It took me a while to find a topic related to information cascades that wasn’t already in the blog, but I eventually came across a 16 page research paper by John H. Miller and Scott E. Page of the Sante Fe Institute about modeling standing ovations. It turns out that an information cascade model similar to the one that we have learned in class can be applied to standing ovations. Note, however, that their information cascade model does differ in some respects from ours, and that it is in general more complicated than ours.

The standing ovation problem is the following: At the end of a presentation, lecture, or other event, there is applause from the audience, and some audience members may or may not choose to stand. How can we determine if the audience will begin to give a standing ovation, or if they will simply remain seated and the applause will calm down. The application of information cascades to this problem should be fairly obvious: If audience member(s) around you choose to stand, your choice of whether or not to stand will probably be influenced by their decision(s) to stand. You probably know from experience that if everyone around you is standing and applauding, you feel pressure to stand as well, even if you didn’t enjoy the lecture or performance.

One version of the standing ovation information cascade model is as follows: Audience members choose to Stand or Sit (referred to as Good and Bad in our class discussions). An audience member can look left and right (to see the people next to him or her) and can see the people in the row directly in front of him or her (a simplified model, but there have to be some simplifications to reduce complexity). An audience member will change his decision from Sit to Stand if there are at least two more people that he can see who are standing than people he can see who are sitting (and vice versa from Stand to Sit). Consider the following example:

(three rows of three audience members, and their actions after the lecture/performance)

Stand Sit Stand

Sit Sit Sit

Sit Sit Sit

In this example, the person in the middle of the front row (he can only see the people to his left and right) chooses to Stand because the two people he can see are standing as well. Then, the people on the outsides of the second row stand up because they see three people in the front row standing, while they only see themselves and the person in the middle of the middle row sitting. Then, the person in the middle of the middle row stands, and so on, until the entire audience is standing. Here are the steps in the cascade:

Stand Stand Stand - Stand Stand Stand - Stand Stand Stand

—-Sit Sit Sit ———> Stand Sit Stand —> Stand Stand Stand

—-Sit Sit Sit ———— Sit Sit Sit —————- Sit Sit Sit

-

— Stand Stand Stand — Stand Stand Stand

-> Stand Stand Stand -> Stand Stand Stand

—–Stand Sit Stand —- Stand Stand Stand

As you can see, even though only two people are initially standing, the entire audience is standing at the end. One important difference to note between this information cascade and the one we discussed in class is that decisions are not made in sequence (anyone can change their decision to stand or sit at any time).

The paper goes on to discuss other ways of modeling standing ovations, including conformity, growth/coordination, and diffusion (see link below for more details). It concludes by pointing out that there are so many dynamics in a real auditorium and a real audience (people leaving the auditorium, people going with their friends, people sitting on different levels (e.g. balconies), where the entrances are located, the fact that people more interested in the talk tend to sit in the front, etc…) that predicting if and how a standing ovation will occur is nearly impossible in many circumstances. However, the paper was still an interesting look at another application of information cascades.

Reference: http://zia.hss.cmu.edu/miller/papers/ovation.pdf