What's the matter with MOOCs?

I've ranted about this before, but it's only gaining more traction, so I find myself unable to not comment on the trend yet again. What trend? MOOCs, or massive open online courses, which I'll happily lump together with any online courses, though at least the open ones don't require you to pay (so far) and have professors from prestigious universities teaching there. Still, the pitfalls are the same.

1. A narrow student body. MOOCs claim to be open to everyone, and sure, anyone can sign up. But how many students can succeed? What does your educational background need to be? How about your motivation level? The creators of MOOCs are themselves highly successful highly motivated people who have excelled in the traditional classroom. I believe they have themselves in mind when they create MOOCs. Perhaps the vision of themselves extends to "what if I had grown up in a remote part of the world without access to great educational resources?" But it is still a narrow vision.
2. Limited pedagogical views. Online courses, because of technological limitations, have a limited number of ways that students can interact with each other and the teacher. These are typically videos, online chat rooms, and forums. There is nothing to mimic in-class group work; working with your peers while asking questions from the professor. There is nothing to mimic in-person immediate interactions; the ability to gesture wildly and draw pictures and receive hints, not answers. To give credit where it's due, the Udacity courses apparently stop the video frequently for quick "check that you're following" quizzes. I do that in my own classes, and it's great - no class should be a lecture. But what if the students can't answer the question on the short quiz, even after they rewind?
3. Emphasizes the mind / body split. bell hooks writes about the mind/body split in teaching. How teachers are viewed purely as minds and lacking in body and lives outside of the classroom. I believe this is frequently true for students as well, and MOOCs make this split harder to overcome and potentially more permanent. There is nothing to mimic longer term mentoring; asking your professor how to get a job, talking about what you should do with your life, or asking for a recommendation. There is no mentor who knows you, in the context of your community and in the context of the class. There is no teacher to determine if you're doing poorly because you're not working hard enough or because you're sick or because you're working three jobs just to pay the rent. And there's no teacher to make sure you get through that, by being lenient or tough, depending on what you need.

Despite all this, the American Council on Education is considering awarding AP credit for courses taken online. Perhaps this will be fine for AP students. In fact, despite my objections, online courses provide a valuable service for many students; they allow motivated, dedicated, intelligent students who have the passion and determination to learn on their own, or have spent years in the workforce learning material on their own, the chance to gain recognition and official acknowledgement of that determination and knowledge. But let's be clear, they learned that information on their own, organizing study groups on their own, watching online help videos on their own, posting to online forums for help on their own. They've gotten help, and perhaps a class helped to organize these forums, but these students are determined individuals who deserve credit for going the extra mile to get an education. They are not the norm.

And when we consider awarding AP credit for online courses, we bring online courses into high school. High schools are not filled with dedicated intelligent students who are determined to get an education no matter the intellectual and emotional difficulty. Yet the right to an education, the right to be taught, is, I believe, a fundamental civil right. Let's not chip away at it.

Back in academia

I'm back in academia - I've taken a position as a visiting assistant professor at Haverford College. So far, that's mostly meant moving cross country and lots of unpacking. But now I'm starting to prep the two intro classes I'll be teaching this fall (one directed towards majors, one for non-majors). It's exciting!

So hello again. I may never be allowed to talk much about all the cool things I saw at Google, but for now, my thoughts are my own.

An Update from the Void

I've been a bit quiet lately, I know. I find it's hard to talk about anything when I'm not allowed to talk about work, and I've been working hard on an unreleased project. It's been wonderfully exciting and interestingly hard, and we launched today so I can finally say (a little) about what I've been spending all my time doing.

I'm part of the indoor maps / location team. Some of you may remember that I'm somewhat obsessed with maps. Now I get to work with lots of other people who are also obsessed with maps. (Did you know that for a long time California was thought to be an island?) Surprisingly to me, there's only a small amount of Computational Geometry being used in all of this, so I've been getting a crash course in Machine Learning, which has been enjoyably educational. I am, of course, planning to inject geometry wherever I can... perhaps eventually I'll be able to give an update on that as well. Until then, I'll have to make do with telling you to download Maps 6.0 and give it a try.

Online Courses

One of the questionable outcomes of the teacher myth has been the increase in online courses at the K-12 and college levels. (Here I mean courses taught entirely online, in which students and teacher never meet face to face and much content is learned by the students on their own by reading through online content.) On the one hand, putting information online is great (if it's open content) - students all over the world who don't have access to formal education can learn from it. Teachers can use it in their own classrooms, thus avoiding reinventing the wheel more than necessary and saving them time. But I worry that it walks a dangerous line inspired by the teacher myth: teachers are unnecessary. What about role models, emotional support, and the ability to ask questions?

Anything perpetuating the idea that teachers are unnecessary also potentially supports other problematic corollaries of the teacher myth, leading to the replacement of tenured unionized teachers with lower paid adjuncts (both at the secondary and college levels). All of this contributes to the de-professionalization of the field. So while we may be pro-online content and generally pro-technology, we should carefully consider the real consequences of being pro-online courses.

SoCG Accepted Motion/Sensing Papers

Yes, this is somewhat delayed. Better late than never?

Here's the list of exciting-looking motion/sensing papers culled from the SoCG 2011 accepted papers list. Also known as my to-read list.

• Mark de Berg, Marcel Roeloffzen and Bettina Speckmann. Kinetic Convex Hulls and Delaunay Triangulations in the Black-Box Model
  
• Umut Acar, Benoît Hudson and Duru Türkoglu. Kinetic Mesh Refinement in 2D
  
• Rishi Gupta, Piotr Indyk, Eric Price and Yaron Rachlin. Compressive sensing with local geometric features
  

Sadly, none of these seem to have a pdf preprint up yet, providing me with yet another excuse for procrastination...

A Full Day of Grading

Teachers, TAs, and professors love to complain about the amount of grading they have. But how much is that, exactly? I've seen studies that say assessments and assessment-related activities (I imagine this includes creation, proctoring, and grading) take up 1/3 - 1/2 of a teacher's time. I certainly believe it. But I thought it might be interesting/useful/fun procrastination to consider a rough estimate of how many hours per week it might be reasonable to expect a teacher (in this case, imagine a high school teacher) to spend just grading. And I found what I came up with to be rather amazing (despite all the time I've personally spent grading).

Here are my basic low-end assumptions:

• A teacher has about 100 students (imagine 4 classes of 25 students each). These days, many teachers teach 5 classes of 30 each. Some pampered private school teachers (like I was) teach 4 classes of 15 students each. Still, 100 is a reasonable estimate and a nice round number.
  
• Each student hands in one page to be graded each day. This will likely be a single sheet of homework. This excludes any tests or quizes, which are generally longer than a sheet, so this is definitely a low-end estimate.
  
• Each page takes the teacher 1 minute to grade. Again, this sounds like a short time to me if you're actually grading the work. And if you also consider the time it takes to determine the full-page grade and enter it into the computer system...
  

These estimates lead to a need to grade 500 pages per week, which I'm estimating could take 500 minutes. That's 8.33 hours, or a full work day. Yes, a teacher could easily spend a full work day each week just grading. I find that astonishing.

SODA: Instance Optimality

The highlight of SODA for me was the plenary talk Computational Geometry for Non-Geometers by Timothy Chan, especially the part in which he discussed instance optimality. The basic idea is to consider algorithm analyses that take all inputs into account and operate optimally over all inputs in comparison to an algorithm designed specifically for that input (but not for that input's order). I believe this was first considered in a geometric setting (Instance-Optimal Geometric Algorithms), so you could imagine that the input is a set of points and the algorithm you're competing against doesn't know in advance in what order the points will be given to you. This wonderfully fixes the problem with worst-case analyses in which an algorithm might perform poorly on most input instances but there is some hard instance for which the performance is reasonable. Instead, under an instance-optimal analysis, the algorithm is expected to perform well on all input instances.

In the study of data structures for moving objects, a similar idea known as motion sensitivity was previously introduced under a motion-specific setting. Motion sensitivity generally describes algorithms that perform well under orderly motion and have performance that degrades as the point motion becomes more random. In my own work, I consider entropy-based analyses in order to capture this concept. In fact, the instance-optimal analysis Timothy described also resulted in an entropy-like inequality, so perhaps this relationship is inevitable/necessary for these analyses.

It's an exciting new area that I'm hoping to see a lot more of. It was certainly an excellent talk.