Penultimate

Hello readers :).

I can’t believe this is my second-to-last post, or that I’m almost done with 16^th grade. Last week feels like it happened a month ago—in the time between this post and last, my three majors stressors have been resolved. Thesis defense on Wednesday. Linear Algebra test on Monday. Intro to Machine Learning test on Tuesday. Wow.

On Wednesday at 2:56pm, I ran from the bus to the Science Center, passing Misha along the way. “You’re in a rush!” she called, as I sprinted up the steps. “Thesis defense!” I yelled back at her. “Are you late?” I heard behind me. “Not yet!”

I had been at office hours for Linear Algebra at MIT, but made it back to Wellesley in time to show up to my defense before any of my invited faculty arrived. I had a great group of people—my two research advisors, my third committee member, the neuroscience department head, and my honors visitor/moderator. These are the professors with whom I have interacted the most at Wellesley, and they have written all of my recent letters of recommendation. It was wonderful to have them all in a group together, to listen to me speak for an hour about what I’d learned in the process of researching and writing this document.

I have to say, I witnessed an amazing range of interactions in that room. Really, I’d almost recommend the thesis process just for the experience of defending. At dinner later that night with six other science-thesising seniors (impromptu, as well—I’d never had a spontaneous gathering with a group that large), I learned that everyone’s defense was quite different. Mine was unusual, and I’m grateful for my experience.

In many science thesis defenses, the student speaks for 10-15 minutes, and the rest of the time is devoted to questions. Usually there are four people in the room—your research advisor (I had two), another member of the department, the department head, and your honors visitor (from outside the department). They have almost all read your thesis, and are familiar with your work to different degrees, depending on when they joined you in the process (your committee members are more familiar with your work than the department head or the honors visitor) and their general knowledge about the topic. Thus, some defenses are filled with difficult questions, some contain mostly clarifying questions, and some finish early. Your honors visitor is there to make sure that everything stays on time, and they are also a moderator, especially in watching out for how much the student gets to speak.

My honors visitor did a great and necessary job during my talk. She told me she was there to look out for me at the beginning of the hour, and I smiled, happy to have her there (I deeply admire her), but privately thought that it probably wouldn’t be necessary. In my committee meetings earlier in the year, I had gotten by just fine without a moderator, and I am generally quite willing to stick my hand in the air and wave it around. But I had collected a great set of invested and adamant professors, and soon enough I was right in the middle of a heated debate about the use of dominant wavelength and what kind of controls I might have needed and should have used. Never a silent moment.

One professor was largely silent, and smiled at me encouragingly throughout my presentation. (I had a thirty minute talk prepared but welcomed people to ask questions throughout.) One professor asked a few clarifying questions, but otherwise listened attentively. One professor again smiled at me encouragingly and nodded in places. One professor had come with a list of issues that needed to be addressed and asked lots of questions when I came to those points in the summary slides. One professor followed up on my responses to those questions and elaborated on both the content of the questions and what specifically I had done. And throughout it all, the questions I received (mainly from that one professor) were focused, quick, and delivered forcefully. There wasn’t anything I couldn’t answer (though I of course could have answered many better) but what threw me off was that many of the questions were not in the direction that I had anticipated. I had prepared a presentation based on what I thought was most relevant in the thesis, and we kept veering into what I thought were details, or tangential areas of color theory I hadn’t explained fully or well in my thesis. I had thought this presentation might progress like my Ruhlman (our student presentations two weeks ago to the public), where I’d present my material and then answer a bunch of follow-up questions. Instead, it turned into the most engaging discussion I’d ever been a part of, because there were ideas flying back and forth, strong opinions being thrown, and I knew how to answer them.

When I attended the CoSyNe (Computational Systems Neuroscience) conference in March, the last discussion I attended had a similar feel. There were four presenters at the front of the room, but the individuals in the audience were fully engaged, and everyone was calling everyone out by name as the questions ping-ponged back and forth. I could only watch this flow of energy in amazement, because my lack of knowledge excluded me from the conversation, even as I could admire the intensity and intellectual depth with which it was happening. But in this thesis defense, I knew what was going on—I understood the questions, I had answers to the questions, and it was just the problem of being able to describe these answers clearly enough to make myself understood. There was also the problem of listening well enough to understand which of the professors’ comments just needed clarification, and which comments I really should have incorporated into my thesis and were faults in my design (the difference between the two is more subtle than I ever imagined). There was so much energy in the room—and surprisingly, it wasn’t stressful, because I was focused on getting my point across and I felt supported by all of the professors. What most surprised me, however, was how I felt I wasn’t being heard.

Oh, I was talking plenty. I was doing a lot of explaining. But then we’d get into a follow-up on a follow-up, or I’d mention that I didn’t think that was the relevant point; I’d rather talk about this (not because I don’t understand what you want to talk about, but because I think this is much more interesting), and then there’d be some disgruntlement and one of the other professors would jump in with an answer. Soon enough, I’d be watching the two professors discussing the points for a while—and watching very closely, for two reasons. First, because I’m a student, and when you’re a student you need to pay attention because everyone knows so much more than you do. (Also, recently, I need to pay attention to make sure to advocate when I don’t understand something or when I really don’t think it would work. I couldn’t do this when I was younger.) Second, because eventually it would be my turn to talk again and I’d have a list of points I’d want to present. As a note, this is not to say that I wasn’t learning things from the discussions between the two professors. But I’ve never literally opened my mouth so many times and then closed it again (three times in a row during one question) because someone else was still speaking. And I’ve seen this before—when I went to a presentation at Harvard where a really talented postdoc was being questioned by an audience of mostly male professors. She nodded really attentively, and started to speak, and then closed her mouth, and then started, and then waited, and only finally began once the speaker was done. In short, my thesis defense was unexpectedly frustrating.

I recounted this story to probably five friends, and my feelings have crystallized every recounting, but what was the most informative was getting input from my father. Because two things that had been interacting in both my story and the postdoc story are 1) power differences, and 2) gender differences. Being at Wellesley, I’m exposed to a lot of discussion on these two points (especially the latter) but the effects are always implicit, and I generally, for better or for worse, stray to the “benefit of the doubt” side. But after emailing my father, he confirmed that I was seeing both power differences and gender differences. Acknowledgement is very important, because I was only able to tell my friends my interpretation, and I didn’t know if I was reading the situation correctly. After my defense, for example, no one (appropriately) mentioned anything about the strange dynamic—everyone offered me a hearty congratulations, and then left to go to other meetings, so I didn’t receive feedback there.

My father, though, supported my tentative interpretation of the situation, gave a stronger interpretation, and then gave me solutions. First, he said, resolving this kind of situation depends on the type of talk one is giving. If you were interested in getting a specific point across, then you need to take control of the talk right at the beginning—remind everyone that to save everyone’s time, you need to continue on with your slides, and that you will take questions at the end. If this were more of a thesis defense kind of talk, when the audience is encouraged to ask questions in the middle, then it might be fine to let the conversation go where it leads. But the overall point is that whatever tone you want to establish, you need to do it at the front, so that everyone is aware of the expectations—and this all needs to be done in a politic way. Meanwhile, I was sitting there staring at this email, and going “huh”. Because this is a skill I have thankfully never needed to develop in my time here.

At the CoSyNe conference, my favorite female presenter was in control of her talk. She would listen to questions, but interrupt the speakers if there was a flaw she saw immediately. Whenever there was a question with a suggestion she hadn’t thought of, she would pause, say “That’s a good point,” pause again, and then continue. My father mentioned in his email that if a presentation’s context allows it to be more audience-based, it’s fine to let the questions go where they lead. And in one of my Machine Learning classes recently, that’s exactly what happened, with the presenter mainly driving his presentation based on the audience’s questions rather than what was on his slides. But that was different from not having control of the presentation. Of opening and shutting my mouth multiple times like I’ve ever only read about. I think the fact that this has never happened to me says a lot about Wellesley’s classroom environment, and the fact that I have never not been encouraged to speak. But then, this thesis defense was a different environment, one where the power dynamics were especially interesting, because I and my advisor knew the most about the topic, I knew the most about the details, one of the other professors was very invested in the topic but had less experience, another had already given me copious feedback and didn’t feel the need to add more at this juncture, and two more weren’t familiar with the topic. Plus everyone was my superior, and all of these facets were interacting at once. It was a thought-provoking experience. None of the other thesising seniors I’ve talked to experienced anything like it, but I was so glad that I experienced it—because this is what I’ll be encountering in the future, and I’m happy to see it for the first time in an environment that is supportive on so many fronts.

In summary, that was my thesis defense. It lasted just one hour, but integrated with what I’ve observed from female scientists at different levels of their careers, it was an extremely informative and valuable experience that I’m glad to have lived for the first time.

…

Hm :). Also this week, I studied a bit for linear algebra, and I studied like mad for Intro to Machine Learning. I have been worried about Machine Learning since March. My poor friends have had to reassure me again and again that it was going to be all right—huge thanks to Emily, Sebiha, and Tiffany for bearing the brunt of it. Because I don’t know if I’ve mentioned it, but I failed my first test in college this semester. 32/60 on my Machine Learning midterm, baby. There was a curve, but I was left pretty far below it. I’d never failed a test in college before—the last one I’d actually failed was in AP Stats in high school—and my unremitting goal throughout these last two months has been to pass this class. Yesterday, the test wasn’t a disaster, and my emails to Sebiha and Emily involved “I PASSED!!!” and a few excited swear words. Sebiha said she’d had no doubt. Since I was the one studying for this thing, I had plenty of doubt, but I am so, so relieved that it’s done.

It’s a strange situation, because I really like Machine Learning. Very brief (and incomplete) primer on what machine learning is: you know how when you write programs, the computer usually does exactly what you tell it to do? What’s also useful is when you write programs that let the computer learn from the data you input. Like, if I go to Amazon and buy a bunch of things, machine learning strategies can be used to try to figure out which people are similar to me based on what they’re buying, and then apply that information to recommend what I should buy next. These are called “recommender systems” and are just one of the many applications of machine learning. Since I want to study something related to social inference in the future—meaning, investigating how we learn about social situations with an algorithmic approach—I like trying to understand how we make computers learn from information. Machine learning is a way of breaking down how we learn, given all of the data we’re presented with in the world.

I have not done well in 6.036 Intro to Machine Learning. My homeworks were great, my psets happened (see the post two weeks ago for the beautiful story of my last pset), but that midterm, man. However, as I was worrying about failing before I took my final, a question occurred to me: who is my peer group here? Who is in this class with me? Because they wear their orchestra shirts and sorority gear and generally participate in all the normal extracurriculars—but MIT students are really, really good. And, uh—this occurred to me about partway through the semester, but I skipped the pre-requisite for this class: 6.01, Introduction to Electrical Engineering and Computer Science I, which itself requires Physics II as a co-requisite. Who knows what that class would have taught me, but I know that I was sitting in 6.036 and asking dumb questions like “What does that upside down A mean?” (my best guess is: “for every [whatever variable follows]”) and “What did you just do with that log function?”. I was also trying to figure out how to use Python for the first time, which made my first pset depressingly difficult— if we hadn’t been trying to implement simple algorithms I would have been in major trouble. (One of my friends, Eduardo, was looking over my first piece of code, and laughed at how Matlab it was :).) In short, I was a little behind in everything: in math and coding and general knowledge. If I hadn’t seen and worked with the second half of the class material previously—I’ve seen Bayes nets and Hidden Markov Models before, in my Computational Cognitive Science class, and I took 18.06, which covered Bayesian probability—I truly believe that I would have been far below the average on this final. In fact, the realization that I’d seen the material before, halfway through the semester, was sudden and strange: the idea that this is so much easier. I actually understand what is being taught to me. I can now focus on the details, and the things that differ, rather than trying to grasp the overall concept and all of the unusual things all at once.

An interesting question is whether I would have taken the class again, and whether I would recommend the class to others. First: if I could go back, would I have still taken the class? Answer: yes. I’ve learned so much, about a topic I’m really interested in, which made the problems worth it. Yes, I was stressed about this class the entire semester (and that’s not something to be underestimated. It changed a lot about my behavior this semester). And yes, 6.036 is not going to be kind to my GPA (though there are advantages to being a second-semester senior!). But the course was just so interesting—it put so many ideas in my head, put a lot of the things I’ve learned in a great context, gave me experience with Python and these algorithms. It was worth it. And just so you know that I don’t say everything that’s hard is worth it: Organic Chemistry II was not worth it. I did not do well in Organic Chemistry I, and was really wavering on taking Orgo II, but did so because my parents encouraged me to and I was, for some odd reason, fulfilling pre-med requirements even though I knew I wasn’t pre-med. The worry and stress there was not worth it (though I admit the lab was awesome). Also, not directly related to if it was worth it or not, but if I were to go back and restart my research career, there would be a lot of changes. Luckily, the sheer amount of mistakes I made meant that research has been my single greatest learning experience in college, and I will take everything that I’ve learned from my undergraduate research experience into the future.

Second: would I recommend the class to other people? This one’s a harder one, because you don’t actually find a lot of people at Wellesley who are interested in what I’m interested in. This is one of the reasons why the people in Prof. Conway’s lab are always interesting—because they want to study neuroscience, but more specifically they want to study computational neuroscience. Basically, we try to understand the brain by sitting at our computers and coding all day. This is an odd intersection of fields: usually you have your computer science people, and you have your neuroscience / biology people, but you don’t find a whole bunch of people who want to do both. Moreover, to do the sort of computational modeling work our lab does, you end up learning information from all over the place—e.g. taking classes in the math department, in the psychology department, in the CS department, in the applied math department, in the neuroscience department—and as you can imagine, it’s a pretty small subpopulation who wants to do all of the above. Remarkably, Isabelle, who’s a year below me, pretty much has the exact same research interests as I do, and she’s told me repeatedly how helpful it has been to have me at Wellesley. She works remotely at a lab at Caltech, and is also different from me in that she’s a Philosophy minor– but she’s taking similar classes to what I took (an eclectic mixture, now that I look back on it.)

I wouldn’t recommend the class to someone who isn’t really interested in the material, because the whole experience was a bit arduous. I’d also be cautious to recommend it to someone who hasn’t had at least some of the background that’s expected for that class—and there’s a lot of it. Like I said, if I hadn’t had two classes on Bayesian probability, the second half of the course would have been just as rough as the first half. There were a lot of implicit pre-reqs—for example, there were references to things like rank, and other linear algebra terms, which were not expected for us to know and yet were helpful in understanding the information. I think this logic applies to almost all of the upper-level courses at MIT, because the thing is: as a Wellesley student—with the interests that brought you to Wellesley—you’re unlikely to have as much background as a typical MIT student, even one who is one or two years below you. So if you’re fine with that, with feeling slightly lost all the time, and you still want to learn it: my advice is go for it.

In bringing it back to my own experience, my final thought on Machine Learning is that I’m so glad I had to struggle. Because I’ve tutored Wellesley students, I’ve had so many friends who have struggled with classes, and I never truly have. I’ve had to work very hard on quite a few of my classes, but I’ve failed a test before. And it’s a stupid thing—one of my first-year acquaintances laughed at me disparagingly when I mentioned I’d never failed a test in college before—but I need the empathy. To know that really, when you’re trying your best, full effort, full stop: this is the best you can do. That it’s not just a matter of: put more things aside, work on it more, just work harder—no, given your situation, given what you know, this is the Best. You. Can. Do. That this occurred senior year, spring semester, when I’d already gotten into graduate school… the perfect timing is almost a blessing. I glad it happened now, and I am glad it happened.

…

It’s so weird to say goodbye to everyone—to say goodbye to my friends of two years at MIT, and longer at Wellesley, and look at them, and be like: well, I guess that’s it— forever. Because our paths likely won’t ever intersect again. The strangest thing is that it can’t be momentous for me—it’s just like leaving them at the end of the class, ready to see them next week. It’s the strangest experience, and the last thought I’ll mention before closing this post.

Thank you all for listening :). This was a long post, so I appreciate you hanging in there! Best of luck with finishing up school, and see you for my final post (nooo) next week :).

Monica