Host John Markoff chats with Wendy K. Tam Cho, a 2019-20 CASBS fellow, Senior Research Scientist at the National Center for Supercomputing Applications, and professor of political science, statistics, mathematics, law, and Asian American studies at University of Illinois at Urbana-Champaign. Cho’s computer modeling produces and analyzes millions of finely tuned district maps. Cho and Markoff discuss the ways in which technology can reshape the process of political redistricting. Shout-out to Drina Adams, finance associate on the CASBS staff, for opening the episode for us!
Familiarize yourself with Optimization Problems
"How to Quantify (and Fight) Gerrymandering” - Quanta Magazine
“Toward a Talismanic Redistricting Tool: A Computational Method for Identifying Extreme Redistricting Plans” - Wendy K. Tam Cho and Yan Y. Liu
Wendy’s reading recommendation “Deep Thinking” by Gary Kasparov"
Announcer: From the Center for Advanced Study in the Behavioral Sciences at Stanford University, this is Human Centered.
Narrator: Our votes are one of our greatest sources of power in a democracy, and how we draw political districts determines just how powerful those votes can be. Today, host John Markoff sits down with Professor Wendy Tam Cho, whose computer models are capable of producing and analyzing millions of possible district maps. We'll get her thoughts on what we're capable of today and what processes we need to develop for tomorrow.
John Markoff: I wanted to start by asking about your path to your research interests now. So, you know, I see a variety of interests going way back, and I've counted at least 5 professional interests ranging from statistics to math to computer science to political science to Asian American studies. And so I wanted to ask what came first.
Wendy K. Tam Cho: Ah, that's a pretty hard question to answer. I guess I can start by saying nothing in my life has been deliberate. Literally nothing. I had no plan ever. And I think that's been actually why I've done what I've done. My parents aren't college educated, so they told me, you know, we want you to go to college because those people go to college, you know, they get money for nothing and we have to work really hard. And so they told me, you know, you go to college. But other than that, they never said, "Study this or do that." They literally thought, you know, you go to college and all this money is going to fall on you. So I went to college and I was already really happy because I thought I had already, you know, I was at the pinnacle. So I majored in math because that's what I loved. And I majored in political science because I thought it was interesting. But I had no plan, right? And so I did what I loved. And it just kind of came together.
John Markoff: Where did your interest in math come from?
Wendy K. Tam Cho: I just love math. I think it's beautiful. I still think it's the most beautiful of all the things that I, you know, ever have ever studied. So I've always loved it. I thought I was good at it. Actually, I'm not as good as I thought I was, which I learned when I got to college. But I thought I was good at it and I loved it. So that's— I don't know where it came from because, you know, my parents didn't really do math, as it were.
John Markoff: At Berkeley, as an undergraduate, were you in applied math? I mean, did you already discover computing at that point? The computer science department was part of the mix?
Wendy K. Tam Cho: No. See, again, I have no plan. I majored in math. I didn't know anything because I didn't come from a background where my parents were telling me stuff. I got there and I thought, "I'm going to major in math." I looked it up and they said you can be pure math or applied math. Actually, I didn't know I don't know the difference. And so I said, "Oh, I'll do applied." And then I looked down and I said, "Okay, so if you do applied math, you have to pick an applied field." And I'm like, "Okay, I never heard of that either, but let me see the list." And I looked down the list and nothing looked very interesting to me. And I saw computer science. And, you know, this was in the '80s. You know, it wasn't really a thing that everybody wanted to do computer science. And I actually didn't know what it was either. And so I thought, "Oh, that looks kind of interesting." You know, I didn't have a computer. I'd never coded in my life. But on the list, it was my pick. So I picked it, and that's how it happened.
John Markoff: Let me jump all the way forward to your gerrymandering work. And I wonder if you could sort of connect the dots and what took you into that particular interest.
Wendy K. Tam Cho: So, you know, nothing is deliberate in my life. So I'll try to connect the dots for you. So I majored in political science out of Duke. Just kind of a pure, I thought it was interesting. I actually wanted to be a politician when I was a little kid. And this was my idea that this would be such a great job and I could do all these things and help the world be a better place. You know, I'm 10, right? And I thought this would be this great thing. And then I got a lot closer to politics. I actually went to some conventions and political conventions and I met some politicians and then I thought, Wow, no, this is not for me. This is not what I thought it was, and this is not who I am. And then I kind of left that idea, but the interest, I think, in politics kind of kept on going. And so I majored in political science, and I think my senior year I wrote a senior honors thesis with actually somebody who's on the Stanford faculty now. He was at Berkeley then, Bruce Cain. But he was my advisor and we were supposed to meet every week and work on a project for the year. So we did that. And Bruce, turns out, had no interest in me. And so sometimes he wouldn't show up for our meetings or whatever. And then sometime in, I think, February that year, I walked into his office with a t-shirt that had a colon and parenthesis on it. It really large, like just really giant. And a friend and I had printed them. We just thought it was funny because, you know, we coded and we sent email with this funny computer smile and no one knew what it was back in the '80s. And so I walk into Bruce's office with this giant, you know, computer smile on my t-shirt. And Bruce says, "What is that?" And I said, "Oh, it's a computer smile." And he says, "What?" And I said, "Yeah, turn your head sideways and you can see it's a smile." The next time I saw him, he said, "Do you program computers?" And I said, "Yeah." He said, "Oh, I've got a project for you." And I said, "Okay. You know, I have a project. It's February and it's a year-long project. We've been meeting every week." And he said to me, "Well, that's a dumb project. This one's going to be a good one." And I thought, "Okay, well, you know, you're giving out the grades, so—" "Hey, what's your project?" And he described basically a redistricting project to me. He said, "I'm going to get you a bunch of data. You're going to write this code. You're going to see, you know, what would be the consequence if we enacted these laws." And I said, "OK." And he kind of laid out the whole project for me. And, you know, I wrote the code. And then, you know, come April, he says, "Oh, you should go to graduate school." And, you know, my parents—I told you they didn't go to college. I'd never heard of graduate school. I literally said to him, "What is that?" And, you know, he kind of described it to me, and I was thinking, "I've never—" I didn't know there was something after college. And so I was like, you know, so that all kind of set the ball rolling. But again, you know, this was not my plan.
John Markoff: The problem of sort of fairly drawing these lines and the kinds of algorithms that you might use to do that or the way you would approach it. And I was wondering, can you draw any parallels to other things in the world? I mean, is this a little bit like the traveling salesman problem, which I think people have a good sense of, or what problems is it like?
Wendy K. Tam Cho: Yeah, so it's exactly like all of those problems. So there are actually a lot of pieces in what we do. So initially, we had started with an optimization problem. So the traveling salesman is an optimization problem. So it uses the same types of algorithms, you know, just optimization heuristics. So we explored that for a long time and tried to figure out how would we draw a map by a computer in a very fast way, because we have to draw a lot of maps, right? So it wasn't just drawing a map. It was drawing a map that is a good map very, very quickly. So that's an optimization.
John Markoff: And you're fitting a definition of— it's fitting fairness? What are you fitting?
Wendy K. Tam Cho: So that piece is completely modular. Modular. You can throw in whatever you want as fairness. It doesn't have to be fairness. It could be, you know, draw me the most competitive plans that you can find, right? It could be draw me the most Republican plan that you can find. It could be the most Democratic plan. So that piece is modular. You just throw that in, and the computer itself doesn't care, right? You just tell it, what do you want, right? And it does what you want.
John Markoff: And is it a little bit like chess or Go that the machine will come up with answers the human wouldn't expect or a human wouldn't draw?
Wendy K. Tam Cho: Do you ever see those kinds of things? The human-drawn maps and computer-drawn maps are actually quite different.
Narrator: What is the biggest difference?
Wendy K. Tam Cho: It's really hard to kind of quantify, but if you— because a human has all sorts of things in its— in a human's mind that a machine does not, right? And it's the same thing with kind of chess and Go. If you think about how you would code a computer to play chess, you have to tell it very specific things and it does those specific things. But a human has these other ideas that are almost like meta ideas, like they're trying to implement a strategy and it's a very broad idea. I'm trying to control the center of the board, would be like a strategy. Things like that you have to think about and overcome. For a machine, For instance, for chess, I think if you try to teach it by showing it games from a grandmaster, one of the things that is kind of a funny thing is when you study chess that way, one of the things you notice is if you sacrifice a queen, that's a great thing to do. Because after you sacrifice a queen, apparently, you win the game right after that. And so a lot of the early chess machines was like, oh, I'm going to sacrifice the queen now. Wait, what happened? Why didn't I just win the game? So that— it's the same kind of idea.
John Markoff: Let me jump again to ask about context. You know, where are we now with more recent Supreme Court rulings and the ability to use tools like yours? We'll get into that in a second, but just sort of take us up to the highest level about where we are in the redistricting question.
Wendy K. Tam Cho: We don't know where we are because we're moving into redistricting 2021. And the Supreme Court ruled that, at least as a federal matter, that partisan gerrymandering is not justiciable, meaning they won't hear these cases, right? It's outside the purview of the court to hear it. For state cases, it's still open, so we can sue or we can bring a partisan gerrymandering case in a state court, but not in the federal courts. That has changed things dramatically, but we don't really actually know how at this point, you know, whether there will be a lot of state cases or whether there will not be. So there's that. And then there is now the districts in 2021. We don't know how they'll be drawn. Like, there are all these new independent redistricting commissions. We don't know what they're going to do. Every state has their own kind of commission. Their own kind of composition for the commission, their own process for the commission. So we don't really know what's going to happen, and everyone's kind of thinking about it.
John Markoff: And when you say "we," are you associated with a— what are you organizationally? How do you fit into the problem?
Wendy K. Tam Cho: I'm not associated with anyone. So I've written about it, how I think we would be best to proceed. But these are just my ideas. I'm not working with someone or employed by someone. Yeah.
John Markoff: And your tool, which is Parallel Evolutionary Algorithm for Redistricting, or PEAR, is it an open-source tool that actually is usable in these situations now?
Wendy K. Tam Cho: It is not open source. It is usable. I would have to use it. I've thought about open source, and I think that's a very thorny kind of issue, partly because I think, as I've written, the technology can be good and it can be bad. And I think if, you know, I had initially thought it would be open source, and I'm kind of rethinking that. It may still be at some point, but I don't know, because what I want to do is I don't want to create something that's used for bad. And if I just throw it out there, I don't know, I have lost complete control of its usage. So I'm still thinking about that. I actually have no idea.
Narrator: Is this because you would essentially be creating an arms race between people who are trying to hide partisan gerrymandering in more complicated districts?
Wendy K. Tam Cho: Yeah, so this tool could create even better gerrymanders than we saw the last decade, right? That now, if you know what the legal criteria are, you can always stay on one side of it.
John Markoff: Because you've basically instructed the computer, "Don't cross this line, but I want this outcome." Nate Persily was a fellow at CASBS a couple of years ago, and he's a political scientist at Stanford, and he served as a special master. Have you had contact with him, or have you traded notes? I mean, he uses these tools in the context.
Wendy K. Tam Cho: Has he used your tool? No, he hasn't used my tool. Nate and I went to graduate school together. We know each other. Yeah, we've talked about redistricting a number of times. And I'm not sure he knows the answer either, but we like to commiserate together.
John Markoff: I was just thinking about your notion that you've built a tool that basically you crunch a— you squish a bunch of human values on top of. And it then takes that advice or guidance and it outputs a map. What could you do to sort of take human bias out of the equation in redistricting?
Wendy K. Tam Cho: Yeah, so that's a pretty hard thing to answer. I think the idea is it's not the algorithm that would take out the bias, it's how it's used. So again, this idea that technology is neither good nor bad, and that what we need to do is formulate a process. That would work. And so I've written a little bit about that, but I'm still thinking about that. And I think my main idea is that the humans and the machines have to collaborate. The machines are really good at producing lots of maps so you can understand what could happen. And if you understand what could happen, then you can— that'll give you an idea of fairness. Like 12 Republican seats is way out, and this isn't really— fair, or at least you could think of it as not fair, or you can argue about whether it's fair or not. At the same time, having 12 Democratic seats would also not be fair. If you had all this information, then I think the humans can then deliberate about it because I think in the end it has to be human deliberation. It can't be tell the machine to do something and the machine comes up with something. The machine can do that, but I think the humans have to go back and say, okay, is that fair? What is our notion of fairness, and then bring in maybe other people who might have another notion, like you got to bring in the minorities. Maybe it's the Republicans and the Democrats talking and they've left out the minorities, or they've left out people who think we should actually just have competitive districts. There are all these different notions, and I think they have to be discussed by humans.
John Markoff: You outlined in a talk I saw you give these situations in where in states One of the parties will have 49% of the vote and 72% of the representatives. What about simply mandating that the algorithm must— those two numbers must match, and then working it from there backwards? Wouldn't that be a democratic and fair outcome?
Wendy K. Tam Cho: You mean to have proportional representation? Yes, yes. Well, there's a reason we don't have proportional representation. I mean, if we wanted that, I think you could mandate that, but that's not what we have. And one of the reasons we don't have that is because it's actually very bad for minorities. And so we've moved to districts largely for minority representation. And so, you know, you can go back to P.R. I think we actually just go back to P.R. People would say, oh, that's why we don't do that, right? And the Supreme Court has said over and over again, we, we don't have a system of proportional representation. So there are reasons for that. And I think You know, if you think about it and you look at the maps and you go through it, you'll see that that actually is not necessarily, you know, the top criterion.
Narrator: Do we have the ability to compute solutions for proportional representation while maintaining minority protections? Yeah.
Wendy K. Tam Cho: So you could, you know, use it on your computer. You could say, how close can I get to PR? And in addition to that, I want to make sure I have these minority districts in place. Then in addition to that, the point is a computer can sit there and take lots of interest and tell you, yes, you can do it or no, you can't. Because a lot of these are competing interests. I think it took the Democrats a long time to figure out that— well, some of the early lawsuits, I think, for racial gerrymandering, people noticed that the Republicans would side with the minorities and say, yes, more minority districts for you because that meant more Republican districts. Then it's like, "Oh, okay. I didn't know that those were competing interests." They figured out those were competing interests and then they decided, "Maybe we should think about this a little bit more. What would be more fair? Do we really want to give up Democratic districts for minority districts?" Sometimes minorities would say yes and sometimes they would say no because they weren't sure if having a non-minority district and just a Democratic one was fair to them. Even though maybe they were going to elect a Democrat, they wanted actually a minority legislature. So this sort of third component of—
Narrator: so you start with some values, plop them into the computer, it produces some results, and then you have this sort of review where people are saying, "OK, what do we think about this?" Are enough of the people involved in that process educated in the language of these models to kind of understand what's going on and trust them? How do you explain to a Supreme Court justice the veracity of these modeling procedures and things like that?
Wendy K. Tam Cho: Well, that's pretty hard, I think.
Narrator: And does it require them to understand it in order to trust it?
Wendy K. Tam Cho: I don't think it really requires them to understand it. A lot of the work is built on mathematical theory, and that mathematical theory has been written up in a mathematical way. And has been peer-reviewed by the people who do that kind of work, which is the way I think it has to be done, because the mathematical theory behind these models I don't think needs to or should be reviewed by a Supreme Court justice. But I think they should trust the scientific community in that regard. I think that's the way it has to proceed. It just, you know, the scientific community who does this kind of work needs to do the peer review of it.
John Markoff: You know, in your Nature essay, you were making the point that computers are impervious to the lure of power. And as I began to think about that, I also began to worry that they can equally be used as instruments of power. In fact, they frequently are. And I'm, you know, I'm not even quite sure of the question to ask, but it's clearly— I mean, we're sort of in this Larry Lessig code-is-law kind of world in which everything is being mediated by these algorithms. Everything in modern life, and it obviously can be used for good and ill. Are we right back where we were before, except surrounded by algorithms?
Wendy K. Tam Cho: No. So here's where I get back to my point that what's really important now is for us to think about process, right? It's not just the machine. It's how we use the machine, right? And that we have to set these processes into place where the technology then produces a good outcome versus just, you know, just having the algorithms out there and then like everybody can use them. That's not a good process, right? There's not enough regulation there around what is being done and how it's being used. And I think we have to think long and hard about that. And I'm not sure people have gotten to that point where we're thinking long and hard about that.
John Markoff: One of the big fights within the battle over voting machines and using electronic voting machines is the value of a paper trail. I began to wonder if you've thought about a paper trail for algorithms just generally.
Wendy K. Tam Cho: Well, there is a paper trail in that there's the code, right? It's there, and it's very specific, right? There's no ambiguity with what is it doing. You can see exactly what it's doing. It is true, it's a little bit hard to read and there's a lot of code, but then, you know, we have to go back to this idea of, you know, what has been written and peer-reviewed, right? You've written an algorithm and you say it does this, does it really do that? That has to be peer-reviewed, right? What is the outcome of doing it this way? Does it actually produce a sample? Does it not actually produce a sample? What are the characteristics of the maps that it produces? Can we actually, you know, have some theoretical basis to believe that? As opposed to, you wrote an algorithm and no one knows what it is and you're running it and nobody knows anything about it. That, I think, would be bad.
John Markoff: Have you used PEER in the real world yet? Has it been used for any political redistricting activity?
Wendy K. Tam Cho: It has been used in, or I have used it as part of my work as an expert witness.
John Markoff: Is it public where you've worked as an expert witness? I think so, because it's a legal case.
Wendy K. Tam Cho: Yeah. What is the relevant case? So the last case I worked was this case in Ohio, which was, I believe, APRi versus somebody, basically whether the congressional districts in the state of Ohio were a partisan gerrymander.
Narrator: And how did the ruling come out on that one?
Wendy K. Tam Cho: So it was a 3-judge panel, and they ruled that it was a partisan gerrymander, and then it was appealed to the Supreme Court, and then the Supreme Court said it's nonjusticiable, so then it was over.
John Markoff: Yeah. I was stunned. In one of your talks, you mentioned this, you know, the reality that 90% of the members of the House of Representatives are reelected as an example of of how uncontested American elections are as a rule. Yeah. That we're sort of working backwards from the way things are set up, I guess.
Wendy K. Tam Cho: Yeah. So that's not entirely redistricting, but it largely can be attributed to things like redistricting.
Narrator: How would redistricting change that?
Wendy K. Tam Cho: Well, one of the things redistricting can do is, you know, produce very safe seats. So like in Ohio, the last decade, the seats were very safe. So, you know, they would be won by very large margins, and they don't necessarily have to be drawn to be won by large margins. Sometimes the parties actually like that, because, you know, they just want to know how many are they going to get, and if they're safe, then that's how many they get, right? So you just kind of set that out at the beginning, and then you're done with it.
Narrator: Earlier you said we're kind of— we don't really have the process in place to do this stuff very well. And I'm wondering, what would it look like if we did? Would it look like more cases in courts? Would it look like centers of—
Wendy K. Tam Cho: I think ideally we avoid the courts, because that means somebody's unhappy, right? And very unhappy. So I don't know what we're going to do going into 2021, but my ideal process is that the technology would bring in more actors. That more people could participate in the process. So I don't know how that will happen. I actually have no idea, but one of the ways it could happen is if the politicians first decided maybe they were going to do something where they would say, "We're gonna use a computer, and, you know, these are the kind of inputs, and we're not even gonna input it. We're just gonna have it, you know, done, and everyone's gonna know the information, and then we will deliberate, and you can kind of hear us deliberate." and do that. I think that's unlikely because they control the process and they don't need to do this. And so, you know, there would have to be some reason to convince them, let's do this other thing, you know, that's more fair or whatever. But even if they don't want to do it, I think if the technology is out there and we have some process for doing it, then we could actually just know— not necessarily politicians, but everyone could kind of know this is what the general parameters are, maybe not the exact map, but the general parameters of what a bunch of people think should be considered can happen. In this state, we can have a map that has these kinds of characteristics all in one map. Then that would somehow constrain then the politicians. Because everybody knows, in the last decade, nobody knew, and then they would come back with a map and they'd say, That's the fairest map we could draw. We worked really hard and that's the best one we could get. They would just say that and you'd look at me like, really? It doesn't seem right, but I have no data. I have no ability to know whether or not that's right. Now we have more ability to know. Then they, I think, would feel constrained. I don't know if they will be constrained, but I would think they would feel more constrained in that information. System.
Narrator: Before we go, do you have a book or paper you've read recently, maybe related to this or not, just something you found captivating?
Wendy K. Tam Cho: Um, yeah, so one of my favorite books that I've read in the last year is this book called Deep Thinking by Garry Kasparov. I think the subtitle is Where Machine Intelligence Ends and Human Creativity Begins. And I really like that book because So it's about— it's his reflections on his game with Big Blue where he lost to a supercomputer. And it was like the first time that, you know, a grandmaster lost to a machine. So he reflects upon that game. So if you're at all into chess, it's a great read, you know, just from a chess perspective. But if you're not into chess, he talks a lot about just how he thought through the game and what AI is becoming, because a lot of what AI is becoming has started with chess. This was like the testbed of AI.
John Markoff: John McCarthy right here at Stanford did early chess-playing programs.
Wendy K. Tam Cho: Yeah. So what's really interesting is Kasparov, who I think is not a coder at all, when you read it, he just nails that AI stuff. It's as if you were talking to someone who writes The algorithms. And that was really fascinating to me because I started reading it mostly because my youngest son is really into chess. So I was like, oh, I'll read Kasparov's latest book. And I was like, this guy has so many great insights about where AI is going, where it has to go, not just for chess, but for society in general. So it's a great read just all around.
John Markoff: That's a great suggestion. Thank you for spending time with us. Yeah, thank you.
Narrator: Thanks again to Professor Wendy Tam Cho. To learn more about her work— and you really, really should learn about these maps, they determine how your vote is counted— check out the episode notes. We've put links to some great articles and papers on the topic. Human Centered is a show from the Center for Advanced Study in the Behavioral Sciences at Stanford University. Visit our website to learn more about the people and projects at the center by going to casbs.stanford.edu, or follow us on Twitter @CASBSStanford. Special thanks this episode to Drina Adams for the opening. From everyone at CASBS, thanks for listening.