Hacker News Reader

Comment by alexpetralia | Hacker News Reader

There must be a term for this, but I imagine no one is average across all attributes. Curse of dimensionality maybe?

“Average pilot fallacy”? https://www.thestar.com/news/insight/2016/01/16/when-us-air-...

When designing cockpits using the average dimensions of pilots the US Air Force found exactly 0 pilots who were average in all 10 dimensions. Even just using 3 dimensions only 3.5% were considered “average”. Take the average dimensions of a human hand and draw that hand. Almost nobody has the dimensions of that hand.

That’s so cool: just occurred to me it’s like the inverse of crowdsourcing. No one guesses the right weight of the cow, but the average of all guesses is almost exactly the right weight.

Well, here you never know the weight of the cow, per se. If the weight-of-cow question were asked, it would "If you were asked to guess the weight of some cow, do you believe you would overestimate it?" And then you find out, that, 63% other people share your self-evaluation, or else do not. There is no info about any actual cow.

On HN, there was recently an article [1] on this topic. You may also be interested in the discussion [2,3].

[1] https://www.thestar.com/news/insight/2016/01/16/when-us-air-...

[2] https://news.ycombinator.com/item?id=22956832

[3] https://news.ycombinator.com/item?id=11230287

I don’t know a term for it, but here's a relevant blog post:

https://www.johndcook.com/blog/2011/09/01/multivariate-norma...

> For a multivariate normal distribution in high dimensions, nearly all the probability mass is concentrated in a thin shell some distance away from the origin.

Reply on news.ycombinator.com