Monday, May 27, 2013

Becoming Part of a MOOC's Solar System

Last summer I started producing my own videos on artificial intelligence (AI) topics. These were low tech, voice-over Powerpoint slideshow presentations, with post-production with iMovie on my Mac. I posted the videos to supplement videos by other people that I was using for my Fall 2012 AI course -- mostly videos by Thrun and Norvig, Daphne Koller, and a few others. My videos often adapted the slides accompanying the free textbook I use by Poole and Mackworth (http://artint.info/html/ArtInt.html) under a Creative Commons Attribution-Noncommercial-Share Alike 2.5 license.

Shortly into the Fall semester,  comments started being left by students of a UC Berkeley AI MOOC that was being offered through edX (https://www.edx.org/). The first views were for a constraint satisfaction video, then later my biggest hit of all, for a video on iterative deepening depth first search! After getting favorable comments for the first of these by students who commented that my video had helped them with the Berkeley MOOC, I went to the MOOC's discussion board and verified that students had left a link there, which had been endorsed by others.

Since then I've posted a few other videos for my database course in Spring 2013. The B+ tree videos are a "hit", though still small potatoes by MOOC standards.

I haven't been able to produce many videos, but I have been getting a high proportion of likes and followers. It's neat.

This experience has made me conscious that there are many resources that students access outside the recognized boundaries of a course, be it online or a good old-fashioned face-to-face class. I can remember finding some astoundingly good videos, each with many thousands of views, mainly in math, produced by "amateurs", including high school and college students, as well as faculty members at various places. I remember watching quite a few good calculus videos one day in 2008 from the Arlington library. The experience was eye-opening -- there were great educational resources on the Web that students were using, and I wanted to add to these resources! There are sites like TeachingTree (http://www.teachingtree.co/), that co-locate educational video resources of based on shared concepts, allowing "the crowd" to vet quality.

I'm most intrigued by the ad hoc community that grows around a course, notably a MOOC. This isn't a stable long-term community like the collection of my Facebook friends, or a Facebook group. Rather, the participants come together because of the MOOC, participating in a shared activity. Moreover -- and this is the most interesting part, I think -- the ad hoc community extends beyond the boundaries of the MOOC itself. By virtue of my videos, for example, I was part of the larger community of the Berkeley AI MOOC. (I've used the expression "ad hoc" communities because the communities that grow around a MOOC seem analogous to what Larry Barsalou called "ad hoc concepts/categories")

I like the metaphor of a solar system, with the MOOC as a center of gravity, and other resources, like YouTube videos, brought into orbit around the MOOC. I could actually make this a truer metaphor by very strategically examining the MOOC (e.g., the Berkeley AI MOOC and/or the Udacity AI MOOC that I just signed up for), with an eye towards identifying those sections of the MOOC(s) that I am sure that some students will have trouble with or those sections that students actually identify as difficult on the MOOC discussion board, and producing remedial videos on those topics. Such strategically-produced videos would be very much in the MOOC's orbit, left there very deliberately by me.

Longitudinal studies for understanding the nature, extent, and evolution of ad hoc communities, typically MOOC centered but not restricted to a single MOOC, undoubtedly require data mining across larger spheres of Web interactions than is currently easy to do (for anyone other than Google and a few others, that is). Nonetheless, the possibilities for understanding and leveraging student patterns in seeking remedial and advanced material, instructor incentives for creating and posting material, and the movement of people between student and teacher roles, are exciting and probably within current technical abilities – if only the data could be accessed. Currently, data mining of student behavior within a MOOC is the norm, but behavior that extends outside the MOOC isn't tracked.

Sunday, May 26, 2013

Better Preparation through MOOCs

I signed up for Udacity's online course in artificial intelligence (AI) at https://www.udacity.com/course/cs271. I hope that I do well! :-)

I used many of the online lectures by the same instructors, Thrun and Norvig, as well as lectures by others (including me!), for my AI class in Fall 2012. And the student ratings of the course were good (4.25/5.0 instructor rating and 4.0/5.0 course rating), holding steady relative to my usual offering of the course.

I signed up for Udacity's AI course to preview it for possible inclusion in my Fall 2013, and not just inclusion of the lectures, but also the quizzes and exams, and perhaps even the programming assignments. My reading of Udacity's current Terms of Service (ToS) is that it does NOT seem to preclude such use, which is a change relative to previous ToSs for Udacity and others (e.g., COURSERA) that I recall, which required written approval for such use (i.e., for student work in the MOOC to be a requirement of an on-campus, for-tuition course). For example, I obtained such written approval for a previous course in Machine Learning (Fall 2012) through COURSERA (https://my.vanderbilt.edu/cs390fall2012/). In any case, I'll be taking a closer look at the ToS this Summer, making sure that if I use the Udacity AI course as part of my course, my students won't be in violation of the ToS. (Parenthetically, if an instructor requires students to complete MOOC activities as part of an on-campus course, its the students who will be in violation of some of the ToSs that I have seen, not the instructor!!! Or so that's how I have read those ToSs).

Regardless of whether I use the actual Udacity course as part of my Fall 2013 AI course, however, the content of the two courses is similar enough that I can give my Fall 2013 students a big head-start, should they want it, by suggesting that they take the Udacity course IN ADVANCE of my AI course. If some do that, I imagine that they will be well prepared to ace my course. Will I worry about grade inflation in that case? Not really, because it won't be grade inflation at all, but rather better preparation, plain and simple!! Used in this way, MOOCs might be a great leveling mechanism, raising the performance of students who are traditionally a bit slower, perhaps only because they are over worked in other areas or otherwise distracted (as an aside, listen to this marvelous talk on the increasing distractions of life for students today and all of us: The Paradox of Choice http://www.youtube.com/watch?v=VO6XEQIsCoM -- listen for the remarkable sentiment regarding student workload starting at about 5:38 min into this video).

I have the luxury of suggesting that the students enrolled in my Fall 2013 AI course take the Udacity course this summer, because the Udacity model of a MOOC is completely self-paced -- students can start at any time and progress at their own pace. Other models (e.g., COURSERA's) have a window of time in which a course should be started, with due dates (really windows of time too) for assignments, quizzes and exams. Both models have pros and cons, which I won't get into now, but one pro of the Udacity model is that I can suggest that students take the course the summer before they will take my course in Fall -- perfect timing!) I can well imagine hybrids of these models, for example by staggering start dates, but thats for another post.

My educated guess is that if I suggest that students take the Udacity course in advance of my AI course, very few, if any, will get very far. In part, this probably would stem from a number of factors -- their confidence that they can get an A in any case; they are busy with jobs this summer; if you can do the thing tomorrow, then why do it today. And this last point is one of the possible cons of the completely self-paced model. I have some experience in this area too. In 1998, Vanderbilt funded my development of an online, completely self-paced version of Vanderbilt's AI course, to include auto grading of computer programs (in the programming language Lisp) and exams (where answers to questions were given as Lisp expressions). The idea was that students, hard-pressed in those days too, could take the AI course at their own pace over Summer and then "test out" of the offering the immediately following Fall. It was a neat idea, and I still have all the Perl scripts, mostly written as I recall by one of my PhD students, Lewis Frey, but only one student finished the AI course in that way, and the option slipped away quickly -- it was an idea that was ahead of its time, even for me -- I didn't see the full implications, or anything close to the full implications of online education, or if I did understand more than I remember now, I am only in the last few years rediscovering those insights, triggered by efforts of others.

Suffice it to say though, that self pacing is really difficult, particularly in a busy world.

So, I am thinking about offering students other incentives to take the Udacity course over the summer in preparation for my course -- perhaps an option to "test out" of (much of) my course? Perhaps an option to "TA" or "tutor" in many of the overlapping areas of my course? Perhaps a "letter of recommendation" at the end of my course? Perhaps some combination? Thoughts are welcome!

Friday, May 24, 2013

Finding the Best Teaching Assistant in the World

I've received several recommendations for this article in the New Yorker (http://www.newyorker.com/reporting/2013/05/20/130520fa_fact_heller), and while its long, it is worth the read if you are interested in online, higher education. Undoubtedly, I'll return to the article from time to time.

This article alludes to the problem of staffing a course with hundreds, thousands, even tens-of-thousands of students from across the globe -- where do you find enough teaching assistants, tutors, and graders! In the case of Gregory Nagy's course on Concepts of the Hero in Classical Greek Civilization,  former instructors and former students (alums) of the traditional, on-campus, Harvard course are asked to help out (http://kleos.chs.harvard.edu/?p=985), presumably as facilitators on the online courses discussion boards, but you can imagine alums in other staff roles, such as tutors and graders. The idea of using alums to actively participate in the educational mission of the university is intriguing!

Another model that has been suggested is to use the best students in a prior offering of a massive open online course (MOOC) as the TAs, tutors, and/or graders of a subsequent offering of the MOOC. If you have 10,000 students who have completed an earlier offering, then take the best performing 500 (5%) and ask them to be graders for the next offering. Not all of them will take you up on this invitation, and the number of acceptances will probably depend critically on the job you ask them to do (grader, tutor, facilitator), but calling upon this human resource of MOOC alums can be an important component in MOOC (re)design.

So, there are examples of using alums from traditional on-campus courses to staff content-similar online courses, and there are examples of using alums of online courses to staff later versions of those same courses -- what's left? As a faculty member at Vanderbilt, I want the option of staffing my on-campus courses with alums of content-similar online courses. 

The database course that I teach using Jennifer Widom's online lectures is a good example (https://my.vanderbilt.edu/cs265/). This course typically has about 30 students and I'm ok with one TA and one grader, if they know their stuff. But what if I could choose from among the very best students in Jennifer Widom's MOOCs on database as my TA(s) and grader(s)? Granted that as a computer science course, there is a lot of auto grading that can happen (and does happen) in a MOOC, which I can and will leverage in my on-campus offering of database, but I also want to consider the utility of human resources from MOOCs -- course alums -- in the design of my on-campus courses.

There are policy, social, and technical hindrances to using people unaffiliated with Vanderbilt as teaching assistants. Ones that I imagine offhand are

  1. privacy concerns -- before opening student work up to "outsiders", student identities would have to be protected
  2. and what of TA identities and accountability?
  3. what of compensation?

You might ask, who would want to be a grader for a course, particularly if they weren't compensated (financially)? My educated guess is that someone would want to be, those someone(s) would be very good, and a "someone" is all I need for my course of 30 students (though 2-3 would be ideal!).

What would they receive as a result of assisting in my course? They would receive a letter of recommendation from me. If there are many students who are willing to work for a relatively generic certificate at completing a MOOC, I am sure that some of the best performers, who are relevantly motivated, perhaps towards education, would happily choose to further reinforce their skills in the area of the MOOC, as my teaching assistant, for a highly personalized letter of recommendation.

I imagine there there are kinks in this approach -- kinks of policy, of culture, of a technical nature -- but the thought of finding the best TA in the world for the benefit of my on-campus students is too exciting to let go of without a lot more thought.

You might also wonder, why go the the cloud for teaching assistants -- if "its not broke, don't fix it." But sometimes the current system is broken -- some TAs don't know the material for an advanced class well enough to be an effective TA, and aren't necessarily motivated to learn at a rate that stays ahead of the students (in part because these TAs are themselves students, who are also taking classes, and who are being encouraged to get out of TAing asap and start research assistantships). And of course, some courses at some institutions won't be staffed at all under the current system, because of a graduate student shortfall. In short, there are kinks in the current, conventional approach too.

Thursday, May 23, 2013

More Bang for the Buck

The onset of highly visible, massive open online courses (MOOCs) has had much to do with re-energizing my teaching, something I posted about as a guest on the ProfHacker blog of the Chronicle of Higher Education (http://bit.ly/10VFZbZ). My interests so far have been much more about using online educational material to improve the quality (along many dimensions) of on-campus education for students through online education, than it has been about decreasing costs of higher education, though I am very concerned with that as well. In the inaugural post of Cloud and Campus, I reprint much of a letter to the editor of the Vanderbilt student newspaper (http://bit.ly/10pF6bD), suggesting to Vanderbilt's current students that even if they won't benefit (directly, immediately) from reduced costs, they can benefit from increased quality.

The following is extracted from http://bit.ly/10pF6bD


I expect that most students know about the boom in online education, particularly massively open online courses (MOOCs). An informal survey of students attending the McGill Hour (http://bit.ly/188lrog) I recently led suggests that quite a few have actually taken or “audited” some of these courses. I opened the McGill Hour with a story of a graduating Vanderbilt senior, in a non-computer science field, who was wandering the halls of the computer science offices looking at bulletin boards the week before last May’s graduation. The long hallway was completely empty except for him and me, and he was clearly looking for something; I asked if I could help. As best as he could articulate, he was looking for some opportunity to learn computer programming in the very near future. I hesitated a little and then suggested that he take one of the excellent and free computer programming courses that were springing up online, offered by other universities. I wanted to tell him, a soon-to-be alum, that such a course was being offered by Vanderbilt, and moreover, that as an alum he could have access to Vanderbilt-produced material throughout his lifetime, perhaps with affordances that others taking Vanderbilt-produced content would not have.


There is much talk of how MOOCs and their descendants will lower the cost of higher education, and that is vitally important, but I am not expecting to see that change in the near future. Nonetheless, there are other ways of increasing “bang-for-the-buck” rather than lowering the buck — and that is by increasing the bang.


I sincerely hope that increasing the “bang” will include the establishment of online, lifelong learning opportunities for our alums, changing the very nature of what it is to be an alum. When I hear from my alma mater, it is with news of their latest and greatest, accompanied by a request for money, which is relatively easy for me to swallow because I paid a few hundred dollars a year for a first-rate education. But if my alma mater doesn’t start approaching me soon with low-cost learning opportunities, I’ll be surprised and disappointed. Moreover, a great treat as a faculty member is hearing from my former students regarding what they are doing professionally, as well as hearing about their families — and I happily reciprocate. However, I would absolutely love to engage in lifelong learning with “former” students, and not just in my offering online courses to them. I recently turned to a former Vanderbilt undergraduate student and current doctoral student at CMU with whom I was sharing a stage and told her and the audience that in five years I wanted to take a MOOC from her!

There are other ways of increasing “bang” that I can provide, even at my station. For example, I can tell a student, “Yes, I will allow you to take the graduate course in artificial intelligence, even without the undergrad course, because you did well in that free online artificial intelligence course over the summer.” In general, I feel comfortable allowing some of these online courses to stand in place of selected prerequisites of courses that I teach, even if I have no power to grant formal university credit for such courses (that latter decision is certainly above my pay grade) but the question of satisfaction of prerequisites is often at my discretion.

Another opportunity for increasing bang is about to begin, as Vanderbilt’s course offerings come online through Coursera. Courses by Douglas Schmidt and David Owens start March 4, with others to follow. Vanderbilt’s Center for Teaching hopes to facilitate local learning communities around these online offerings, to include Vanderbilt students, staff and faculty. While I’ll see my colleagues on the big screen, I will be in the audience, learning new material side by side with students. Most faculty members are learning new things all the time of course, most obviously through their research and through teaching courses for the first time, but I am excited about being embedded in the learning community, modeling lifelong learning — or so I hope. Online learning may bring a sea change to on-campus education culture, again increasing bang for the buck.

I think much creativity stems from dealing with discomfort, and higher education costs must be contributing to the substantial discomfort of many students and families, in spite of some commendable efforts made in good faith by Vanderbilt to ease financial burdens. Still, I hope that students, alums and their families are all active in ideating on what Vanderbilt can do in the way of increasing the educational bang for the buck, both by leveraging online learning and in other ways.