3 ms·

There is a great book explaning the philosophy, story and problem of using averages in many fields - "The End of Average" by Todd Rose. [1] I originally found

by divan 4y ago

There is a great book explaning the philosophy, story and problem of using averages in many fields - "The End of Average" by Todd Rose. [1] I originally found it from the fascinating article on how using averages in the design of military plane cockpit resulted in many pilot deaths. That's the best introduction I've seen about uselessness of averages with multiple dimensions (with more than 3 dimensions it goes bananas). The article is actually is an excerpt from the book, so you can get the sense of the book level from it. [2] [1] http://www.toddrose.com/endofaverage http://www.toddrose.com/endofaverage [2] https://www.thestar.com/news/insight/2016/01/16/when-us-air-force-discovered-the-flaw-of-averages.html?rf https://www.thestar.com/news/insight/2016/01/16/when-us-air-...

taneq 4y ago
I tend to explain this with car analogies. If you’re about to drive through an intersection and the light goes orange, you either want to brake hard, or maintain a steady speed, but taking the average of these actions will not end well.
- sokoloff 4y ago
  (When I was younger) My friends colloquially used “orange” to mean a light that was a mix of amber (yellow) and red (meaning “too late to be entering the intersection legally”).
- jldugger 4y ago
  Three economists are on the football field, practicing field goals. The first kicks the ball was to the left, widely missing the goal. The second one kicks and shanks it to the right, again missing the goal. The third one runs off the field shouting "we got it, we got it!"
HFguy 4y ago
Thanks for the suggestion. I think this problem is just a matter of degrees with any summary statistic. You throwing away information in an attempt to synthesize an insight.
jldugger 4y ago
I read that book, and as I recall it takes a left turn into education policy after chapter one. Part 1 sells you a story of personalization over one size fits all, and the next parts tell you how individuals vary, mostly in ways understood by educators, and emphasizing the authors personal story of getting a PhD as a high school dropout and how to remove barriers to educational attainment. It's probably a fine book if education policy is your jam, and its fine for scholars to publish monographs on their research but I felt a bit mislead given the title and PR excerpts. For example, maybe an examination of Average Treatment Effect and Individual Treatment Effect in medicine, and followed by an adventure into causal reasoning. Or perhaps a treatment on diminishing returns of personalization. And perhaps if the "how we succeed" portion investigated productivity instead of defining success as getting a degree from Harvard. As software designers we come to recognize that there's diminishing returns on customization in our software, and develop hierarchies of customization -- commonly adjusted settings are easily found, and others may be dotfile / envvar / registry only. There's probably a happy medium between Apple inspired "it just works" philosophy and Gentoo's 'every software custom built' approach.

by divan 4y ago

taneq 4y ago

I tend to explain this with car analogies. If you’re about to drive through an intersection and the light goes orange, you either want to brake hard, or maintain a steady speed, but taking the average of these actions will not end well.

sokoloff 4y ago
(When I was younger) My friends colloquially used “orange” to mean a light that was a mix of amber (yellow) and red (meaning “too late to be entering the intersection legally”).
jldugger 4y ago
Three economists are on the football field, practicing field goals. The first kicks the ball was to the left, widely missing the goal. The second one kicks and shanks it to the right, again missing the goal. The third one runs off the field shouting "we got it, we got it!"

HFguy 4y ago

Thanks for the suggestion. I think this problem is just a matter of degrees with any summary statistic. You throwing away information in an attempt to synthesize an insight.

jldugger 4y ago

I read that book, and as I recall it takes a left turn into education policy after chapter one. Part 1 sells you a story of personalization over one size fits all, and the next parts tell you how individuals vary, mostly in ways understood by educators, and emphasizing the authors personal story of getting a PhD as a high school dropout and how to remove barriers to educational attainment. It's probably a fine book if education policy is your jam, and its fine for scholars to publish monographs on their research but I felt a bit mislead given the title and PR excerpts. For example, maybe an examination of Average Treatment Effect and Individual Treatment Effect in medicine, and followed by an adventure into causal reasoning. Or perhaps a treatment on diminishing returns of personalization. And perhaps if the "how we succeed" portion investigated productivity instead of defining success as getting a degree from Harvard. As software designers we come to recognize that there's diminishing returns on customization in our software, and develop hierarchies of customization -- commonly adjusted settings are easily found, and others may be dotfile / envvar / registry only. There's probably a happy medium between Apple inspired "it just works" philosophy and Gentoo's 'every software custom built' approach.