Monday, May 30, 2005

Conference Presentations

The quality of conference presentations have, on average, much improved over the past decade or two. Why? Certainly technological improvements like PowerPoint and advanced LaTeX macros have helped. As our field gets more specialized, talks in general theory conferences have to appeal to a wider audience which tend to improve the presentation. Or perhaps I'm just remembering only the bad talks from the good old days.

Despite the increase in quality, I find myself going to fewer and fewer talks in general theory conferences. I learn much more talking directly to my fellow computer scientists. As for the presentations, I can read the papers later.

A fellow computer scientist suggested that we hire a company to videotape the talks and make them available on the web. A back of the envelope calculation suggested we could make this happen for about $10 extra per participant for a reasonably sized conference. I am not a fan of making talks available on the web. Outside of a conference, who has time to sit at a computer screen and watch talks. I also worry about giving people yet another reason not to go to a conference. Remember the most important aspect of a scientific conference are not the talks and papers but bringing members of the community together.

20 comments:

  1. I'll also bet that in the past decade or so there are a lot more theory researchers; thus increasing the quality of published works too.

    Perhaps there is something to the peer pressure effect, though: As technology enables people to make better presentations, someone eventually uses it and raises the bar. In general I would say that PowerPoint has led to a decline in the quality of presentations.

    ReplyDelete
  2. I was wondering how Latex Macros were used in presentations

    ReplyDelete
  3. Same here. As a student, it's often impossible to attend conferences. Listening to a speaker while browsing through his slides has helped me a lot in understanding viewpoints and approaches which I didn't find in the papers. The videos on MSRI are an excellent example of what is possible and as far as I'm concerned, is useful.

    I hope the CS community gives the idea some thought. Maybe even try out a little experiment and see how popular it becomes?

    ReplyDelete
  4. I am not a fan of making talks available on the web. Outside of a conference, who has time to sit at a computer screen and watch talks.

    Actually, this works better. For one, you don't have to listen to all the talks at once. Second, it is easier to "walk out" so-to-speak if the talk turns out not to be of your interest or at the wrong level (too detailed/too general).

    I also worry about giving people yet another reason not to go to a conference.

    Bigger conferences (by federating them) can be an antidote against this. If one holds in a single place a large meeting of theoreticians then (a) it is much more likely one would have a paper in at least one of them and (b) it is easier to justify the time and expense to attend the event.

    For example CC (or SoCG) could federate with satellite workshops or even STOC.

    In other areas of CS (e.g. AI, Networks), as well as many other sciences (Medicine, Bio, Chem) scientific meetings with thousands of attendants are common place.

    Remember the most important aspect of a scientific conference are not the talks and papers but bringing members of the community together.

    A colleague half-jocularly says that the quality of a conference is measured by the number of sessions a conscientious attendant skips. I.e. skipping lots of talks means there were plenty of interesting people around to talk to, which means the conference was a success because it attracted the main participants in the field.

    Alex Lopez-Ortiz

    ReplyDelete
  5. I hope the CS community gives the idea some thought. Maybe even try out a little experiment and see how popular it becomes?

    Wasn't STOC'94 in Montreal videotaped? I seem to remember camera operators recording the talks.

    Alex Lopez-Ortiz

    ReplyDelete
  6. I would love to see videos on the web of the talks delivered in theory conferences. There would certainly be much more people benefiting from this than the number of people that would skip the conference. I am a big fan of the MSRI taped archive. It is ironic, however, that computer scientists are among the last ones to make good use of technology. Look for example at the STOC submission system...

    ReplyDelete
  7. Videotaping lectures is a great idea, and especially helpful for people outside north america. It would also be great if there were more videotapes of longer seminar talks.

    However a start would be to simply make available all the conference papers on the web, so that people that are unable to go to the conference can still read the papers.

    ReplyDelete
  8. I would like to disagree with Lance in one respect.

    It is true that in recent conferences the use of presentation technologies has significantly improved the average talk. In large measure the technology has prevented authors from cramming too much on their slides.

    However, at STOC I (and other people I talked to) noticed a disturbing trend: a tendency for authors to remove almost everything technical from their talks. This seemed to 'dumb-down' some of the conference talks I attended.

    A really good conference talk will spend some amount of time motivating the problems considered and discussing related work. However it will also highlight the really key ideas from the paper as a guide to potential readers. What is it about the techniques in this paper that make it different from all the others? Too often, speakers are now completely leaving out this critical function of the conference talk.

    ReplyDelete
  9. So much to discuss here... On the improvement issue I don't see how the technology caused it (more accurately how 90s technology caused it). It is certainly not on account of Powerpoint fonts (will people drop the "pseudoscript" font already). I think the improvement is more due to people gaining experience, both theirs and others'.

    I second the opinion about not having enough technical details in talks. I like to see a "drawing of the proof". A talk whose summary is "I know how to prove something cool" always leaves me feeling disenfranchized.

    ReplyDelete
  10. Actually, one of the better conference talks I've seen was at FOCS03 - the speaker did not use PowerPoint, only transparencies. This was Chan's "On Levels in Arrangements of Curves, II: A Simple Inequality and Its Consequences."

    Not a promising title for a non-specialist, right? I mean, it has a "II" in the title. Still, the speaker managed to lead us through the problem and the approach step by step. That's the kind of talk I wish were available on video for guidance when I'm doing my own talks...

    ReplyDelete
  11. Lance's worry is accurate (in that it would encourage fewer people to attend the conferences), but its at about the same level as the argument that cars should not be made (and used), since they discourage people from walking.

    Personally, I think it would be fantastic as it would give innumerable more people opportunity to learn and understand current research.

    ReplyDelete
  12. One additional point - videotaping the talks may cause them to improve in quality even more. Knowing that your talk will be recorded for posterity is another motivation to prepare it well.

    It's too bad Yao's FOCS 82 talk was not taped..

    ReplyDelete
  13. With regards to videotaping I cannot as readily endorse the prevalent viewpoint. It is true that I would feel personally uneasy with being videotaped, but it is not my only issue with it. Videotaping would make us a little bit like rock stars, remembered more by our public performance than by the mathematical content we leave behind. This is one instance of social convergence that I prefer not to take place.

    ReplyDelete
  14. PowerPoint has HELPED? Excuse me, I think my brain just exploded. Just once, can't we have a blackboard-only conference?

    Blackboard only conferences are called pure math conferences, and yes, judging from my limited sample they are generally inferior to PowerPoint presentations.

    One of the best CS talks I've seen was by Rolf Klein at a seminar we organized in Dagstuhl, using only the blackboard. Having said that IMHO only the best speakers do better on the blackboard. We average folk do better with prepared powerpoint talks.

    Also for data structures/algorithmic talks the animation facilities from PP are invaluable.

    Alex Lopez-Ortiz

    ReplyDelete
  15. I was wondering how Latex Macros were used in presentations

    The seminar and prosper LaTeX packages have been around for a while now, but the beamer package (https://sourceforge.net/projects/latex-beamer/) blows both of them away.

    I'll agree that PowerPoint, in general (I can't say anything about CS conferences, since CCC05 will be my first), harmed presentations. Far too often, presenters overuse and misuse the flashy animations, fonts, colors etc. Too much in a presentation distracts from the content (or perhaps the content was not too good to begin with and this is the presenter's purpose).

    ReplyDelete
  16. Overusing animations and fonts is a common rookie mistake. Most non-junior academic presenters know enough *not* to do that.

    ReplyDelete
  17. I believe PowerPoint (or similar things such as Latex/beamer etc.) has helped improve quality of talks.

    First, it has the advantage over transparencies that it's very easy to modify your talk, and so you're not tempted to stick with the first version you wrote.

    For 20 minute talks, I believe transparencies or PowerPoint are superior over blackboard. Blackboard talks may be better for 1-2 hour format.

    Unfortunately, I don't think there is a danger of us becoming "rock stars" - even if conference talks were taped and on the web, the number of "hits" most talks would ge will be in the single to double digits. With few exceptions, the reason people will want to view a particular talk is because of the result.

    The ability to look at the taped talk just before reading the paper
    could be very useful.

    Perhaps someone on the FOCS 05 committee could suggest this idea?

    ReplyDelete
  18. Videotaping in a large ballroom is
    extraordinarily difficult and surprisingly expensive. On the other hand many conferences and workshops collect slides in some format or other. That may be a better idea than getting a talking head.

    ReplyDelete
  19. The expense of videotaping is a fair point. How about making the slides available as the previous message suggests together with the sound track? I'd presume sound recording is much cheaper...

    ReplyDelete
  20. This last option (scanned slides + audio recording) has been standard in theoretical physics conferences at least since '97, possibly earlier. The work of scanning and recording is done by student volunteers at the conferences.

    Amit Chakrabarti

    ReplyDelete