Sunday, July 12, 2009

Hollywood-Style Lectures

Teaching is hard, and teaching well is REALLY hard. If you’ve never done it before, you have no idea how many hours of preparation each lecture takes. The trick to counter that is to teach the same class over and over. By the n-th time you teach it, you can get away with much less preparation, but even then some of the lectures truly suck.

I teach the same class every other semester: Great Theoretical Ideas in Computer Science. If you’re a CMU undergrad, you surely have strong memories about this class (positive or negative). If you’re not, all you need to know is that it’s a discrete math course required for all computer science majors and usually has about 200 students in it. When I spend millions of hours preparing for the class, it ends up being pretty good (teaching awards, very high student evaluations, etc.) Unfortunately, some weeks I don’t have millions of hours to spend on it so the lectures are not as good as I would want them to be (and the students fall asleep!). But let’s all be honest here: even when I prepare a lot, the lectures are not all that great. I make mistakes, I forget to say some things, my handwriting is bad, my jokes fall flat, etc. Every semester there are maybe 3-4 lectures that I am happy with afterwards, and of the rest about 50% totally suck in my mind and 50% are just barely passable. The fact that I am considered one of the better teachers of the department is, truthfully, sad.

So, a good fraction of my lectures totally suck. I am also quite tired of repeating almost the same thing over and over every semester (and what sucks more is that sometimes it comes out great and sometimes it doesn’t!). So here’s my proposal: instead of my amateurish attempts at making good lectures that fail most of the time, and instead of repeating the same crap every semester like a broken record, why don’t I just produce really good video lectures?

Now, I know what you will say: “Video lectures suck! They tend to put the students to sleep even more than real life professors, the audio quality is poor, you can’t see the board, etc.” And I agree. There is just something about being there in real life that cannot be captured by a video and this makes recorded lectures be even crappier than their real life counterparts.

But I’m not talking about simply recording myself giving a regular lecture. I think that would suck. I am talking about making a high production value movie for every lecture. I’m talking about professional script writers (those guys that make the Daily Show or the Colbert Report so funny), about special effects and computer graphics to illustrate the concepts instead of the board, about high end directors, cameramen, and producers (like the guys who made my Nova special).

I’ve spoken with some of my friends in the movie industry (writers, producers and directors of Hollywood blockbusters), and they all seem quite excited about trying to do this. The biggest problem seems to be the cost. It’s hard to estimate how much each lecture would cost, and it clearly depends on how much quality you want, but it seems each lecture can be done for between $75k to $300k. If we make 30 lectures to cover the whole semester, that amounts to something between $2 million and $9 million. This type of investment is probably not worth it for higher-level classes that are only taken by a few people each semester. But for a discrete math class taken by 200 students every semester at CMU alone (and tens of thousands of students throughout the world every year), I think it’s well worth it.



  1. Sounds great. I wish there was something like this when I was between high school and University and eciding what to study; having a sampler of first year university courses of very high quality would have been great.

    It seems unlikely you would be able to charge the students for the video lectures. In CS it doesnt seem far fetched to get a large corporation to sponsor them (for scale; sponsoring 10 differet courses like this would be equivalent to the marketing expenditures of the bing launch).

    I wonder about the macro effects if lectures like this where to become very popular. If everyone had seen the same intro to CS then two effects I can see arizing: (1) You can comunicate btter with your peers as there is more in the way of shared knowledge and representation. (2) You would get more homogenous perspectives on problems. the fact that students are exposed to some random subset of good and bad exposition problably biases them to learn more about some approaches than others, this creates a more diverse background when you have a group of students who took ifferent versions of essentially the same class. Ensambles of weak learners and all that.

  2. Indeed.

    I don't understand why we don't yet have more canned video lectures like this. (I have seen some "educations DVDs" for sale for they are for "IT education", basically, how to use MS Excell and stuff like that). The MIT opencoursware videos are also OK, but they are just one prof. talking in front of a blackboard.

    I see these online lectures as the modern equivalent of the textbook: something the teacher has the students buy and read/view on their own. At the beginning of the semester I, the teacher, would choose a textbook and set of video lectures which the students would be required to purchase. Then I would be free to focus the class on providing the students with one-on-one help as well as generating/grading tests and homeworks, which currently takes most of my time anyway.

    The only problem is that different schools follow different curricula: different topics at different time. But this should not be a problem if we can buy lectures a la-cart.

    I used vimeo for screencasting my lectures in HD and it works fine. HD resolution is more than enough. The students loved it because of the time-shifting: they could continue to sleep in. Of course, my lectures are crappy. A $300K video lecture would be much better, and if it the cost to the students is equivalent to that of a textbook it should not be much of a problem.

    Please, someone go to work on this!

  3. Here's some thoughts: Rather than producing something feature length, what if each major concept you would deliver in your lecture was a stand-alone skit. Each skit could then digestible (for students and online distribution) and upgradeable (to make refreshing changes such as if one joke "falls flat" you won't have to change the entire feature). For a feature-length, what if you had someone or yourself introduce the concepts then show the skits (think Chappelle Show for academics).

  4. Yes, that could be a better way to do it because it would make the skits interchangeable among lectures.

  5. I just reached to your blog,, and yesterday I was reading about investing 100 millions in filming text books or something like this!!

    I can say repeating same lectures every semester suck from personal experience :(

    So I think this is a good idea! heard before about teaching elementary students through games, to add excitement to the class.

    first of all this can be a very lucrative business for remote teaching, so that the university itself will have its own studios, producers, writers and so on. Or maybe for those boring teaching channels, this would be good for them and this could be for a code to buy so they can cover their expenses :)

    But in classroom, believe me nothing is like teacher-student interaction and regardless of your preparation some questions or gestures from students can bring with unpredicted enthusiasm a great deal of information, something will be built inside you, questions that are coming to your mind, and lots and lots of other benefits.

    another thing, who knows if the content will remain the same after all, something might add, applications, changes and those sort of things! So, maintenance cost should be considered too.

    At the end, looks good idea but it needs to be implemented in some situations..
    For sleepy students I think they need a cup of cold water :) or maybe a rain inside the class :D

  6. That's an interesting idea, but there's a couple things about it that bother me. First is that you lose the interactivity of a real lecture - students can't ask questions of a video screen. And there have been times when the answers to students' questions provide some key information that wasn't in the lecture, or are just interesting.

    Following on from that, if it turns out that some key information is missing from the video (which could happen no matter how much effort is put into producing them), then you're stuck with an expensive, incomplete video. You could change it, but that would take a lot more time and effort than changing the content of a traditional lecture.

    Second main thing that bothers me is the stigma (at least in my mind) of educational videos. I've always found them to be a bit cheesy, even if they're well-produced and informative. It feels like the video is talking down to the audience. That might be partly because educational videos are typically shown to little kids.

    Videos that are supplementary to real lectures seem like a better idea, since that gives students something they can refer back to on their own time, and the benefits of the real lecture aren't lost. It'd almost certainly be better than most classes' lecture slides.

    I think you'd also find that videos are inherently harder to pay attention to. Without a lecturer, there is no element of human contact in the lecture, which makes it much easier to ignore. No matter how well-produced the video is, I think, it's easier for students to tune it out than it would be for them to tune out a person who's physically there. And this is precisely because, as I mentioned before, with a video there can be no back-and-forth or exchange between the audience and the lecturer. Even leaving aside questions and answers, a human lecturer can watch the audience, can vary the pace according to the audience's need, can do a variety of things to rally the audience's attention if need be. There's a lot of nonverbal exchange between a lecturer and an audience, which holds the attention of both (try giving a lecture to no audience but a video camera; it's much harder than having a human audience) and, I believe, leads to much better information retention by the audience.

  7. Great Idea!
    And for the advance courses may be you can have the simple low budget video lectures.
    Another reason why video lectures are good is that you always have the freedom to pause/rewind. ;)

  8. This is a terrific idea. Instead of paying professional people to do this, why not let arts/drama/design students do this by making it as some sort of senior project for them?

  9. Interesting. My gut reaction is that even a Hollywood-style recorded lecture would suck, but it's difficult to figure out why. Well, here are more-subtle kinds of interactivity that can exist in a live lecture: 1. You can change the pace of the lecture based on non-verbal feedback from students. (Maybe this isn't possible in 200-person classes; I've taught only small classes.) 2. Even if #1 isn't possible, I do think that non-verbal cues, e.g., your excitement about the lecture, help maintain student interest. 3. When you forget something or screw up, the students get to see your thought process as you fix it in real time. I think this is actually really valuable, and underappreciated. If you wanted everything slick and perfect, why not just read the book?

  10. Another thought: Once you get to that level of production values, why do YOU need to be the one reading the script? You would certainly need to supervise the scriptwriting, but you may as well get professional voice talent at that point.

  11. Was just thinking about something like this the other day. Virginia and I are currently teaching lectures at Rutgers that introduce CS Theory to NJ public high school students. (It's based on 15-251/Andrew's Leap, but with a bunch of stuff cut out, and the ideas get emphasized more than details.) The kids are very excited about the material.

    It would be awesome if we could produce quality videos of these lectures (polished, with better jokes, etc.) and offer them essentially for free.

    I would bet that a well-executed program would increase future enrollment in CS courses in colleges. Just exposing the kids to the ideas seems already effective. Most of our lectures are not "real" college lectures, but kind of a mix between "popular talks" and 251 lectures. It might be much easier to turn this into a Hollywood production than the technical material. Also it should not upset universities to distribute these, because the lectures aren't suited for undergrads anyway. (Rather, the colleges should get more applications from prospective CS majors, after many high schoolers watch them.)

  12. I really like the "Chappelle Show" idea. 2-3 ~10 minute sketches in an hour long lecture could really spice up the ideas. Hopefully it would cut down on prep-time since you'll only have to prep for introducing concepts and answering questions.

    Also, it may make it easier to convince the university to subsidize the production. If you made ~2 lectures worth of skits. Then check the grades on questions related to those topics relative to the "standard" lectures you could get quantitative evidence to support your claim. You could couple that with surveys from students about their preference for each lecture type.

    I'm not if CMU has a film/graphics design (or its quality) but you might be able to swing a few students to produce the initial skits as a senior design project.

  13. The yale online lectures - - are great. Audio quality is fantastic, resolution is high, you can see the board perfectly, subtitles can be overlayed on the lecture, and a transcript of the lecture is available for you. I also doubt they cost very much to produce. Sounds like you overestimate the costs.

  14. This is a great idea.. I just imagined camera zooming into a section of the blackboard with a sound effect for a point that needs emphasis or parts of a math formula lighting up in different colors as the professor talks about them. Bullet time recording. It would be great.

  15. I'm moving towards creating online lectures for some topics in my War, State and Society option, and I'm coming round to thinking in terms of making 10 min short podcasts rather than 1 hour lectures. it is easier to replace a 10 minute segment than a hour, and since students often only 'get' half to a third of the content of a one hour lecture; 3 ten minute segments are probably as useful as an hour of video.

    I think 75k to 300k is an overestimate; I figure I need a day to do a good ten minute segment, mixing video and powerpoints into a flash movie using something like Captivate.

  16. Another thought: Once you get to that level of production values, why do YOU need to be the one reading the script? You would certainly need to supervise the scriptwriting, but you may as well get professional voice talent at that point.

    Oh, I totally agree. Heck, the "professor" should be a professional actor that can be super charismatic and even good looking!

  17. A couple of you have mentioned something that I expected to hear as a criticism of my idea: interactivity. I agree that asking questions can be powerful, but I don't see that having recorded lectures prevents students from asking things. The professor's job under my scheme would be to sit there during the video presentation and answer any questions (either during the video or afterwards).

  18. I think 75k to 300k is an overestimate; I figure I need a day to do a good ten minute segment, mixing video and powerpoints into a flash movie using something like Captivate.

    No, 75k-300k is actually pretty precise estimate for the type of production I'm talking about (which would not take a day to do 10 minutes of, unless you had a huge crew). I'm not talking about a shabby recording of a standard lecture. I'm talking about writing a script that is truly funny (gotta pay the professional writers for that), about animations, etc. A 10-minute vignete would probably take about a week if you had a 10-person crew.

  19. I like the idea. But wouldn't it be easier if you just made vanilla lectures videos (that are cheap) and spent the countless hours answering stuff that students didn't get. For example, have a website with questions that students should be able to answer after watching a video on say recursion. A pdf of student response to different questions will allow you to work on concepts that clearly didn't get through in class in real time.

  20. Well the interaction goes back and forth, and as I think it might change the direction of the lecture!
    Just to get along with the idea, can you give an example, so the point will be clearer :) one mentioned Zoom, board, formula's .. and I keep thinking about this as a context about how we got to that point!
    would you?

  21. From my personal experience, one of the reasons you win teaching awards is because of the exceptional job you do during office hours, as well as the exceptional questions that come from the students at CMU. Seeing the thought process of your peers as you learn the material and solve those ridiculously murderous problems trains your mind to try those previously-trodden and succesful paths in future problems that show similarities. By creating a movie, you lose the most important aspect of the classroom: student-professor interation, and student-student interaction (TAs I'm lumping you into students).

    A movie also allows a student to get a general understanding of the subject material as it is presented by the professor. However, judging from general student opinion about 251, a movie would not be any more useful than the slides. Students study and scrape every detail from the slides and I don't believe a video just on a lecture would be more beneficial than a set a slides with a professor they can interact with.

    In addition, a movie, no matter how professionally done, can never replace the soul-crushing homework in 251.

  22. You speak almost as though money were the only problem, as if all you need is an extra $3 million (which, by the way, is a whole lot of money!). But in reality, money only solves the easy problems. The main difficulties of the format remain.

    There are, for example, several institutes out there that record all of their talks professionally, including a separate video for the slides, if any. And yet, in my opinion, they work terribly. Professional scriptwriters would not help.

    The problems with money can be avoided by using some creativity. The necessary equipment, camera, mikes, lighting, etc., is not so expensive. The people are expensive. Everyone in Hollywood is part of a union. But at a university, you have slave labor, I mean graduate students. Hand out free ramen packets and they'll be jumping over each other to help out.

  23. Are you familiar with The Teaching Company -- They make lecture series DVDs with nice graphics and good production values. The material, however, is straight out of the professors' lesson plans, it hasn't been Hollywood-ified and sugar-coated like you suggest. So it's a lot dryer than watching The Colbert Report.

    What you're describing is edutainment. Material that's been broken into small chunks for short attention spans and interwoven with humor. I'm not sure that there's a lot of precedent for edutainment in higher education, although many successful professors probably aspire to that. TED talks come to mind, as do the "Cartoon Guide to Genetics" style books.

  24. By the way, I think the money is a solvable problem. The existence of the teaching company proves that there's a market for high-quality adult educational video. Their production costs are surely a lot lower, but their material is also a lot less accessible than what you've proposed, so their market is smaller.

    For some fields, I could imagine selling corporate subscriptions to this video. Companies spend a lot of money on employee education. $50k to have access to incredible teaching material from the world's best professors is not unreasonable.

    Also parents in the US have shown that they will spend enormous amounts of money to give their kids an educational edge.

    I don't know whether an ad-supported model would work. Probably not.

  25. I like the idea. But wouldn't it be easier if you just made vanilla lectures videos (that are cheap) and spent the countless hours answering stuff that students didn't get. For example, have a website with questions that students should be able to answer after watching a video on say recursion. A pdf of student response to different questions will allow you to work on concepts that clearly didn't get through in class in real time.

    Sure, that can be done the semester after we're finished with the videos :) Part of the beauty of this scheme is that the professor would have a lot more time to work on things like this.

  26. Have you seen Walter Lewin's videos? Not sure if the links at the bottom of that page are the videos I used to watch on MIT Cable but he is very well known for his intro physics lectures. Don't know what the production costs are but they've surely amortized well over time.

  27. I don't think video lectures are a great long term investment unless you're talking about remote students. In a live environment I think it would be much more beneficial to do something along the lines of a training video. Each lecture, instead of recording a lecture you record an instructional video on presenting that lecture. Each class you rotate out a student to watch the video ahead of time and present the material. I think this would do a lot more for interaction and retention.

    Also I think in either case the videos should be modularized so that regular updates to keep the videos relevant are easy. That way you don't have to start over completely each time you want to throw something currently relevant in to a part of the lecture.

  28. I honestly really like the video lectures offered by the likes of MIT OCW, Open Yale Courses, Stanford on iTunesU, and webcase@berkeley. Those programs are great. And as a CMU student, I feel ashamed that our school does not have a similar service. I'm a CS major at CMU, so I took 251 and concepts here. They were great classes, but I didn't have as much available to me as I did when I took 451 here. I kept up with Prof. Blum's lectures, but I also used the 6.046J lectures they had on MIT OCW. A lot of times the two complemented one another very well, and really helped me learn the material. Those lectures didn't have any Hollywood effects, yet they worked very well.

    If you abandoned your slides, started teach math on the chalkboard, and released video lectures of 251 online you would be provided a much needed service to the people on the web trying to learn discrete math. In the process you would also be helping CMU positively advertise itself; which we all know is something that this university desperately needs.

    What I would really like to see Mackey's Concepts and your 251 lectures online as part of a new CMU video lecture web service.

  29. Hmmm.. It seems to me that the goal of the project discussed here needs to be changed. I don't think teachers of any kind should talk about a project unless the goal stated is clear: to improve the quality of the education the students receive, not so much the enjoyment of the prof. :)

    I like the idea of having a video available mostly for a review/study standpoint. Maybe you could just start recording your lectures with the help of some A/V oriented students. Then over time you could edit to produce a master that has the best explanations, the clearest examples, and maybe some great questions from the students.

    I don't think the prof should ever just sit there playing Tetris through the whole class, but some excerpts from the "master video" followed by discussion could be a valuable tool. Like VH1's "What and Discuss" ;)

    At any rate I still think it should be in Neon that the goal is to improve the education and experience of the students... the prof's life should come second to them. :)

  30. I'd like to follow up on Emil Sit's comment about Walter Lewin. I watched this lecture as an MIT undergrad, and I think it is still one of the best lectures I have ever seen (though I remember it being higher fidelity at the time).

    I think you should just go for the low-budget option, for several reasons:

    1) It doesn't cost as much!

    2) It's more sustainable and generalizable -- others will be able to continue the tradition more easily.

    3) Even low-budget lectures can be great -- viz. the Walter Lewin lectures.

    4) Take a lesson from software engineering: get a quick-and-dirty version out early, learn what's wrong with it from actual feedback, and repeat. Why spend the money when, even if you do, there's a significant chance you won't get it right? Better to do 10 cheap revisions and get it great than 1 expensive one and get it wrong.

    On another note, these mini-lectures on category theory are a great example of how 10-minute modular lectures can be done well (including being able to see the blackboard).

  31. Why limit the lectures to you alone? Why limit this to CMU alone? Wouldn't a multi-institutional or even multi-national course syndicate be even more compelling?

  32. I totally agree that there is no need to limit things to me or CMU alone.

  33. The lectures of Richard Feynman at is a good example of how to go even beyond the standard "recorded lecture". Not only the lecture is recorded but there are also subtitles and commentary, you can see pointers to external resources appearing next to the screen when the lecturer talks about a specific topic, etc.

  34. Luis, there's no point being modest about your teaching quality; your slides are fantastic, and you complement them very well.

    I would also note that with your slides, you actually do come pretty close to a lecture that could be recorded and distributed. I know a couple years ago Scherlis and Langmead recorded their 211 lectures and posted them online (I think they were only available to course students). You could easily do something similar; record the lecture, put it in a video, and put the slides (synchronized) on the page next to the video.

  35. To the commenters concerned with the interaction: yes, it is not as good that the video lecture is not tailored to an individual student, but neither are live lectures. Research in tutoting systems shows that even observing others asking questions helps learning (vicarious learning).

  36. Luis: I just came across this and thought about you:

    Doesn't it corresponds to what you were describing?

  37. TED has the right mix for the lecture format, in my opinion. Anything longer or more verbose is better accomplished on an individual scale in some other format.

  38. With the advent of high quality graphics software, movie-making from the desktop shouldn't be too expensive, and is probably worth considering as an alternative to a big-budget production. It is of course, time-consuming (probably more so than a regular movie), but you could get a few grad students excited enough to do it.

  39. I think video lecture are a great idea, for the reasons you explained. A video and a good book would almost do the job.
    On the other hand I do NOT think that hollywood style lectures are needed. I think instead that for every lecture (or group of lectures) there should be some sort of review session or open discussion between professor or TA and all students in order to go over again the hard concetps and expecially answer the students question/comments.

  40. Well, we have a good drama school. They're slave labor; why not try them for a lower budget version? But, yeah, cinematic production values would be very cool. That said, I'm not sure that it wouldn't hurt interactivity. The special effects, etc seem like they would stretch things out a lot, leaving less time for discussion.

  41. Ralph Johnson at UIUC shows some of his lectures in class. The TA comes to answer interactive questions. Sounds weird at first, but I like his argument. This way he saves his teaching time and spends it on the homeworks. You present the homework to him in person and get to interact with him in a very small group (3-4 pepople).

    He does it only with the "perfect" lectures. The other ones he record over and over again.

    To connect it to your proposal - maybe you need to record the lectures you like. 3-4 lectures a semester and you'll be done soon.

  42. The special effects, etc seem like they would stretch things out a lot, leaving less time for discussion.
    Assignments | Dissertations | Essays

  43. movie-making from the desktop shouldn't be too expensive, and is probably worth considering as an alternative to a big-budget production. It is of course
    Theses | GCSE Coursework

  44. This has happened to some degree in India - over 1,300 academic lectures by Bollywood star Salman Khan are on YouTube. See for example his take on the Electron Transport Chain: