Surprise! The Element Of Surprise Has Been Mathematically Modeled
from the there's-just-no-spontaneity-any-more dept
Answering the question of whether or not something can be a surprise if it’s predictable, some researchers claim to have mathematically modeled the likelihood of surprise — currently based on watching a video. The description isn’t entirely clear (and those who are more mathematically minded are urged to weigh in with just how innovative this is, or isn’t), but it suggests the mathematics of surprise could be useful in trying to gain attention by standing out from the pack. The idea is that using this model, you could see if your “surprise” is actually all that surprising. Either way, saying you’ve mathematically modeled the element of surprise probably doesn’t play well on first dates.
Comments on “Surprise! The Element Of Surprise Has Been Mathematically Modeled”
“Data that does not change your beliefs is not surprising,” says itti
So they created a method to determine what peoples beliefs are, analyzed it and created a method for determining what would and wouldn’t be a normal belief?
Hmmm, sounds like they are working on a method to reach a specific demographic based on beliefs instead of the norm, i.e. male, female, age, and locality?
I’m surprised I’m still awake after reading this post…
No Subject Given
There’s not enough in that article to make a meaningful critique, but it really depends on the assumptions that their model makes for it to be applicable. All mathematical models — representations of reality — are bounded with assumptions, and it’s when those assumptions start significantly oversimplifying the complexities of the real world that we start scoffing at the models. Whether this one is breakthrough or just something to get published, we can’t tell without reading the paper.
Looks like a perceptual equivalent of measuring entropy to me.
Re: Seeing Shannon
Here is a presentation from a conference: http://www.cnse.caltech.edu/Industry/Conferences/2004/AIC/pdf/Laurent-Itti.pdf
This looks like the actual article: