So, as I pointed out before, last week was the annual APS March Meeting. For those of you unfamiliar with the conference, it consists of thousands of presentations given by professors, post-doctoral students, graduate students, and even undergraduate students. Most presentations are scheduled to last ten minutes followed by two minutes of questions for the speaker. This schedule is enforced by the chair of the session. This person is tasked with deciding how many questions should be asked and informing the speaker if they are running out of time. I noticed as the week went on that the talks always seemed to be rushed toward the end and there were often very few questions allowed by the session chair. This observation brought out the scientist side of me. I decided to take data and determine how long the talks were actually taking. So, this is what I did: starting on Thursday, I timed each presentation I went to and recorded the time at the end of the presentation and then again at the end of the question period. To see the results of the experiment read on!
First, a disclaimer: I started this experiment on Thursday, so the sample size is not as large as I would like (N=36), but the results are still interesting. First, I made a histogram of the length of the presentation. Then I fit the data to a Gaussian function. This provides an estimate for the average length of the presentation and also for the spread in the length of the talks. Here is the histogram with fitted Gaussian function for the length of the presentation:
As you can see, the data is not very well distributed in a bell curve, but the peak of the Gaussian does look to be about centered where one would think that the average lies. So, this fit gives an average length of presentation of 650 seconds, or 10 minutes 50 seconds. This means that the average presentation lasts almost one minute longer than is allotted. However, when we look at the same data for the end of the question period, we see that this lack of time for questions does not pose a problem for the total time allotted to a speaker.
This fit looks a bit better than the previous fit. More importantly, the center of this distribution is located at 727 seconds, or 12 minutes 7 seconds. This is almost exactly the scheduled length of talks. Also, from looking at the fit for the data, it seems to be symmetrically distributed so that in a session with many talks, there are going to be roughly as many long talks as short ones so that the session stays on schedule.
Something to notice about this information is that the second fit has a center of only 1 minute 17 seconds longer than the first fit. This means that even though the talks go longer than scheduled, the session chair succeeds at intervening through limiting the amount of questions to keep the session on schedule. So, in the end, the session chair plays an important role in keeping the conference on track. With so many talks being given over the course of the week, if even one session ran late, then it would create a domino effect through the rest of the day and mess up the schedule of the conference. So, a job well done for the APS session chairs, keeping the conference moving and also a great job by the presenters giving some super interesting talks. It was a great experience and I look forward to future conferences.