3 ms·

Is your software racist?

6 points by dsr12 9y ago

anonytrary 9y ago
More like "Does your software use math to establish statistical correlations in data, and does that make you angry?" Relevant excerpt: > [A Microsoft researcher] said that omitting variables like gender and race in different algorithms isn’t always the solution to countering bias. In some cases, like medical predictions, these variables could be important to accuracy. And there can be other variables, like ZIP codes, that can correlate with race and introduce bias into models that don’t explicitly include race.
- jrs95 9y ago
  Yeah, this has way more to do with limits in software implementations than anything else imo. But of course the implication in some of the language in this article is that we wouldn’t have this problem if we just got rid of the white men. I have a different suggestion. Let’s just give control of the software development process to the HR department and see if they can eliminate these issues and effectively deliver products at the same time.
jimmywanger 9y ago
Take this example. Black people are 37.8% of the prison population[0] and 12.3% of the general population[1]. They also have 72% of babies being born out of wedlock[2], compared to 29% of white-non hispanic babies[2]. Both those factors correlate to higher probabilities of crime in the future[3] and higher incarceration rates[4] So they are ~3x overrepresented in the prison population. So no. Even if you drop race, black people will generally be more highly targeted by any sort of fair algorithm. There is no thing as a racist algorithm or racist data. There can be racist systems and policy causing this data to be generated and can be looked for and targeted by algorithms, but even without taking race into account, just dealing with the current algorithm, the software is accurate and doing its job. Most modern science was caused by people questioning the status quo and challenging political orthodoxy. [0]https://www.bop.gov/about/statistics/statistics_inmate_race.jsp https://www.bop.gov/about/statistics/statistics_inmate_race.... [1]https://www.google.com/search?q=percentage+of+blacks+in+america&rlz=1CAHPZR_enUS781US781&oq=percentage+of+black&aqs=chrome.2.69i57j0l5.5520j0j1&sourceid=chrome&ie=UTF-8 https://www.google.com/search?q=percentage+of+blacks+in+amer... [2]https://www.google.com/search?q=percentage+of+blacks+in+america&rlz=1CAHPZR_enUS781US781&oq=percentage+of+black&aqs=chrome.2.69i57j0l5.5520j0j1&sourceid=chrome&ie=UTF-8 https://www.google.com/search?q=percentage+of+blacks+in+amer... [3]http://www.politifact.com/truth-o-meter/statements/2013/jul/29/don-lemon/cnns-don-lemon-says-more-72-percent-african-americ/ http://www.politifact.com/truth-o-meter/statements/2013/jul/... [4]https://familyinequality.wordpress.com/2011/10/20/single-parents-crime-and-incarceration/ https://familyinequality.wordpress.com/2011/10/20/single-par...

6 points by dsr12 9y ago

More like "Does your software use math to establish statistical correlations in data, and does that make you angry?" Relevant excerpt: > [A Microsoft researcher] said that omitting variables like gender and race in different algorithms isn’t always the solution to countering bias. In some cases, like medical predictions, these variables could be important to accuracy. And there can be other variables, like ZIP codes, that can correlate with race and introduce bias into models that don’t explicitly include race.

jrs95 9y ago
Yeah, this has way more to do with limits in software implementations than anything else imo. But of course the implication in some of the language in this article is that we wouldn’t have this problem if we just got rid of the white men. I have a different suggestion. Let’s just give control of the software development process to the HR department and see if they can eliminate these issues and effectively deliver products at the same time.

jimmywanger 9y ago

Take this example. Black people are 37.8% of the prison population[0] and 12.3% of the general population[1]. They also have 72% of babies being born out of wedlock[2], compared to 29% of white-non hispanic babies[2]. Both those factors correlate to higher probabilities of crime in the future[3] and higher incarceration rates[4] So they are ~3x overrepresented in the prison population. So no. Even if you drop race, black people will generally be more highly targeted by any sort of fair algorithm. There is no thing as a racist algorithm or racist data. There can be racist systems and policy causing this data to be generated and can be looked for and targeted by algorithms, but even without taking race into account, just dealing with the current algorithm, the software is accurate and doing its job. Most modern science was caused by people questioning the status quo and challenging political orthodoxy. [0]https://www.bop.gov/about/statistics/statistics_inmate_race.jsp https://www.bop.gov/about/statistics/statistics_inmate_race.... [1]https://www.google.com/search?q=percentage+of+blacks+in+america&rlz=1CAHPZR_enUS781US781&oq=percentage+of+black&aqs=chrome.2.69i57j0l5.5520j0j1&sourceid=chrome&ie=UTF-8 https://www.google.com/search?q=percentage+of+blacks+in+amer... [2]https://www.google.com/search?q=percentage+of+blacks+in+america&rlz=1CAHPZR_enUS781US781&oq=percentage+of+black&aqs=chrome.2.69i57j0l5.5520j0j1&sourceid=chrome&ie=UTF-8 https://www.google.com/search?q=percentage+of+blacks+in+amer... [3]http://www.politifact.com/truth-o-meter/statements/2013/jul/29/don-lemon/cnns-don-lemon-says-more-72-percent-african-americ/ http://www.politifact.com/truth-o-meter/statements/2013/jul/... [4]https://familyinequality.wordpress.com/2011/10/20/single-parents-crime-and-incarceration/ https://familyinequality.wordpress.com/2011/10/20/single-par...