and it just really wasn't possible to do the kind of quantitative work that we're interested in in the lab and I started working on trafficking in like the late 90s, right? So it's been a really long time but I'll say that now in the modern data economy, we actually are invested in expanding what we think of as trafficking-related data, right? So yes, we do want to have those representative surveys. We want to find the hidden populations. We want to find those hard-to-reach populations that's critically important but we also can take advantage of the era that we're living in where we are like drinking from a fire hose when it comes to data streams. Yes, we are. And a lot of that data that isn't traditionally used to work on intractable issues like trafficking, I think if you reimagine what that work looks like, you can actually do quite a lot in this space. So thinking about, you know, these reamed archives of like, you know, labor code violations that aren't related to trafficking but our strong predictor of who's going to be a trafficker, right? Or satellite imagery or, you know, sort of administrative databases from social safety net programs and things like that. You can bring all of that to bear. And with the technologies that we have available to us now, the AIs and the LLMs, you can take, you know, volumes and volumes of legal reports and, you know, court records and things like that and then actually transform that into data that's useful for research, right? So it's an exciting time and we work with an expansive amount of data on the issue of trafficking when it comes to actually finding the hidden populations themselves. You know, and like identifying people who are in trafficking, that's really difficult. And it actually starts with- And dangerous, I would imagine. Yeah, well actually, I mean, trafficking happens everywhere, right? And so it's not happening in like, CD back alleys and, you know, places it is for sure, but it's also happening right under our noses. And so one of the important pieces of work that we're doing is really to under, to build an understanding that is evident, that is rooted in data, that is rooted in evidence about where exactly that trafficking is happening. So we just finished this huge household survey of representative sample survey down in Brazil where we just, you know, use the traditional tools of survey analysis to like get a population representative estimate of where is trafficking happening? What does trafficking look like in practice, right? What like, it may be very different from the types of cases that police are interested in or police prosecute, right? What does it really look like for people on the ground? And what are their experiences that are relevant? So we just completed that in December and we're working through all of that data now and it's really exciting and it's giving us a very interesting window into what everyday exploitation looks like and what trafficking looks like for people who experience it year on year, trafficked over and over again, and sort of, yeah. So that's not very uplifting. But the reality is, so I mean, not everything that comes out of our mouth needs to be inspiring and uplifting. What I do notice in the end, the data on the lab site is the estimates are between 27 million and 46 million people worldwide who are held in modern slavery. That's a lot of people, no matter how, I mean, that's a 10th of the population of the United States, if you were to look at that as a sample size. And it's not going to be uplifting when we talk about that and when we talk about combating that, but hopefully the technology today, actually this is a question for you, is our technology today and data science today making your job easier than it was in the 90s? Absolutely, there's no question about that. And it's making it easier to, well traffickers use this kind of technology as well, but it is making it easier for frontline stakeholders, law enforcement, civil society actors, all those kind of people who are doing the heavy lifting in the field every single day to try to help people. Their job is really hard. They have a lot of cases in their backlog. So we can bring the tools of data science actually to make their lives easier. We like to say that the tools that we build at the Trafficking Data Lab are basically like a fancy dishwasher that really just helps them to make their life easier to help them process through all of their cases in a more efficient way to target the cases in a better way and to really just enhance the very important work that frontline people are doing. And so I do think it is an exciting time because we have access to a lot of technologies and tools that just were not available to us before. And so it is very exciting time to be working in the space. I love that dishwasher analogy. We talk about data hygiene, but when you really think about that, it is accurate though. It is, yeah. And so let's dig in there a little bit more. In terms of the frontline, what sort of data decision making are you able to empower through this? Yeah, sure, absolutely. So I'll give you an example. And I just spoke about this inside the main conference. We have a project that uses satellite imagery to find illegal worksites in the arc of deforestation in Brazil. So part of, everybody knows that deforestation is happening. It's environmentally devastating. It's really terrible, but they don't know that there are, it is rife with human rights abuses, right? So people who are deforesting are not law abiding by nature and they don't care about people generally, right? They're converting this lot. Yeah, and so there's a ton of human trafficking in that process of converting all this forest and Savannah land into agricultural use. It's really hard for law enforcement to actually enforce the regulations and the laws because it's so remote, they just cannot find the sites and they don't know where to go. And they don't find it until it's after the fact. They don't find it until after the fact, right? These sites, by the time they identify where it's happening, it's moved on six months ago, right? And so, fortunately, the earth is imaged every single day, the entire earth, right? And so we can take advantage of this new source of data, the satellite imagery and these object detection algorithms that function really, really well to actually provide a service to our frontline partners where we can say, okay, you don't have to spend two months searching for this site that somebody thought might be in this area. We can actually just deploy our model in this area and tell you the coordinates and then you can go there, right? So that's an example of one type of application of data science that's providing a real value to- So much more efficient. It's incredibly efficient. In fact, we pilot tested it and the very first operation was completed in three days as opposed to the typical two, three weeks or something like that. We were talking about order of my suit. Yeah, and in resource constrained settings, time is my, everybody in the field is that takes their attention away from other cases. It takes, it costs money, right? And all those things are really constrained. So any way that we can help things work in a better way, help people do their jobs in a better way is what we want to invest in, yeah. Yeah, I mean, that's a perfect example. And it allows you to have, to be able to respond real time. Absolutely. By the time you show up in the past, everyone's gone. Yeah. Or whatever, in that particular use case. And in a proactive way, right? Yeah. I mean, you don't have to wait for somebody who's in a really bad situation to like get to a church and like ask for some help, right? Right. You can actually intervene in a proactive way and know and have a complete landscape of what's going on in the area. We talk about how to find hard to reach populations. Yeah. How do you find hard to reach like encampments, right? Yeah. It's really hard. So, but if you can use satellite imagery and these really well performing object detection algorithms to populate the entire state and identify where they all are, every single month that's a tremendous, that's a tremendous advance that we're Absolutely. So excited about. See, it is uplifting. We just had to get to the right part of the conversation. Don't worry about that. How does, how do, how do technological advancements like this in the narrative, the storytelling basically that you're able to do with the data like you're talking about? How does that affect policy making? Yeah, that's great. So we are fortunate that we have an incredible partnership and we're working in a context where policy makers are very invested in creating evidence-based policy and actually bringing data into their workflows, right? So, you know, us in academia and sometimes in tech, whatever you can like make most beautiful product, the most beautiful technology and people then don't use it if they're not invested in it, right? So we look for partners that are actually truly invested in a meaningful way and in this case in Brazil, you know, we formed the lab together and so there is that partnership baked in from the beginning and so there is that investment and then I think you demonstrate the usefulness of this and then that translates right away to better policy and in fact our partners start using it and then everybody else says, hey, what are they doing? Why do they have that tool? How can we get that tool, right? And so then you kind of start to expand in that way and so I will say that the work that we do in the lab is quite rare in that the academic research and the technological development that we're doing is designed for and has a direct pipeline right to that policy making from the beginning and we don't do anything unless we do formative work with our partners and know that what we're doing, the research that we're doing is needed and is helpful and will lead to some policy changes which is quite exciting to be involved in. Yeah, it is really exciting to be involved in especially having seen the course of time and the way that you have in what was possible I'm sure you were beating an emotional drum and now you have a very clear data-driven argument for a lot of the things that you're advocating for not that there shouldn't be that in human rights anyway but we've all been alive long enough to know it's not always the case. I'm curious, where do you see, where do you hope we will be in five to 10 years given the technological order of magnitude shift we're seeing right now? Yeah, that's such a difficult question because I think just the pace of technological development and the pace of how things are evolving, I think I am actually most excited for the questions I can't even begin to formulate right now, right? So I'm so excited that there are things on the horizon that are beyond my imagination at this point and so I want to be in a position to understand those technologies, understand those advancements and look for ways to induct them into our work and bring that to people who can really benefit from that. And so I guess in five years what I would love to see is an expansion of our lab, of our work and additional, more collaborations with so many of the incredibly talented women here at this conference. I mean, can you imagine bringing all this talent into your, I mean, it's an- Just the feeling in this room, I wish I could bottle it and share with the folks watching at home. It's so calm, there's no toxic masculinity here. It's very empowering. It's very empowering. Very uplifting. I'm not just on a cheesy, hey girl, you look cute in the bathroom way but in a, we are doing interesting things and how can we empower each other even more. Creativity and the energy is just unbelievable. Yeah, no, it's really inspiring. I'm curious, have you noticed or surfaced any trends or patterns in all of your research at the lab that were particularly surprising to you? Yeah, absolutely, we definitely have. I think I mentioned a few minutes ago that we just completed a large survey in the field and we're comparing the results of that survey to all of the various streams of administrative data that we have been using. So, you know, thinking about using like LLMs to process and digest and like feature extract from big reams of court records and prosecution records and whatnot. So that part, the administrative data gives us a really great sense of the corpus of cases that are being prosecuted, that are being, that law enforcement is interested in and that they're following up on. When you compare that to actually the results of the research that we did on the ground just by asking everyday workers a random sample of workers representative of the agricultural workforce, tell me about your experiences, tell me about how you were paid, tell me about who took your cell phone in the middle of the day or whatever, things like that. What we're getting is a very, actually a very different picture of everyday exploitation that people are enduring, that compares very differently to the cases that are the focus of law enforcement and prosecution. And so I think that that's been- You realize it's probably much more systemic than people. Huge, I mean we find, so we use a definition of trafficking that includes a couple dozen indicators, right? So it's really hard, you can't go to somebody and say, are you being trafficked, right? People don't know what trafficking is, you know? And so we use an indicator structure where we have to have a certain threshold in order to, for a person to sort of meet that definition that internationally accepted definition of a trafficked person, right? But what I found is even among the people that didn't meet that definition, that didn't sort of cross that threshold, just the experience of exploitation 95% of the people in our survey experienced that least one indicator of trafficking even if they didn't. I mean that's just incredible to me, just the level of what people are living with. Abuse, that's what you're saying, and the poor behavior that's out there, manipulative behavior, because I'm sure there's a lot of abortion. And I think there's a great policy opportunity there to address some of the exploitative practices that can then contribute to trafficking before it becomes trafficking, yeah. Yes, I love that. That's one of the great advantages, I think of AI and general ML is we may be able to stop things before they start. Yes. And stop the bad things. Absolutely, and being able to predict where a new hotspot is going to emerge, or if we know about seasonality and we have this long sort of time series of seasonal observations, where do we need to position our resources, right? What do we need to be prepared for in April versus October, something like that? And that's incredibly powerful. That is powerful. And it gives the upper hand to regulators who otherwise wouldn't have that advanced sort of predictive ability, right? So it's great, yeah. We've talked a bit about how you and the team and partners are using tech for good. You also mentioned that some of the nefarious actors here in traffickers are using tech for bad. How do we as a society ensure that AI is used in an ethical way? That's another really great question. It's a tricky one. It's a really tricky one. In fact, we just had a wonderful, you know, very high level panel discussion where the entire focus was on ethical AI and how do we make sure that AI is serving the human rights agenda that we want to have? So it actually goes beyond ethics to human rights, right? So it's not just this obligation to protect people and not do any more harm. It's actually an obligation to improve people's lives and help people. So how do we make sure that our AI and our data science research is actually working to improve people's lives, right? And that is on us, right? As researchers and it's not easy. There are no best practices out there yet, right? This is the Wild West. And so we need to do a very good job of being stewards, first of all, of the data and the tasks that are given to us. And then we always need to be interrogating our work. And in our lab, we have a working relationship and a close partnership with survivor advocates. And that's tremendously important because, you know, I haven't been trafficked. So I don't know. You know, it's not part of my experience. And so we bring in survivor advocates to gut check what we're doing and to sort of really give us an, like an independent perspective on whether we're doing enough, right? Are we really- Such a good pulse check there. Yeah, absolutely. Are we, we're auditing our results. Okay, but like, are we really seeing the bias that somebody else might see? You know, and that kind of thing. So it's really important work to do, but it's, there's, we need best practices. We need protocols. You know, we have plenty of clinical trial protocols for how like protect patients. We need those for this kind of work in AI when we're working with very vulnerable populations. And yeah, so it's really challenging, but very important. And it's something that we take very seriously in our lab. I can tell even just from talking to you. And like you said, it's imperative. Otherwise we just exacerbate the digital divide. That's right. That's right. This is our chance to close the chasm. We're absolutely alienated people forever. And I would say even more than closing that to digital divide, can we use like this AI and this technology to actually help leapfrog, right? Yes, yeah, yeah, yeah. We actually can use the technology that we have here to actually bypass some of the like bottlenecks and the sort of legacy things that interrupt our work here in context that haven't really implemented those yet. And so we can definitely achieve that if we have enough people with that type of vision. Yeah. No, that's very, very well said. I love the notion of how we could leapfrog that too. Yeah, absolutely. Everyone together at the same time too. It doesn't have to be this incremental linear stuff that we get stuck with a lot of the time. Three final questions for you. You were very inspiring and I feel like, especially when you're young, your heart's so open to things like trafficking as a cause to invest your time in, hopefully it stays that way throughout the course of your jaded adult life, but generally speaking, what would be your advice to a young woman or a woman of any age who's considering a career in data science or the world that we live in? Yeah, okay, well, that's, again, another great question. And I will say that it didn't take long for me to become jaded, if you want to become jaded real quick, start working on human trafficking. Right? I can only imagine, bless you for what you do seriously on the whole team. I think that I was given some very good advice when I was a young person, or like probably a freshman in college even, in that you need to follow what your purpose is and your passion, right? And then the human rights movement needs advocates in every single sphere, including data science, including AI, right? And so for me, I've always been a quantitative person. I always knew that I was a quantitative person, right? It wasn't compatible with doing human rights work at the time, but I followed that desire. I went through the graduate school, I positioned myself to sort of be in that position, and now when it's possible to work on the issues that I really care about, that I have always cared about, it's possible, right? So I would say that for young people, you know, you have to know thyself, right? And you have to follow that natural talent and that natural passion, and then look for opportunities to express that in your own personal way that is uniquely them, right? Yeah, absolutely. To that own self be true. Very well. That's right, exactly. Very well stated. And follow that passion. I mean, you don't have to know how your passion is going to make you money yet, as long as you keep your passion in front of mind. Absolutely. And going back to what we talked about before, you can't even imagine what's around the future. Exactly. So you just want to take opportunities and position yourself to take advantage of those opportunities you can't imagine yet. Exactly. All right, that's a beautiful answer. Final two questions. What is your advice for the allies in our world, looking to empower the women in data science around their team, or just even women in their lives curious about tech? Yeah, so I would say that community is so important. We're here at this conference building that community. I had mentors in my life that were really critical to me in my early development, and I love having mentor relationships with young people now. That is so important. And it's important to have women in community with other women, right? Yes. So if you want to have somebody who understands your perspective and experience, and is like a safe sounding board or a safe space for advice and that kind of thing. And so I would say building that community is critical around yourself, around others in your space. And looking for opportunities to kind of create that. Those mentor-mentee relationships is what I would recommend, yeah. Absolutely, I love that. Okay, last question for you. You're obviously very successful. Is there anyone you would like to give a shout out to today on International Women's Day to say thanks? Oh yeah, of course. I would like to say thank you to my partner in crime at the data lab, Jesse Brunner. She's the director of research at the Center for Human Rights here, and she has just been such an incredible inspiration, a friend, and just a very talented woman who is an inspiration to many people, including me, and does an absolutely herculean effort to run the research at her center. And so I'll shout her out. Shout out to my sister and my mother and all the other inspirational women, the mentors that I've had over the years. It's great, it's wonderful to be surrounded by inspirational women like that. And I feel that sitting here at the desk today. Thank you so much. It's been an absolute pleasure, and good luck with the rest of your work and your research, obviously very important. End up lifting as you have an incredible problem that you're trying to solve. Well, thank you for having me. Yes, our pleasure at any time here on theCUBE. And thank all of you for tuning in live to theCUBE's coverage here from Women in Data Science Worldwide Annual Event at Stanford. My name's Savannah Peterson. You're watching theCUBE, the leading source for enterprise tech.