## Monday, November 13, 2006

The Chicago Tribune today had a few articles on Second Life, a growing virtual world that even has its own economic markets with its own currency exchangable with US dollars. As the article says, Harvard Law School taught a class in Second Life and I have heard of many other universities in the process of establishing Second Life courses as part of their on-line degrees. Taken to an extreme, why do we need hundreds of graduate complexity courses taught world-wide every year when everyone can sit in on one of a handful of the best lecturers giving the class?

Indeed we now have the technology to have virtual seminars or even entire conferences on-line complete with "coffee breaks", business meetings and dance parties. Why not even hold an established conference, like STOC or SODA in a virtual world? The total cost would be much less than traveling to a real-world meeting and nearly every aspect of the conference experience could be simulated.

One advantage of a real-world conference is not so much what one can do but what a real-world conference prevents you from doing. While away at STOC you can't teach your course, attend committee meetings, hold office hours, meet with students, etc. You are forced by circumstance to reschedule these activities and open up your calendar to see talks and meet with your colleagues. But at a virtual meeting, can you tell your chair you have to miss your class and the faculty meeting while you sit at your computer, your body in your office but your mind in a different place?

1. There is a somewhat related post at cosmic variance

2. If virtual meetings were going to replace real ones in the foreseeable future, then I think they would have done so already. We've had the technology since at least the early 90's, yet I've seen no movement whatsoever in that direction.

Face it, Lance: as much as people gripe about it, they like flying to conferences all over the world.

3. Taken to an extreme, why do we need hundreds of graduate complexity courses taught world-wide every year when everyone can sit in on one of a handful of the best lecturers giving the class?

As is often the case, Lance's post hits on two issues simultaneously; I'm surprised nobody has latched on to this one. It's a great question, and I personally am not sure what the great answer is. I believe face-to-face learning is generally much better than distance-education. Is that a good enough answer?

Scott said:

Face it, Lance: as much as people gripe about it, they like flying to conferences all over the world.

As Lance points out, there are certainly advantages to face-to-face conference meetings. But there are so many disadvantages, specifically (these days) the flying. I disagree with Scott -- I think this hasn't been done yet because (1) the technical overhead is higher than perceived (but headed downward quickly) (2) nobody's ever done things this way, so there's a lot of intertia and (3) the risk of a flop seems high, so nobody wants to be the first.

But I'd love it if somebody tried this. Sign me up!

4. Here is one reason that non-virtual lectures and classes can hold their own: In a class with less than 100 people (and usually less than 50) the lecturer reacts to the audience, making each talk somewhat a custom made one. Virtual lectures as described here will have 1-way communication, and so will have a TV-show essence and no dialogue. Indeed, why "send your video-cassette" conferences were never tried? That technology for virtual academics was already with us for quite a while.

5. Virtual talk is a first step of virtual conferences. Virtual talks are somewhat being done. For an example, Microsoft Research, make most of the talks available to everybody with a broadband connection.

Even Google does that too (But unfortunately does not support the community attempt.) Everybody of us could also post their talks on the web.

Microsoft Research and many other companies and universities use researchchannel.com to post their talks. The advantage of researchchannel.com over proprietary solution such as Google video is two fold:

1. The UI of researchchannel.com is more appropriate for academic talks.

2. Researchchannel.com is a more cooperative solution. It has support of many universities. (Luca and others, you may want your university to participate in researchchannel.com.)

Note: The commentator is a Microsoft employee and used this opportunity to point out one goodness of the employer, which is aligned with the interest of the community of this blog.

6. Eldar,

I don't understand your comment. Lance talks about classes on Second Life, and conference with virtual coffee breaks and business meetings. He is talking about having full 2-way interactive communication, which is pretty much entirely possible (albeit I still think not entirely easy) with today's technology.

I agree that "class-by-videotape" loses something. But what about 2-way interactive video lectures (either real-time video, or slides with your own personal avatar). Why not hold a conference that way? (Like a real conference, most questions would come at the end -- the 2-way interactions could be fairly simple, and side-chats could occur among participants.) And if there is an appropriate question-asking medium, Lance's question on the need for hundreds of graduate complexity courses remains challenging.

7. I agree that face-to-face lectures are generally better than distant ones.

However, a distant lecture by a good lecturer and/or an expert in the subject is MUCH better than a face-to-face lecture by a less qualified instructor.

8. Michael,

Yes, you can add some biderectionality kludges on top of your lecture, but it will be like trying to pet a kitten while wearing oven mittens (I'm going for a kid-friendly analogy here). Having a much larger "class" than in non-virtual talks won't help either (as well as not being a witness to in-class interactions, or other types of "3rd party communication").

As for "virtual coffee breaks", what is that? Am I already too old to understand?

9. There is a complexity-theoretic answer (assuming P|= PSPACE) to why "distance learning" classes are inferior to traditional ones, one that Lance is QUITE familiar with: IP=PSPACE.

The problem is not as much "interaction at a distance" (although
most interaction-providing software is still kludgy) but size of the audience is. Typical complexity theory classes are small, students can ask questions, and the instructor can taylor explanations to the level of understanding of the audience. This is much harder to do in a virtual class.

My favorite complaint about beamed lectures is that in an actual lecture my eyes wander from the speaker to the slide, (and, in transparency projectors, to the slide on the second projector that has the definitions I had already forgotten.) Most software sytems provide a single image. I have used a system of "smart rooms" between Chicago and Argonne that has several images in each direction, and it was much better.

Still, a good class has between half a dozen and two dozen students in it: this allows real interaction. Perhaps, with good gadgets, such a course could be taught in a location-independent way, but would not substitute the hundreds of courses.

I could imagine a course offered in the Oxford model: a "best lecturer" lecturing, and small discussion groups led by a local expert.

There is a selfish reason for teaching hundreds of courses. It is a great way for the instructor to learn. This is not a minor consideration: forcing yourslef to understand in minute detail new proofs, and thinking about what results are important is a great way to keep up to date, and to think about good research problems.

10. Some models to consider:

1) SOME people GOTO the conference but
grants) get to go virtually'' for a
eventually most people not showing up and
THEN its all virtual.

2) Classrooms: A Complexity Course would
be hard to have virtual- not enough students,
topics change alot and not standarized.
BUT a course like CALCULUS has major
potential to be online in some form
(and already is at some schools- I'm
sure its on line at the online school
whose name I dont' recall)

3) MY HOPE: As virtual conferencs and other ways of BEING THERE without
being there evolve, flying will diminish
so much that the airlines will HAVE
to provide better service at cheaper
prices and (more importantly) less
hassle. However, Note that airlines
(it seems like they are often in a state
of bankrupcy but still flying)

bill g

11. Pretty soon, people will talk about outsourcing
professors' job to foreign
countries, we will all
lose our jobs:-(

12. I like Bill's idea of allowing "virtual access" to the talks at a conference. There are plenty of reasons other than attending talks to go to a conference (socializing,...) but sometimes one just cannot afford the time/money etc and may still want to see the talks (to understand a particular paper better, for example, or see an invited lecture). It would be nice to have a recording of the talks in this case.

A few years back a group at CMU experimented with multicasting the sigcomm conference live and I heard that it went pretty well. (Not sure why they decided not to continue this; perhaps a lack of volunteers.) Of course this would end up costing the conference and we might see reduced attendence. (I don't think the latter would be significant.) But if people were charged a nominal amount, say \$5 for unlimited access to all talks at that conference, I would bet there would be sufficient interest to make this viable.

13. As an undergraduate, in many humanities courses the professor would deliver a lecture and then there would be some time for questions at the end. Now often these lectures were the same that had been given for years. This is not to say that the delivery was dry, just that the professor had delivered more or less this exact lecture more than ten times before.

Lets face it, besides the good feeling that the professor gets from holding the attention of so many students, another equally good presenter that has memorized the material, but has no idea what it means. The only time he is using his knowledge is the last five-ten minutes for questions.

Maybe this is the answer. Not to replace professors with technology, but with actors. These actors can liven up the lectures, and then the actual professor can come in at the end for 5 minutes to answer questions, and stay after class and talk to the students more.

Let’s look at another extreme. At Oxford students often have one-on-one, or two-on-one weekly meetings with teachers. Now, for certain things, this is great. However, it is often overkill, and a waste of the teacher’s time (they need to spend a lot of time tutoring students).

My point is that it is important to maximize the use of professor’s time. Maybe their time is best spent preparing lectures and delivering them. But maybe not. I think that a lot of the on-line Universities try very hard to not use professor’s time when it is not necessary, but then to provide more feedback (than a normal university) when it is more beneficial. The question is not, can technology help us skimp; but rather, can professors better help students to learn than they currently do.

Grant S

14. Two reasons why virtual meetings have not replaced the real ones:
1. (Specific to some set of universities/labs which include Waterloo) The density of real seminars, courses, etc. is high enough to consume close to 100% of the attention/time for talks that people have. As a result, virtual meetings are hardly ever explored.

If virtual meetings take off, I expect it to start from institutions/fields with lower density of real talks.

2. Inertia is a major factor. Some wireless executive once said that the speed of technology adoption is now primarily determined by customer readiness, rather than technological possibility. Weblogs were technologically possible at least 5 years before they really took off and virtual meetings are much bigger change than weblogs were.

Andris

15. Interactivity is essential in a teaching setting. However, what proportion of conference talks have you attended where you asked a question or there was significant interactivity?

I second Kamal's recommendation for ResearchChannel. We do all our colloquia through them. The quality of the experience is actually very good. It's kind of like being stuck in the overflow room when a popular lecturer comes to campus - you lose many of the human cues that draw your attention but if you really are interested you find it very worthwhile.

My ideal for a virtual talk would be:
* high quality large projected slides
* audio
* low quality small talking head to cue that this is a real human being doing the talking.

I'd go to conferences anyway but in this format it would be great even to review conference talks I had attended (or, even better, the talks I couldn't attend in the other parallel session!)

A big problem for conferences still is cost. A/V already is a surprising high cost to conferences. To produce good virtual talks still requires expensive production or infrastructure, even with the changes in technology, and would multiply those costs several times over.

16. In response to Paul Beame's comment: I would want more from a virtual talk--and, as technology becomes cheaper, we should be able to get it:

1)Large screen with high quality slides
2)A decent video of the speaker (not only a talking head--many good speakers have very expressive gestures and it's worth seeing and hearing them
3)Wi-Fi access to the paper itself (or paper copy)--to look up the exact definition of something, perhaps not on the slides
4)Access (same wireless connection)
to the slides previously shown. This would be very useful even in current conferences. A poor substitute would be a downloadable copy of the slides, but speakers would not like this, and they would be right.