3 ms·

For social-science applications, I think the information-theoretic notion of entropy is more appropriate. There's no social analog to the laws of thermodynamic

For social-science applications, I think the information-theoretic notion of entropy is more appropriate. There's no social analog to the laws of thermodynamics, but we can use the notion of entropy to make statements about information content. In particular, the idea (along with the adjacent subject of Bayesian statistics) has sharpened my thinking about news: there's less information (to me) in otherwise trustworthy headlines/stories that agree with my prior notions. In an analytic sense this isn't surprising, but it runs against the emotional impulse to seek out affirming information.

contravariant 6y ago
You still need to be a bit careful, because what you're describing as entropy isn't how much information is in it, but rather how much you didn't already know. Hence random noise has the maximal 'information content' but isn't useful in the slightest. And really I wouldn't worry too much about finding some mathematical justification for everything. It's an easy way to get caught up in technicalities for something that you already knew.