ethical-algorithms   15

Book Review by John Derbyshire: Doesn’t Add Up
Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy
https://twitter.com/HoustonEuler/status/887479542360702977
My embedded opinion is that Cathy O'Neil frequently writes foolish things
She's a former mathematician/finance quant who dresses up a lot of progressive dogma with phony skepticism

http://www.vdare.com/tag/minority-mortgage-meltdown
http://www.slate.com/articles/business/moneybox/2008/10/subprime_suspects.html
http://voxeu.org/article/minority-mortgage-market-and-crisis

Causes of the Financial Crisis: https://spottedtoad.wordpress.com/2016/01/17/causes-of-the-financial-crisis/
Look, Wall Street was definitely a bad actor from 2000-2008. But the idea that they were solely responsible for the crisis has got to go. There were four main factors, in descending order:

a) A huge global savings glut that meant there were vast amounts of cash people were eager to lend out, combined with…

b) Enormous pressures to make use of all that money to increase lending and reduce standards for lower-income and minority households.

This book has some of the story, mostly focusing on the expansion and growing political power of Freddie and Fannie in the 1990s, but it really wasn’t any secret that reducing credit-worthiness (sorry, barriers to homeownership) was an explicit goal of the Clinton and Bush administrations and affiliated banks. Bush gave a long speech on these goals in mid-2002. Countrywide, led by Angelo Mozilo, pledged $600 billion in loans to low-income and minority homeowners in early 2003. Then, the Bush administration was bragging in late 2004 about the commitments they had elicited from lenders to expand low-income and minority lending by over $1 Trillion. Then, a few months later, in 2005, Countrywide, with a former HUD secretary on its board, released a press release bragging that they were going to increase their book of lending to minority and low-income households to $1 Trillion. Looking back on the crisis, liberal sociologists find, unsurprisingly, that subprime lending and the subsequent foreclosures were concentrated in minority households.
c) The efforts to extend massive amounts of credit to non-creditworthy families were abetted by fraud and irresponsible borrowing by those same households. See, for example, Atif Mian’s papers on widespread fraud in mortgage applications.
d) Bad actions by Wall Street (Inside Job is probably a good version of this.)

The idea that only d matters is just nuts.
pdf  essay  org:mag  books  review  summary  critique  westminster  culture-war  migration  cycles  housing  policy  government  politics  ethical-algorithms  gnon  isteveish  albion  right-wing  rant  attaq  paleocon  multi  twitter  social  commentary  unaffiliated  chart  news  org:lite  list  stream  org:ngo  econotariat  economics  macro  roots  big-picture  finance  race  diversity  track-record  ratty  debt  usa  statesmen  events  letters 
may 2017 by nhaliday
A computer program used for bail and sentencing decisions was labeled biased against blacks. It’s actually not that clear. - The Washington Post
How We Analyzed the COMPAS Recidivism Algorithm: https://www.propublica.org/article/how-we-analyzed-the-compas-recidivism-algorithm
Automatic Justice: http://www.theamericanconservative.com/articles/automatic-justice/
A.I. ‘BIAS’ DOESN’T MEAN WHAT JOURNALISTS SAY IT MEANS: https://jacobitemag.com/2017/08/29/a-i-bias-doesnt-mean-what-journalists-want-you-to-think-it-means/
When a journalist discusses bias, they typically do not mean it in the same manner that statisticians do. As described in the examples above, a journalist typically uses the term “bias” when an algorithm’s output fails to live up to the journalist’s ideal reality.
Machine Learning and Human Bias: https://www.youtube.com/watch?v=59bMh59JQDo
analysis  ethical-algorithms  crime  critique  news  debate  org:rec  technocracy  multi  media  data  org:mag  investigative-journo  right-wing  douthatish  policy  law  criminal-justice  criminology  madisonian  org:popup  gnon  techtariat  rhetoric  biases  meta:rhetoric  propaganda  bias-variance  data-science  video  presentation  google  censorship  sv  tech 
november 2016 by nhaliday

related tags

academia  acmtariat  akrasia  albion  algorithms  alignment  analysis  applications  attaq  attention  bandits  bias-variance  biases  big-picture  bio  biotech  blog  books  browser  business  censorship  chart  civil-liberty  clever-rats  commentary  computer-vision  concept  cool  creative  crime  criminal-justice  criminology  critique  crypto  culture-war  culture  current-events  cycles  data-science  data  dataset  debate  debt  degrees-of-freedom  demographics  design  differential-privacy  diversity  douthatish  economics  econotariat  embeddings  essay  ethics  events  externalities  facebook  features  finance  genetics  genomics  geography  geometry  gnon  gnxp  google  government  graphs  hci  hi-order-bits  hmm  housing  huge-data-the-biggest  identity  inhibition  integration-extension  intel  interdisciplinary  internet  interview  investigative-journo  isteveish  justice  kinship  law  legibility  lesswrong  letters  leviathan  list  machine-learning  macro  madisonian  managerial-state  maps  matching  math.at  math.mg  math  meaningness  measurement  media  meta:rhetoric  methodology  metrics  migration  minimalism  mit  mrtz  multi  news  nibble  nlp  online-learning  operational  org:bleg  org:edu  org:inst  org:lite  org:mag  org:med  org:ngo  org:popup  org:rec  org:sci  paleocon  pdf  philosophy  policy  polisci  politics  population  preprint  presentation  privacy  profile  propaganda  proposal  protocol  race  rant  ratty  reflection  regulation  review  rhetoric  right-wing  risk  roots  sanctity-degradation  scale  scitariat  security  social-choice  social  society  software  spreading  startups  state-of-art  statesmen  stream  study  summary  sv  systems  tcs  tech  technocracy  technology  techtariat  the-monster  the-watchers  thinking  time-preference  time-use  tools  top-n  topology  track-record  trees  twitter  unaffiliated  unintended-consequences  usa  vague  values  video  visualization  volo-avolo  westminster  whole-partial-many  wonkish  worrydream  🖥  🦉 

Copy this bookmark:



description:


tags: