>>Welcome everyone to the cubes coverage of women in data science, worldwide conference 2022. I'm Lisa Martin, coming to you live from Stanford university at the Arriaga alumni center. It's great to be back at widths in person, and I'm pleased to welcome fresh from the main stage Tiara Bill's assistant professor at UCLA Tierra. Welcome to the program. >>I'm glad to be here. Thank you for having me. Tell >>Me a little bit about your background. You're a civil engineer and I was telling you, so it was my dad. So I'm, I'm partial to civil engineers, but give our audience an overview of your background, what you studied and all that. Good. >>Yeah. So I'm a civil engineer, um, specifically transportation engineer, um, at UCLA. I also have an appointment in the public policy department. And so, um, I'm split between the two, my work focuses on travel demand modeling and how to use these tools to better inform, uh, and learn more about transportation equity and how to advance transportation equity. Um, and what that means is that we are prioritizing the needs of vulnerable communities, um, in terms of the data that we're using, the models that we're using to guide decision-making, um, in terms of the very projects that we evaluate and ultimately the decisions that we make to invest in certain transportation improvements. How >>Did you get interested in transportation equity? >>Yeah, so I think it, it stems from growing up, uh, in Detroit, some or Detroit born and raised native, and it stems from growing up in an environment where it was very clear that space matters that where you live the most, that you have access to, uh, whether you have a car or not. Um, whether you have flexibility in your, in your travel, it all matters. And it all governs the opportunities that you have access to. So it was very clear to me, um, when I would realize that certain certain kids didn't really leave their neighborhood, you know, they didn't travel about the city, let alone outside of the city and abroad. And so, um, and there are also other, you know, examples of, um, there are examples and cases after case where it's clear that communities are, um, being exposed to a high level of emissions, for example, um, that might result from transportation, but they're not positioned to benefit, um, in the same ways that the people who own the infrastructure on the freight or what have you. So, um, these are all very real experiences that have motivated my interest in transportation equity. >>Interesting. It's something I actually had never thought about, but you bring up a great point. How are talk to me about the travel demand models, how they're relevant and, and where some of the biases are in travel data, >>Right? So travel demand models, they are they're computational tools. They're empirically estimated meaning that their estimated from raw data, um, everything about them is driven by the data that you have access to. And how they're used is in largely in regional transportation planning, when it is necessary for regions to assess 10, maybe 15, 20 years into the future. Um, how is transportation going to change as a result of changes in travel patterns, growth in the population, um, changes and how firms are distributed across the landscape. Um, environmental changes, all sorts of changes that, um, that guide and direct our transportation decisions at an individual level. So regions are assessing these things over time and they need these powerful travel demand models in order to perform those assessments. And then they also, once they have an understanding of what the need is, because for example, they expect traffic congestion to improve, or sorry to increase over time. Um, there needs to be a means of assessing alternatives for mitigating those issues. And so they use the same types of models to understand if we expand highway capacity, if we, uh, build a new form of transit, is that going to mitigate, uh, the challenges that we're going to face in the future >>And travel demand, modeling and equity? What's the connection there? I imagine there's a pretty good >>Deep connection, right? So the connection is that. So we're using these tools to decide on the future of transportation investments and because of a history of understanding that we have around how ignoring the conditions for vulnerable communities, ignoring how, um, uh, transportation decisions might differentially impact different, different groups, different segments. Um, if we ignore that, then it can lead to devastating outcomes. And so I'm citing, um, examples of the construction of the Eisenhower interstate system back in the fifties and sixties, where, uh, we know today that there were millions of black and minority communities that were, uh, displace. Um, they weren't fairly compensated all because of lack of consideration for, for outcomes to these communities and the planning process. And so we are aware that these kinds of things can happen. Um, and because of that, we now have federal regulations that require, uh, equity analysis to occur for any project that's going to leverage federal funding. And so it's, it's tied to our understanding of what can happen when we don't focus on equity is also tied to what the current regulations are, but challenge is that we need better guidance on how to do this, how to perform the equity analysis. What types of improvements are actually going to move the needle and advance us toward a state where we can prioritize the needs of the vulnerable travelers and residents? What >>Excites you about the work that you're doing? >>You know, I, I have a vested interest in seeing conditions improve for, um, for the underdog, if you will, for folks who, um, they, they work hard, but they still struggle, um, for folks who experience discrimination in different forms. Um, and so I have a vested interest in seeing conditions improve for them. And so I'm really excited about, uh, the time that we're in, I'm excited that equity is now at the height of many discussions, um, because it's opening up resources, right? To have, uh, more folks paying attention, more folks, researching more folks, developing methods and processes that will actually help to advance equity, >>Advancing equity. We definitely need that. And you're right. There's, there's good V visibility on it right now. And let's take advantage of that for the good things that can come out of it. Talk to me a little bit about what you talked about in your talk earlier today here at widths. >>Right? So today I got a chance to elaborate on how travel demand models can end up, um, uh, with, with issues of bias and under-representation, and it's tied to a number of things, but one of them is the data that reusing, because these are, uh, empirically estimated tools. They take their form, they take their, uh, significance. Everything about them is shaped by the data that we use. Um, and at the same time, we are aware that vulnerable communities are more prone to issues that contribute to data bias. And under-representation so issues, for example, like non-response, um, issues like coverage bias with means that, um, certain groups are for whatever reason, not in the sample link frame. Um, and so, because we know that these types of errors are more prevalent for vulnerable communities, it brings, uh, it raises questions about, um, the quality of the decisions that come out of these models that we estimate based on these data. >>And so I'm interested in weaving these parts together. Um, and part of it has to do with understanding the conditions that, um, that underlie the data. So what do I mean by conditions? I gave an example of, uh, cases where there is discrimination and as evidenced by the data that we have available as evidenced, uh, for example, by examining, um, the quality of service across racial groups, um, using Uber and Lyft, right? So we have information that, that, that presents this to us, but that information is still outside of what we typically use to estimate travel, demand models. That information is not being used to understand the context under which people are making decisions. It's not being used to better understand the constraints that people are facing when they're making, uh, decisions. And so what is the connection that means that we are using data, um, that does not will capture the target group. >>People who are low income, elderly, um, transit dependent, uh, we're not capturing these groups very well because of the prevalence of, of various types of survey bias. Um, and it is shaping our models in unknown ways. And so my group is really trying to make that connection between, okay, how do we collect Bader, better data, first of all, but second, what does that mean? What are the ramifications for prediction, accuracy for VR, for various groups, and then beyond that, what are the policy implications? Right. Um, I think that the risk is that we might be making wrong decisions, right? We might be assuming that, uh, certain types of improvements are actually going to improve quality of service for vulnerable communities when they actually don't. Right. Um, and so that's the worry and that's part of the unknown, and that's why I'm working in this >>Part of the anonymity. Also, I'm sure part of your passion and your interest international women's day is tomorrow. And the theme this year is break the bias of breaking the bias with >>Mercy back >>To travel equity. Where do you think we are on, on being able to start mitigating some of the biases that you've talked about? >>I think that it's all about phasing. I think that there are things that we can do now, right? And so, um, at the point of making decisions, um, we can view the results that we have through this lens, that it might be an incomplete picture. We can view it through a historical lens. We can also view it, um, using emerging data that allows for us to explore some of these constraints that, you know, might be exogenous to the models or X, not in, not included in how we estimate the models. Um, and so that's one thing that we can do in practice is okay. We already know that there are some challenges let's view this from a different lens, as opposed to assuming that it's giving us the complete picture. Right. Um, and that's kind of been my theme, uh, today is that, you know, as decision-makers, as analysts, as data scientists, as researchers, we do have tendency of assuming that the data that we have, the results that we have is giving us the complete picture when we know, but it's not, we know that we act as if it is, but we know that it's not right. >>So, you know, we need to, there's a lot of learning and changing of behaviors, um, that that has to happen. >>Changing behaviors is challenging. >>It is behavior changes is tough, but it's necessary, but it's necessary. It's necessary. And it's urgent. And it's critical, especially if you're going to, uh, improve conditions for vulnerable community. >>What are some of the things that excite you, that looking at where we are now, we've got a nice visibility on equity. There, there's the conscious understanding of the bias and data and the work to help to mitigate that. What are some of the things that excite you about what you're doing and maybe even some of the policies that you think should be enacted as a result of more encompassing datasets? >>It's a good question. Um, one thing I will say is what excites me is it's also tied to the emerging data that we have available. So I'm trying to go back to an example that I gave about measuring constraints. Think that we can now do that in interesting ways, because we're collecting data about everything we're collecting data about, um, not just about where we travel, but how we travel, why we travel. Um, you know, we, we collect information on who we're traveling with, you know, so there's a lot more information that we can make use of, um, in particular to understand constraints. So it's, it's really exciting to me. And when I say that again, um, talking about, um, how would we make a choice to take a certain mode of transportation or to leave our house at a certain time in the morning to, to get to work. >>Um, we're making that under some conditions, right? Right. And those conditions aren't always observed and traditional data sets. I think now we're at a time where emerging data sources can start to capture some of that. And so we can ask questions that we weren't able to, or answer questions that we weren't able to answer before. And the reason why it's important in the modeling is because in the models, you have this sort of choice driven side and you have the alternatives. So you're making a choice amongst some set of alternatives. We model the choices and we spend a lot of time and pay a lot of attention to the decision process. And what factors goes into making the choice, assuming that everyone really has the same set of universal choices. Right. I think that we need to take a little, pay a little more intention, um, to understanding the constraints that people have, um, and how that guides the overall outcomes. Right? So, so that's what I'm excited about. I mean, it's basically leveraging the new data in new ways that we weren't able to before >>Leveraging the data in new ways. Love it. Tierra, thank you for joining me, talking about transportation equity, what you're doing there, the opportunities and kind of where we are on that road. If you will. Thank you so much for having me, my pleasure. I'm Lisa Martin. You're watching the cubes coverage of women in data science conference, 2022. We'll be right back with our next guest.