Re:Google Fires Employee Behind Controversial Diversity Memo
by oh_sigh on Tuesday August 08, 2017 @01:08AM
(#14954575)
You're somewhat correct with your analysis. The bar stays the same, but on average those who get a more detailed analysis done will be closer to the bar than those who are subjected to coarser grained measures. So, it's wrong to say it lowers the bar, but it's right to say that it ends up giving you more people who are very close to the bar. Perhaps a better way to word it would be that Google's diversity hiring method(of looping back around for a closer look on "diverse" candidates) shrinks the error bars on the bar
Imagine it this way: You're a fisher. The Law says that you can only catch fish which are >= 12" long. If you're caught with < 12" fish, bad things happen, but otherwise you want to maximize the amount of fish you catch.
You have two ways of measuring length: a crude wooden stick with some fat lines on it which mark approximately 12" - this is quick but only approximate. You also have a high precision caliper which can tell you the length down to .001", but it takes 2x as long to make the measurement.
You're interested in catching lots of fish, so if any fish you catch are a little close to that line, you just throw them back with one measurement.
90% of the fish you catch are red, but about 10% of the times you catch a green fish, which are really sparkly and fun, so you change your standard: if the green fish on first measurement is really close to 12"(not close enough to keep), instead of throwing it back like a red fish, you bust out the caliper and measure it more precisely, and throw out anything < 12".
So, at the end of the day, assuming red and green fish are distributed equally across lengths in the ocean, what would the distribution of kept fish look like? Would the mean and median length of your green fish be higher or lower than the mean and median of your red fish?
Imagine it this way: You're a fisher. The Law says that you can only catch fish which are >= 12" long. If you're caught with < 12" fish, bad things happen, but otherwise you want to maximize the amount of fish you catch.
You have two ways of measuring length: a crude wooden stick with some fat lines on it which mark approximately 12" - this is quick but only approximate. You also have a high precision caliper which can tell you the length down to .001", but it takes 2x as long to make the measurement.
You're interested in catching lots of fish, so if any fish you catch are a little close to that line, you just throw them back with one measurement.
90% of the fish you catch are red, but about 10% of the times you catch a green fish, which are really sparkly and fun, so you change your standard: if the green fish on first measurement is really close to 12"(not close enough to keep), instead of throwing it back like a red fish, you bust out the caliper and measure it more precisely, and throw out anything < 12".
So, at the end of the day, assuming red and green fish are distributed equally across lengths in the ocean, what would the distribution of kept fish look like? Would the mean and median length of your green fish be higher or lower than the mean and median of your red fish?