Interviews

Deborah Estrin

Dag Spicer Computer History Museum Editor: Dag Spicer

Deborah Estrin is a computer scien- UCLA Computer Science Department was first growing. tist who has made major technical There were a lot of social interactions at my parents’ contributions to networking, mul- house, cocktail parties and dinners, and what became ticast routing, embedded sensing known as my father’s “Probability Seminar,” which was, and computing, wireless sensor net- of course, a poker game with Len Kleinrock and other works, and mobile and electronic faculty from CS. health. She has been a professor her My father was very affected by the women’s move- entire career, has served on count- ment, initially through my mother. I used to think both less panels and advisory boards, and my parents had their consciousness raised in the 1970s continues to mentor students in electrical engineering while I was in middle school and high school. But and computer science. She is one of the most accom- recently looking through some of my mother’s old let- plished and visionary people in computing today and ters, I realize that she was a feminist long before that. also the first professor hired at the Cornell NYC Tech Nevertheless, although my mother had done her PhD campus. In 2012, Wired magazine named her one of the at the same time as my father at the University of Wis- “50 People Who Will Change the World.” consin, her career had always come second, and I think that was one of my father’s deepest regrets. My father was probably the least sexist person I’ve ever met, of David Walden: Please tell me a little bit about your any age or gender. And that was a very powerful way to youth, the places you lived, and your early educa- grow up as a young woman. tion, family, hobbies, siblings, and so on. What were my hobbies? I did a lot of Israeli folk Estrin: I was born and raised less than a mile from here, dancing, and we spent summers and sabbaticals travel- on the other side of the University of , Los ing together as a family around Europe and Israel, Angeles campus, to Thelma and Gerald Estrin. I have where my parents both had strong professional two older sisters, Margo and Judy. From the age of three, connections. I lived across the street from UCLA and went to the local My parents got married when my mother was 17 and public schools. I then went to UCLA my last year in my father was 20. They met when they were students at high school and on to UC Berkeley Engineering and CCNY [City College of New York] and were both active MIT. From my early childhood, I always remember my in political movements when World War II broke out. parents telling the story that I wanted to grow up and They had married before the hostilities began and were get my driver’s license and my PhD. This was at the age both studying history I believe. During the war, my of six or seven, reflecting both that I was growing up in mother started working in electronics—the Rosie the —where a driver’s license was an indication Riveter phenomenon. After the war, they both decided of independence—and that a PhD was the norm with to go on into electrical engineering and, through the GI the set of people they always had over to dinner and to Bill, went to the University of Wisconsin together and various social engagements. It was a time when the obtained bachelor’s, master’s, and doctorate degrees.

Background of Deborah Estrin Born: 6 December 1959, Los Angeles cofounder (2011–present); Cornel NYC Tech, professor Education: BS (EECS), University of California, Ber- (2013–present) keley, 1980; MS (technology policy), Massachusetts Awards and Honors: National Science Founda- Institute of Technology, 1982; PhD (EECS), MIT, 1985. tion’s Presidential Young Investigator Award, 1987; Professional Experience: University of Southern ACM Fellow, 2000; American Association for the California, assistant professor (1986–1992), associate Advancement of Science (AAAS) Fellow, 2001; IEEE Fel- professor (1992–1998), professor (1998–2000); Univer- low, 2004; ACM-W Athena Lecturer, 2006–2007; Ameri- sity of California, Los Angeles, professor (2000–2012); can Academy of Arts & Sciences Fellow, 2007; Anita Center for Embedded Networked Sensing (CENS), Borg Institute Women of Vision Award for Innovation, founding director (2002–2012); Open mHealth, 2007; National Academy of Engineering Member, 2009.

64 IEEE Annals of the History of Computing Published by the IEEE Computer Society 1058-6180/14/$31.00 c 2014 IEEE Walden: The government can do some end I was more drawn to designing and good things. inventing things than to analyzing the impli- Estrin: Yes, very much so. And after the war, cations of what someone else had invented. at that time, anti-Semitism was a factor when they were looking for jobs. They ended up Walden: You also got a master’s from with at the Institute for MIT, though. Advanced Study at Princeton. From there, Estrin: I did, in technology policy. And then I they went to the Weizman Institute in Israel; went on and did my PhD in Course 6 (EECS). Chaim Pekeris visited IAS and wanted a machine for his research and government Walden: Jerry Saltzer was your advisor, at use, so they brought my father over to run least for your PhD? the project and build the Weizac. Estrin: Yes, he was stickler for both vision and detail—quite a combination. Walden: After high school, you went on to Berkeley and chose electrical engineering Walden: Not for your master’s? and computer science. Was this because Estrin: Jerry was on my master’s project along it was your parents’ field? Often children with Marvin Sirbu. run away from their parents’ field. Estrin: Well, I ran away from them physically Walden: How did you end up working and left LA. I was about 17 and I remember with Jerry? my first phone call home to my parents from Estrin: At that time, Jerry was starting to have Berkeley was, “Why did you move to Los more multidisciplinary interests, so when I Angeles and raise me there?” I felt somehow went to look at the Laboratory for Computer a little bit more at home in the Bay Area than Science (since merged with the AI lab to form growing up in West Los Angeles in that era of CSAIL) for possible advisors someone Beverly Hills and Hollywood and all of that. directed me his way. He had an interest in Now the whole world is like that. cable TV and data over cable—this was It was a time when many more women 1981–1982. I did my master’s around data were doing things that were not just the tradi- communications over cable TV. I explored tional things they had done, and I loved math adapting multiaccess communication proto- and science. Also, I always had this drive to cols (such as what makes Ethernet work) to design and invent things. So with a desire to longer distance CATV networks as well as create new things, engineering and EECS was some nontechnical issues. In particular, back just a natural; so much was happening in then, there were all sorts of technological and terms of new technologies. And that became institutional questions about whether the even more true—that was really the draw. Mom and Pop cable operators were going to be able to upgrade their facilities and manage Walden: Did you focus more on hardware advanced data services. So I started that proj- or software there? ect with him and Marvin Sirbu, and then I Estrin: At the time, it was a double EECS continued on with my PhD with Saltzer. major but it was more software than hard- ware. In my last year, I had an incredible class Walden: What was your topic? with George Turin on the communication Estrin: The title of my PhD was “Intra- side of things. That really set the stage for my organizational Networks.” It was about doing interest in networking. Internet connectively across administrative boundaries and involved topics related to net- Walden: After Berkeley, you went to MIT. work security policies such as how to define What did you do at MIT? the access control and information flow rules Estrin: I got to Berkeley and went through it and mechanisms for remote operations inside quickly—I don’t know why I was in such a a different administrative domain. hurry. I had always had a little bit of my parents’ politics or activism in me so when I Walden: After graduating from MIT with went to MIT, I entered the technology policy your PhD, you ended up back on the program to study the social implications and West Coast at the University of Southern policy around technology. But I ended up California. doing my PhD at MIT in EECS (although it Estrin: I was hired at USC by then Chair was rather multidisciplinary) because in the George Bekey. That turned out to be a

July–September 2014 65 Interviews

tremendous stroke of fortune because, after a can do relevant work. I wanted to work in year or two, I started spending more time at areas where I felt there was a chance the work the affiliated ISI [Information Sciences Insti- would have impact and where I would have a tute]. It was actually Danny Cohen who, I better chance of learning from deployment think, said to Jon Postel and their group, “We and use. It’s sort of what caused me to shift should invite Deborah over and give her an my focus to study wireless sensor networks. office.” I started spending most of my time In retrospect, I was wrong in general but there and, most importantly, having my right for me. Wrong in that many academics graduate students there too, giving them the have had tremendous influence on the shape opportunity to be apprentices on advanced of networking through their research in the development projects ISI was doing. That was decades since—Scott Shenker, Dave Clark, the basis of the 15 years I spent at USC. Jennifer Rexford, to name just a few. Right, in that I personally was ready for a change because that’s how I am. Walden: For our readers who don’t know what ISI is, could you just say a word about how that’s distinct from, say, the Walden: Tell me about that. Computer Science Department at USC? Estrin: I ran an ISAT [DARPA’s Information Sci- Estrin: The Information Sciences Institute was ence and Technology advisory committee] started long before I was there and was affili- study called “simple systems,” and it was one ated with USC as a kind of research lab, sort of the things that led DARPA to launch the of what Lincoln Labs is to MIT or SRI was to SenseIT program. Over a pretty short period Stanford. They did some very early important of time, it became the entire focus of my own work related to speech and the Internet. Jon research projects. Our SenseIT research project Postel was the person who ran the domain at USC/ISI focused on mechanisms for coordi- name system and operated a lot of other nated distributed sensing, and I agreed to seminal infrastructure around the Internet; chair a National Research Council study that he used to personally give out IP addresses produced a report “Embedded Everywhere.” and domain names—he was the name and Through this early work, it became in- number czar. He was the governance of the creasingly clear to me that doing experimental Internet when it was too small to have any- distributed sensing without a strong connec- thing else. The DNS developed by Paul Mock- tion to the application domain for which the apetris was built and run at ISI. The RFC datawerebeingusedjustmadenosense.For [Request for Comments] system was also run the previous 15 years, I had safely and com- out of ISI. It was a place where advanced fortably been working at a level of abstraction infrastructure was being designed and built below the application, but now I felt like there and run in production as a way to do com- was some circuit not being closed. So I had the puter science research. ISI shaped the way I idea of developing a center that would bring think about and do research. the users and designers of distributed sensing That was an amazing time of people doing under one roof and decided to apply for an research and academic work but with output NSF Science and Technology Center grant to that was not solely academic. One of the last make that feasible. Center scale grants are core networking projects I did was related to unique in that they have the scale of funding multicast routing, which was a really interest- and duration to bring together collaborators ing design opportunity. I had the chance to for productive multidsciplinary work that work with incredible minds like Van Jacob- would not happen otherwise. son and Steve Deering. Ironically, it was also I came to UCLA and immediately began that experience with multicast routing that the long competitive application process for caused me to move away from core network- Center for Embedded Networked Sensing ing because, in the end, multicast didn’t get (CENS) to develop environmental monitor- deployed. And since I’m an odd sort of aca- ing applications and the underlying sensing demic—things don’t interest me solely for systems. UCLA was the prime institution, academic reasons—-that was a bit traumatic with USC, UC Riverside and UC Merced as and I thought, well, if the Internet has partners. After a long review process, we become such a critical infrastructure that it’s were awarded the grant. CENS wasn’t about difficult to influence the core technologies funding technologists to have small encoun- from academia, then it’s not clear how as an ters with applications; there was both the applied academic doing Internet routing I time and the resources to support true

66 IEEE Annals of the History of Computing co-innovation between application and tech- to enable a kind of full above and below nology. The NSF Science Technology Program ground ecosystem measurement system for is fantastic and unique. It gave us a tremen- the purposes of modeling. dous opportunity to do application-driven Initially, we came in with a naive set of research. From 2001 to 2005, I focused on dis- notions about what the scientists needed and tributed sensor arrays and then, around 2006 what were interesting distributed algorithm or so, became interested in mobile phones as problems that we could solve. Because of the a type of sensing data source, which has led commitment to actually do things that were to the work that we ended up doing in partic- relevant to the applications, and a style that ipatory sensing and mobile health. involved rapid prototyping and pilots in the field, we became much more flexible in our Walden: Stepping back for a second, you technical infrastructures and algorithms and mentioned that you moved from USC to much more focused on what would actually make a difference to the scientists. UCLA. How did that come about? Estrin: You know, they asked and it was just at Walden: What is your evaluation of the one of those times. I’d been at USC for 15 impact this work has had? Is it ongoing? years, and I was starting to think about this Has it helped? more application-driven work. I began that Estrin: Yes, I think that there are two branches work with some wonderful colleagues at USC of the work. With respect to the scientific in marine biology, actually. It wasn’t some- domains involved, it did have a lot of impact thing I was seeking, but when UCLA asked, it on our colleagues in those fields and what felt like the right thing to do: moving on from those fields are doing. We influenced the where I’d started at the age of 26 and going to design of NEON, which is a huge piece of sci- a larger university that had an even larger entific instrument infrastructure for the base of application domains in the sciences country. And I am told that we had a lot of and things. It was just time for a change, and impact on the way in which ecologists are it was a change that fit in with the fact that I using these kinds of distributed arrays and was a single mom. Being close to my parents cameras and so forth. In terms of computer was also very important to my son and me so science, I think we raised the bar of what it I wasn’t looking to leave LA then. means to do applied and experimental work and engage with real applications. Walden: In the sensor area, one of the One of the frustrating things was the fun- areas you worked was eco-sensing, damental economics of fixed sensing. When putting people with sensors out in the you put a sensor someplace in the physical world. How big a thrust of the sensor environment, it is measuring that cubic cen- effort was that, or is that just a little timeter or meter and that’s it. It better be a piece of what you did? really important cubic centimeter or meter, Estrin: Environmental monitoring and devel- maybe because it represents something in a oping innovative, multimodal, distributed model or it’s a hotspot in a server room or on sensing systems was the thrust of the center. a factory floor. Bill Kaiser and Greg Pottie ini- Ecology and ecosystem monitoring was the tiated a project that I was part of on robotic- biggest application domain. I was also very based sensors where we started moving sen- involved with NEON, the National Ecological sors on aerial cables to be able to cover more Observatory Network, which is now an NSF of a two-dimensional and, in some cases, major research and equipment facility. three-dimensional area to get a better, fuller They’re building something that will live for model of what’s going on and not just be 20 to 40 years across the continental US; a set stuck in certain fixed places. The economics of distributed ecological observatories in for mobile sensing happened to occur at the which people can do systematic exploration same time that mobile phones were getting and long-term research around things like more and more powerful. the impact of climate change on ecosystems. For example, phones were newly program- It was in that context that we started to mable, knew where they were thanks to GPS look at imagers as biological sensors—using components, and were something that peo- cameras in observations of foliage and their ple plugged in and recharged at the end of transitions and of birds and using infrared the day. So we started to do something we cameras to pick up soil phenomena com- called “participatory sensing,” a combination bined with below-ground soil arrays—trying of automated and human input. It was a

July–September 2014 67 Interviews

even one person at a time benefits from the innovation. Over time, as the approach pro- When it comes to scale, liferates, we can benefit from the opportuni- ties that come at scale as well (population you can’t iterate based models, discovery, and so on) Now, this approach has its downsides; it experimentally in made my focus more short term. It’s very important to the field that we take a range of the context of some perspectives, so while this was the right path for me, I am not suggesting it prescriptively. future situation that We need a full spectrum of research styles.

doesn’t yet exist. Walden: Could this work be used, say, in treating diabetes? Estrin: Yes. In fact, we had a showcase in December 2012 at which one of the scenarios natural next move to conclude that not only we demonstrated was for Type 1 diabetes; our was the environment in which the people are use case was an individual who needed more moving around of interest but so were the information to better manage her condition. people themselves. It was a natural, not pre- We demonstrated the insights she and her meditated, progression from distributed sens- clinician gained by integrating diverse data ing to mobile sensing, to mobile phones, to streams: blood glucose measurements, physi- participatory sensing, to mobile health. cal activity monitoring, and data from apps for journaling sleep and mood. The ability to bring all those streams together to inform Walden: Was mobile health still happen- self-care and provide clinically actionable ing within the CENS center? data is what mobile healthcare is about. Estrin: Yes, being a NSF Center, CENS allowed for a lot of flexibility. We added participatory sensing to the existing domains of seis- Walden: Since NSF Centers only last 10 mology, environmental engineering, marine years and now that CENS has closed, ple- biology, and ecology. That included mobile ase tell me what you think might happen health and came toward the end of CENS. next. Will you bring some of what you’ve We pursued the participatory sensing and been doing to that? mobile health work with applications that Estrin: I’m delighted to be continuing and would produce benefit for users even on a growing my work at Cornell NYC Tech. Cor- small scale, rather than starting with applica- nell Tech is a bold vision inspired by a New tions like epidemiology that require large- York City initiative under then-Mayor scale usage before they can deliver results. Michael Bloomberg and enacted by Dan Hut- When it comes to scale, you can’t iterate tenlocher, dean of the Cornell NYC Tech experimentally in the context of some future campus. The vision was to draw on an aca- situation that doesn’t yet exist. I learned demic institution with great depth and his- through distributed sensing that it’s hard to tory, Cornell, and create a new center for do meaningful applied experimental work digital innovation rooted in the urban con- when it is all contingent on a posited future text. The three initial research hubs are in (such as the massive distribution of sensors) Healthier Life, Connected Media, and Built that isn’t yet buildable. I wanted problems Environment. When I first looked at the pro- that scaled down as well as up. Because if a posal back in 2012, I thought, “They’re sing- problem doesn’t scale down, we can’t build ing my tune, singing my song, whatever it is, and use it. I can’t experiment with it; I can’t humming my tune.” When they asked me if I explore it and then build up from the suc- would be interested, it was really just such a cesses if it doesn’t scale down in number. clear cultural match in the way they wanted So our work in mobile health started at the to see the work done there. scale of the individual. By them running an I’m excited about doing that kind of application, there’s an intensity of informa- enmeshed entrepreneurial work. We are tion, a time series, and a temporal resolution about commercial entrepreneurship as well that you can understand about them as an as social entrepreneurship, so my work with individual. We focused on applications where Open mHealth (http://openmhealth.org/), a

68 IEEE Annals of the History of Computing nonprofit startup that is all about catalyzing time being a single parent. We’re very close commercial innovation and being a part of and part of what stimulated me to embark on the open architecture world, is very much new endeavors (like distributed sensing), ear- applicable. lier than I might have otherwise, was that I liked having exciting things to tell my then young son about. Being a mother and having Walden: What part has teaching played this young man as my close friend has been a in your life? very important part of my life as well. Estrin: The best part of being a faculty mem- ber is working with fantastic colleagues and incredible students. Sometimes it’s more Reference and Note PhD students and other times it’s more 1. This interview of Deborah Estrin was initially master’s students, depending on the mix of conducted by David Walden on 19 Nov. 2011 problems. In my last few years at UCLA, I’d in Estrin’s office at UCLA. This original been working with undergrads and high interview was 7,400 words long, and it was school students to add to that mix. I really edited down to the length of this column by like teaching through apprenticeship, where Dag Spicer with the approval of the approval students become experts through scholarship of the interviewee. and building, and we are bringing that culture toourformalclassesaswellasindependent Dag Spicer is senior curator at the Computer History research work at Cornell Tech. Museum. Contact him at [email protected].

Walden: Is there anything final you’d like to say? Estrin: Well, for a complete picture of things, I have a 27-year-old son and I spent a lot of

July–September 2014 69