This article is part of STEMinism in the Spotlight, a monthly interview series.

Gireeja Ranade is an assistant teaching professor in Electrical Engineering and Computer Sciences (EECS). She received her PhD in EECS from UC Berkeley as well. She is an accomplished and thoughtful instructor: she designed and taught the first offering of EECS 16A/B and received the 2017 UC Berkeley Electrical Engineering Award for Outstanding Teaching. Prior to becoming a professor at UC Berkeley, she was a Researcher at Microsoft Research (MSR) AI. Her research focuses on a number of areas including artificial intelligence (AI), information theory, and control theory.

Amanda Glazer (AG): Could you give us an overview of your research?

Gireeja Ranade (GR): I am broadly interested in lots of different things. My research has evolved over the years, and I have worked on a very wide variety of things.

I started out as an undergrad working in biomedical engineering. I was seriously considering going to medical school. My undergraduate thesis involved thinking about developing non-invasive methods for monitoring a quantity called cardiac output, which is the amount of blood that your heart pumps out per unit time. Typically, this is measured by having an incision in your chest, putting in a catheter, and measuring how much blood is being pumped out. This is, of course, not a procedure that you want to have done, so what I was working on was whether you can develop models measuring things that are easier to measure such as your heart rate, respiration rate, pulse, blood pressure, for your cardiovascular system. We ended up building on this model called the Windkessel Model which uses electronic circuit elements, like resistors and capacitors, surprisingly enough, to model the cardiovascular system. I really was planning to move forward in biomedical research. My master’s thesis was about neural signal processing and brain machine interfaces. It was fun work, but then I took information theory and control theory classes and I realized that was something I was even more interested in.

AG: What about those fields were you drawn to?

GR: For information theory I think it was just the elegance of the field. It’s so pretty. Everything fits together. It’s very elegant. I really like that about it. I also thought that this intersection between control and information theory is fundamentally very interesting. In information theory you have a sender Alice and a receiver Bob and you care about how many bits Alice can send. You don’t care about what happens to those bits once Bob receives them. Bob can put them in the bank, make a painting out of them, or do whatever he wants with them, and that’s that. But when you think about information theory or communication in the context of control, what starts to matter is, what are you going to do with that information? If you’re a controller and I’m sending you information and then you need to take actions based on that information, suddenly why the information is being sent becomes a central part of the problem. I thought that was very, very interesting. Largely I’ve been interested in problems around uncertainty in control.

A reason to be interested in this is because thinking about communication and control is a very simplified way of understanding interactions in large scale organizations and systems with people in them. If you cannot understand multiple little controllers that are communicating with each other, maybe some have one objective and others have another objective and you are trying to understand what kind of communications are happening between them and how they can collectively achieve their objective, how are you ever going to understand larger questions in society where you have multiple people interacting and multiple organizations with multiple objectives? These problems of control and communication are very cool as a very simple way of thinking about interactions.

One of the very first problems that I worked on in my PhD was this example called Witsenhausen’s counterexample. I got interested in it because a friend of mine was working on the problem. This problem is basically a problem about two controllers. One controller is really strong and can put a lot of power into the system and change the system, but this controller has glasses, or blurry vision, so it has only noisy observations of the system. Whereas you have another controller that has very, very precise observations on the system, but cannot put as much power in. So, how can these collaborate to achieve the best control performance? This is the basic setup of Witsenhausen’s problem. One of the senior students had just made some really nice contributions in this area and told me about it, and I worked on a little extension of this, which is what got me interested in the field. That’s my story about getting interested in information theory and control, but I’ve been broadly interested in a lot of different fields.

I did a lot of internships when I was a graduate student. I interned, for example, at IBM with Lav Varshney. There I worked on understanding crowdsourcing. Mechanical Turk was just starting back then, and these crowdsourcing contests were just starting to pick up—topcoder was new, for example. We were trying to understand this very fundamental question of when should one crowdsource. To crowdsource or not to crowdsource? I think of this as an interesting problem of how to organize different entities, different people, to come together? It is very different from control but has a similar type of flavor.

I’ve worked in wireless communications, more applied questions around how to think about 5G networks. That was work I did at Berkeley as a PhD student. When I went to Microsoft Research that was a whole new world. There were all these new problems to work on and new perspectives. I did a lot of work in control and that’s where I got interested in some of this work on misinformation and fake news.

I would love to hear more about your work on fake news.

This is a project that started just out of casual conversations with friends and colleagues. We were reading a lot of media reports about whether “fake news” influenced the election. We didn’t even really know how to define what fake news was. It was just a term that started coming up in the media. I try to avoid using it now as much. We thought we should try to put a scientific perspective on it. A lot of the stories that we would read were anecdotal and we thought, as engineers and scientists, let’s see if we can put numbers to it. I was really lucky to be at MSR at the time and have a very supportive group of people who were really excited about trying to understand what was going on. The nice thing about being at MSR was that we were able to actually have access to data to study this. We were able to really look at what web traffic was like to some of these websites that were being talked about in the news as spreading misinformation. How many people did visit them? How often did they visit them? Where did people come to these websites from? The media talked a lot about social media playing a big role. One of the most surprising things for me from that research was the fact that it turns out when you look at the data there were actually many misinformation websites, which were sharing misinformation and fake news, and that got a lot of traffic, where people did not arrive at those websites from social media.

How were people arriving at those fake news websites?

They were coming from other sources, like email or blogs. Some cases we don’t know where they were coming from, but it was not social media. There were significant websites, such as infowars.com, that got lots of traffic that was not necessarily coming from social media. That was something that I thought was very interesting.

How did you characterize what misinformation or fake news is and how did you define a fake news website?

We basically punted on the question entirely. We are computer scientists and engineers, and we are not trained in journalism or identifying misinformation, and so we relied on third party lists. Now there is a lot more systematization to fact checking and people are more aware of it. At the time, this was December 2016/January 2017, there was a list put out by Snopes of particular stories that they knew were fake. Buzzfeed had put out a list of stories that they knew were fake. Wikipedia had a list. We used the Wikipedia list, largely, of websites that were known to share misinformation. We did not go through and curate the list. One large reason being we didn’t want our personal opinions to influence the curation of a list. We wanted to be as scientific as we could, so by using classifications from someone else and just studying traffic to that classification we were able to try and be neutral on the matter.

This was research that you did at Microsoft Research?

This is research I did at MSR. I’m still collaborating with some of the folks there after moving to Berkeley, but a large volume of that was conducted while I was at MSR.

How did being at MSR differ from being here? Generally, how does industry and academia differ? It seems like there is a lot of similar research.

There’s a lot of differences at one level and almost no differences at another level. The big difference is that the pulls on my time here are very different. I spend a lot of time with students and teaching. The sheer number of responsibilities that I have is larger. For example, I am teaching a class with 1,100 students this semester. So, I get a lot of email! The course staff that I’m working with is about 100 people. At Microsoft, I wasn’t managing anybody. It’s a very different job description from that perspective. Research-wise, there’s not that much difference. At MSR you have the really great ability to work with certain types of data that is not available to academics. The fact that MSR makes that available to the researchers there is really cool. That was very nice. Beyond that, there were not a lot of big differences. At MSR work with interns is focused during the summer months, whereas at Berkeley this happens all the time.

Did you know you wanted to come back to academia when you went to MSR?

I don’t know that there was necessarily a plan. When I was graduating, I got this amazing offer from Microsoft Research that seemed like the best postdoc offer that I could have ever gotten. I was thrilled to go there. If you had asked me at the beginning of my PhD if I was thinking of being at a place like Microsoft Research it would never have even crossed my mind.

Why?

I don’t even think I knew about it. I didn’t know very much about what industry research even meant. It wasn’t like I had a plan to go there, it just sort of happened. The same thing happened with the move back to academia. I’ve always been interested in teaching and academia, but I wasn’t necessarily sure that I was going to make the jump. It was something that seemed fun and that I wanted to try.

What do you like about teaching? What is your approach to teaching?

One of the exciting things about being at Berkeley in particular is the opportunity to impact a very large number of students and do it at a very early time during their careers. At a point where you maybe have some ability to give them a good experience and have them like a particular subject. I know there are a lot of stereotypes and preconceived notions about computer science and engineering. I know that students come in with all sorts of preconceived notions about what the field is. The reality of what I do now and what I thought engineering was when I was a freshman is different.

The class that I teach in the fall semester is a large, introductory freshman class. Our target population is freshman and incoming junior transfers. It’s the first class they take in EECS. There’s two first classes: 61A which I think is the largest course on the Berkeley campus—this semester I think they have 2,000 students—and 16A which is the one I teach. A lot of our students are concurrently taking 61A and 16A. This is my chance to expose them to some of the really cool mathematical parts of the department, that I think are very exciting. I find that quite nice.

When I started as a freshman in college, I knew that I really liked math, but I didn’t really have that much exposure to programming, whereas a lot of people around me had a lot more exposure to programming.

EECS is such a vast field and I want to try and show people the different parts of the field early on in their undergraduate careers. Sometimes people come in with preconceived notions of what EECS is and who should be in it, but I want to try to show them that the field and the people in it are actually a bunch of diverse personalities with different intellectual interests and are coming from different backgrounds. I keep telling my students: you don’t have to fit into a box that someone else made for you, you can make your own box, and you don’t have to fit into that either.

I like the challenge of trying to reach many different personalities and trying to teach a class where you want to make sure there is something interesting for every single student. Some students are coming from having tons of exposure, from coding, robotics, electronics, they’ve had that exposure in their life so far. But there’s other students that have never had access to AP Computer Science or AP Calculus classes. There are many first-generation students coming to Berkeley. The life experiences of these students are so vastly different. There are things that work well for one group that don’t necessarily work well for another group, so trying to make a course that can offer something to every group is what I want to do.

One of the things we do in the class is have a lot of different types of homework. The class is called “Designing Information Devices and Systems.” We basically provide a basic introduction to machine learning and design. The first module in the class is essentially an introduction to systems and modeling. How do you take something that’s real world, unformulated, and turn it into a mathematical model that then you can collect data about or reason about or simplify? For this, the first module talks about linear algebra and tries to build up from first principles. Walking in, the students are not expected to know what a vector or matrix is. By the end of the course, we hope that they will learn how the least squares algorithm works and how matching pursuit works. That’s the flow of the course. For each of these abstract linear algebra concepts, for example an eigenvalue, we try to create homework problems that are very applied and also have lab modules. For example, in the first module the students do an imaging lab where they are basically building a single pixel camera. You have a picture, you shine light on it from a projector and then you measure the reflected light on the sensor. Let’s say I only have one sensor: how would I try to image this? Well, I would project really, really small dots of light. Let’s say my image is just a checkerboard of black and white pixels. Every entry is a 0 or 1. I can measure pixel by pixel. An alternative might be to simultaneously measure multiple pixels but then take multiple readings. Basically, build a system of linear equations of the pixels. Then what we talk about is how do you build the system of linear equations and can you build better systems of linear equations than others? If you set this up as a system of linear equations, the quality of the reconstructed image is related to the eigenvalues of the masking matrix. If you have larger eigenvalues, you are going to get a better reconstruction.

When I learned engineering, signal processing and linear algebra, I was told here’s a matrix, algorithm and do these steps because there’s a problem on the homework. For someone who is already committed to studying CS, engineering or math, then you have a motivation to grind through this problem. But if you aren’t sure whether you want to study CS or engineering at all, if you are like, well I could study this or I could study physics or economics or history or biology, if you have all these other options, it’s very hard to get motivated about an integral, unless you have some other external force that’s motivating you to learn about that integral. What we are trying to do is provide a motivation for the mathematics. To explain why this mathematics matters. By showing that, we are trying to draw in people who might not have previously seen themselves as engineers or computer scientists.

What is it that drew you into EECS when you were in college?

I spent a lot of time trying to figure out what I wanted to major in. I almost majored in physics. I took a lot of biology classes as well. I ended up studying EECS because one of the things people told me and I also realized was that EECS was the broadest major that I could do. It left the most doors open. If I studied engineering, I could still apply to medical school afterwards, or do something more physics like if I took physics courses too, or do something more policy related. I couldn’t decide among the majors and this was the happy medium, where I get to study some math, physics, etc. It was this averaging decision.

Did you like it a lot and then decide you wanted to go to grad school?

No, I wish it was that clean of a story. I just got lucky. I definitely did not know what I wanted to do. When I was graduating I applied to all kinds of jobs. I applied to software engineering, consulting and teaching jobs. One of my professors, actually—well there’s actually three of them that I owe a lot to: George Verghese, Denny Freeman and Colin Stultz. In particular, George Verghese, I was doing research in his lab, and he said, “Have you thought about grad school? You should really consider applying. Why don’t you try this thing out?” One thing led to another and here I am now. It really mattered to me that I had a mentor that cared. It made a big difference. The fact that I enjoyed working in his lab and had a positive experience and that I enjoyed it more than just doing homework for a class. I’m not a person who enjoys taking exams, but I really enjoyed the open-ended part of this and the fact that there wasn’t necessarily a deadline and I was just expected to try. That meant a lot to me as a student.

I feel like that’s a common experience.

Yeah, I don’t know if your experience is the same but for me it mattered a lot. The fact that people said, “I think you can do it.” Just that little nudge made a big difference.

I wanted to ask about your experience teaching with the Meltwater Entrepreneurial School of Technology (MEST) and the Prison University Project (PUP). How did you get involved with those organizations?

PUP was something I did as a grad student here. I don’t know if you know but there’s a prison, San Quentin. There’s this amazing group, it’s run by Jody Lewen, and a large number of the volunteers are actually Berkeley graduate students and staff.

Prisons are not allowed to receive funding for inmates’ higher education. They are not allowed to receive federal or state funding for any education for inmates beyond their GED. So what PUP does is it offers inmates the opportunity to get an associate’s degree, but it is entirely run based on donations and all of the teaching is done by volunteers. It’s run by an amazing and dedicated team. Berkeley is a large recruitment center for their volunteers. The first time you teach you do the job of a teaching assistant (TA). You’re holding large scale office hours and tutoring on basic algebra. I did that for one semester then the following semester I was an instructor for a geometry course. It was me and two other friends. We planned the course and taught a full course on geometry. That was really cool and another completely life-changing experience in terms of opening my eyes to how different parts of the world are. I came from a super supportive family. My parents supported me studying. It was not a challenge for me to study. I came from a family that wanted me to do well. Not everyone is that lucky. I went to a school where my teachers cared about me learning and we had textbooks, bathrooms, and materials. A lot of people simply don’t. A lot of people have undiagnosed learning disabilities, are mistreated at home or at school. It’s very clear to me that if I’d grown up in the environment some of the people in prison had grown up in, I’m pretty sure I would be there too. It’s the system. It’s not like there’s good people and bad people, and bad people go to prison and good people make lots of money. That’s just not true.

I completely agree. That sounds like a wonderful program. What about MEST?

When I started grad school, as you now know, I was not 100 percent confident that I wanted to get a PhD. I was very interested in teaching and so after my master’s I decided to leave grad school for a little bit and work in some sort of teaching-related job. I applied to a lot of different jobs around the world. I wanted to travel. I wanted to do something with the fact that I had now learned a bunch of technical stuff. I applied to be an instructor at MEST which at the time was a one-year old institution in Accra, Ghana, and I got the job. It was very different. I’d never been to Ghana before.

So, what is MEST? MEST is an organization started by a Norwegian donor who wanted to develop Accra as the Silicon Valley of West Africa. He thought if we can try and provide students there with technology skills as well as basic business training, they can start companies. I was there as a computer science instructor. There were other folks with me that were more on the business side. There were a couple of us that were teaching software engineering and more computer science skills, and there was a couple of people teaching more things on the entrepreneurship side. It was an amazing experience. I was challenged in ways I had never expected. I grew. It was also just hard because we were starting with nothing. I really appreciated the fact that MIT at the time had open courseware, I relied a lot on it. I would go and look up what I’d studied in my course and develop a lecture plan. I’d never had any formal training in how to teach. It was building an entire computer science class from scratch solo with no training. In retrospect, I don’t know how I did that but it was an experience.

Has MEST been successful?

I think it is. A couple of students there have started companies and moved out of Ghana. There are multiple students that are now pursuing higher degrees in the United States. It’s really pretty incredible. The impact has been quite positive. There’s a couple of students that I’m still in touch with. It’s really nice also, the group of instructors that were there, we ended up getting to know each other and that was really nice.

One of the things that I realized when I was there was that I really wanted to go through with my PhD. I realized that there’s a lot of things you can do at the grassroots level, on the ground, but a lot of challenges are systemic and are rooted at a much higher level.

So how do you solve them?

So, that was the question I wanted to work on when I came back to grad school, which I think led to part of my interest in control, communication, organizations. It’s not necessarily a linear pathway but I think that’s how I ended up along that direction. I ended up in education because it feels like that’s one solid way to try and create positive change.

Featured image: Gireeja Ranade by Adam Lau from the College of Engineering