Friday, November 22, 2019

Survivorship Bias and User Ratings

Every episode in a TV series has an innate tendency to draw better user ratings than the previous episode in places like IMDB.com, due to Survivorship bias. People who dislike - or don't particularly love - a show gradually drop out, leaving an increasingly appreciative remainder. Same for show seasons (though of course particularly bad and good episodes and seasons will dip or peak accordingly). It works the same for anything consecutive, including, say, YouTube multi-part series, movie franchises with multiple sequels, etc.

Survivorship bias is a deep trove of insight in the guise of a logical fallacy. You can deconstruct much of the world by pondering Survivorship bias, and, as with all the best insight, the result is reliably "duh" (like the above!). Survivorship bias is so pandemic and intellectually fertile that I've never finished any of the many postings I’ve started writing about it. It's just too much.

People these days can't stop talking about Dunning–Kruger effect and Confirmation Bias (understandable, as they're the intellectual maladies of our day, though I think "everyone's an aristocrat, too sensitive and opinionated to recognize the Utopia they're in" is the more fundamental insight underpinning both). But Survivorship bias is the shizzle.


Another oddity with multi-part series: if Jesus, Moses, and Buddha returned to Earth to offer sure-fire instructions for making everything totally okay, and the message was presented via a series of YouTube videos, installment #2 would have fewer views than installment #1, and installment #3 would have fewer still, etc. etc.. The Law of Dwindling YouTube Viewership (which I just made up, qualifying this post for the tag "Leff's Laws") is rock solid regardless of interest level (though it doesn't seem to apply to movies or TV for reasons mysterious to me). In this case, returning to my previous point, I think user-ratings would remain consistently high (despite those damned satanists trying to game the results).

No comments:

Blog Archive