No, that's not discrimination, fuckhead.
Article in MIT Tech Review
starts to make a good point but misses it by a thousand miles.
Talking about the Facebook selection bias, the author says that some kinds of bias are unavoidably baked into the code, so it doesn't matter how you use it. True of some types of bias (eg misuse of stats), but none of her examples fit the point.
First she claims...
One area of potential bias comes from the fact that so many of the programmers creating these programs, especially machine-learning experts, are male.
Machine-learning types are Nietzscheans who see themselves as GOD, fully authorized to exterminate all Negative Externalities (IQ<200). This bias shows up in the code of autonomous vehicles, which are specifically designed to exterminate the slow and poor and old. But this is not male vs female, this is psychotic vs normal. The author doesn't give any examples of male vs female, so we have to conclude that the author is an anti-male bigot.
Harvard professor Latanya Sweeney looked at the Google AdSense ads that came up during searches of names associated with white babies (Geoffrey, Jill, Emma) and names associated with black babies (DeShawn, Darnell, Jermaine). She found that ads containing the word “arrest” were shown next to more than 80 percent of “black” name searches but fewer than 30 percent of “white” name searches.
Sounds about right. If black parents would stop giving their kids black names, the problem would disappear. When you identify yourself right up front and publicly by race, advertisers will naturally hit you with ads that are likely to fit your needs.
Rehashing a point I've already beaten to death:
Before 1968 black people didn't have black names. Here's the first few columns of two pages from the 1940 Enid census again. Each page is entirely one race. See if you can tell which is which before you right-click the names of the JPG files. I've copied out the names into text to make the task even easier. (Skipped a couple names that I couldn't decipher.)
Florence, Cache, Eugene, Charles, Bertha, Opal, Robert, Shelley, Pearlie, Carrie, Amy, John, Ida, John, Lillie, Edith, William, Lawrence, Clarence, Johnnie, Teddie, Lorenzo, Annie, Geneva, Jay, Florence, Mariah, Lona, Alonzo, Clarence, Inez, Webster, Laura, Reginald, W.B.
John, Berniece, Roy, Mary, Bobby, Freida, Patricia, Elmer, Ida, Cloyd, Geneva, Billy, Lon, Susie, Billy, Edith, James, Madge, Joe Ann, Jerry, LeRoy, William, Etta, Ina, George, Mary, Henry, George, Lillian, Willie, Charles, Orville, Ernest, Emma, Delores, Ramona, Ernest Jr, William.
There's no way an algorithm could sort those names AS names. If the algorithm also knew the Zip code, it might be able to make a better guess; but even in mostly segregated 1940 you'd really need LOCAL knowledge of which blocks were black.
In other words, the example given by the author is NOT discrimination by algorithm. It is simply an accurate use of FACTS. It is SCIENCE.