Sunday, February 22, 2009

Bad Research Talks

I'll admit that I'm particularly bad at understanding presentations because I have no patience -- "This is boOOooring. Hmmm, what am I having for dinner tonight? Oh crap, what did they just say?" ... and then I'm lost for the rest of the talk. But people, please, at least try to make your talks accessible. Many conferences now have "best paper" awards; I think they should also have "best talk" awards so that everybody tries harder. Related to that: WHATEVER YOU DO, UNDER ANY CIRCUMSTANCES, DO NOT USE BULLETS IN POWERPOINT. ALSO, ALWAYS TURN OFF AUTO-RESIZING OF FONTS. Your fonts should be 28pts or higher, and the size should be uniform throughout the presentation.

If more than 75% of the audience thinks a talk really sucks, we should ban the speaker from that conference for the next three years. After three such infractions, the speaker should be voted off the research island.

25 comments:

  1. Why no bullets? I don't understand how bullets affect the quality of a talk.

    ReplyDelete
  2. Hmm. You admit that you have no patience, then you blame the speaker for your inability to concentrate on their talk?

    "Best Paper" awards given in real-time are the biggest bogosity ever. You cannot tell which is the best paper at a conference until many years later. It's just an empty political game.

    Replacing "Best Paper" awards by "Best Talk" awards does makes sense, though, because then you are honoring the event of the talk, not the content of the paper.

    But it will encourage empty theatrics that emphasize form over content. I can think of several people in my field who would walk away with every "best talk" award every time, but whose papers rarely are all that important even two or three years later. They're good at being popular. Is that what we want to encourage?

    ReplyDelete
  3. You said: But it will encourage empty theatrics that emphasize form over content.

    Yes, that's what I want to encourage. I don't care how good the content is if I can't follow the talk. Inaccessible talks are a waste of everybody's time; I'll just go and read the paper.

    And yes, I blame the speakers for bad talks even if I'm impatient: somehow some good speakers manage to keep my attention.

    ReplyDelete
  4. Looks like you've been through a really lousy talk.

    ReplyDelete
  5. In order to accurately judge such a contest, I think all talks would need to be videotaped. To the extent that this encourages conferences
    that don't already do so to videotape their talks and make them available online, I think this is a good thing.

    But to some extent we already have informal "best talk" awards — those are the ones that get reported on by the bloggers at the conferences.

    As for Existential Type's "don't blame the speaker, blame the audience" approach: I think we've all had to sit through objectively
    bad talks. Talks that proceeded in a monotone mumble, that packed long paragraphs of text onto each slide, that started immediately into
    piles of dry equations with no attempt at motivation or intuition, or that consisted of lots of pretty stock art and fancy powerpoint
    transitions with little connection to the subject and no technical meat. I'm not sure whether it's appropriate to explicitly discourage bad presenters, though — won't they get what they deserve anyway, in decreased impact of their paper?

    ReplyDelete
  6. I totally agree with the idea of videotaping to judge the speakers. It sucks to be an audience of some great researcher only to realize that his talk just blows but can't leave just to show him some respect.

    ReplyDelete
  7. Luis, how about posting some of your own slides as examples? =)

    ReplyDelete
  8. Are you serious about this post??? Does good talk imply good research??? What do you mean by banning from research island??? No one has monopoly over the research island (atleast I think so). No one wants to give bad talks, its just that they get nervous or not native english speakers.

    ReplyDelete
  9. It's not a problem with talks, it's a problem with attendees that don't walk away when they feel they're wasting their time.

    ReplyDelete
  10. Are you serious about this post??? Does good talk imply good research???

    I'm mostly serious :) Clearly we'll never vote anybody off the research island. But I do think there should be "Best Talk" awards.

    I don't think a good talk implies good research at all. But I don't want to hear bad talks. I don't understand why talks are needed if they are going to be completely inaccessible. The paper is enough in that case.

    No one wants to give bad talks, its just that they get nervous or not native english speakers.

    I buy the nervous bit. (Although, especially if you plan to become a professor, you better get over that fast.) But I simply do not buy the "not native English speaker" bit. The overwhelming majority of bad talks that I've seen have nothing to do with the person's accent.

    I think part of the problem is with universities: we're not making our students give enough practice talks.

    ReplyDelete
  11. One learns these skills through practice. Someone has to give a bad talk, get embarrassed and then learn. So unfortunately or fortunately this cycle has to go on. Sorry Luis but this is a harsh reality.

    ReplyDelete
  12. There is some truth to what you're saying, but I think this learning curve is way too slow in too many cases. I honestly think some people just don't think it's that important to give good talks.

    ReplyDelete
  13. I thought that it was the exclusive right of the students to wonder in lectures and talks.
    ;)
    Nice to know some profs. share it too

    ReplyDelete
  14. I completely agree with Luis. Bad talks are just a waste of time. A researcher who can't explain properly what he is doing is simply a bad researcher, because that is part of his job.
    What bothers me in particular are talks that are simply prepared badly -- I suggest that a publication only is accepted definitely for a conference AFTER the talk.

    ReplyDelete
  15. On an abstract level, a good talk is no different than a paper well written. And we already have a system which favors such papers regardless of their 'real' content- I even doubt if it can be measured for most cases. But we do need some heuristics/metrics and dependent incentives for almost everything. We already have them for papers, why not have them for talks too?

    ReplyDelete
  16. A lot of talks are universally bad, but just to play devil's advocate:

    If 75% of the general audience thinks the talk is good, but the 25% most closely related to the research thinks it sucks, is this really better?

    ReplyDelete
  17. Yup, no bullets:
    http://www.cs.berkeley.edu/~jrs/speaking.html

    Sadly, bullets+lots-o-text on slides is a really tough habit to kick.

    ReplyDelete
  18. I believe that reputation should be a natural incentive for giving good talks. When attending a venue, I don't read all of the papers but listen to most of the talks. What is likely to give a permanent impression is thus the quality of the talk; people are likely remember it. Out of own interest, it should be a motivation to give a good talk in order to let people associate something good with a particular paper / speaker and/or motivate them to read the paper and, therefore, hopefully reuse and cite it in their own work. It is just about advertisement.
    But I agree with you, having a best presentation award as well established as the best paper award is nowadays, would be an additional incentive to give good talks. This is also something one can place in CV's and on web pages ;-)

    ReplyDelete
  19. A lot of people appear to try to make their talk replace the need to read the paper. That is difficult to do well.

    The strategy that seems to work better is making the talk almost an advertisement for the paper. The goal of these talks seems to be to get people who otherwise would not read the paper excited about the work. These talks focus on motivating the problem and describing the intuition behind the approach used. They expect people who want the details to read the paper.

    ReplyDelete
  20. Greg made an excellent point (as Luis did with this provocative post).

    A talk should be about advertising your research paper: provide a 15-20 min entertainment experience for the audience and let them know the key ideas of what they will find (detailed) in the paper.

    Please do not try to explain the hidden variables of the magical formula of the evaluation Section 4.3.1

    I would blame (partly) Microsoft Powerpoint (and competitors) for their influence in preparing talks with auto-size bullets.

    Anyone interested in understanding why such talks do not do their work properly or anyone just wanting to improve their ideas delivery with more visual and natural talks:

    http://www.presentationzen.com

    ReplyDelete
  21. Many institutions limit access to their online information. Making this information available will be an asset to all.

    ReplyDelete
  22. The strategy that seems to work better is making the talk almost an advertisement for the paper.

    ReplyDelete