Why it's time to stop blindly believe in big data
A Life / / December 19, 2019
algorithms decide now who approve the loan, insurance or who receive an invitation for an interview, but often they do so unfairly. And it only increases the gap between the layers of the population.
Katie O'Neill (Cathy O'Neil)
The mathematician, on the analysis of financial markets specialist, author of the book "Weapons of mathematical defeat."
To construct an algorithm, we need two things: data (what happened in the past) and the definition of a successful outcome (what you want to find by using this algorithm). It then determines which criteria lead to a successful outcome. But the definition of success can not be universal.
Algorithm - is someone else's opinion, the built-in code.
We used to think that algorithms are objective and reliable, but it is only a marketing gimmick designed to intimidate us and make us trust in algorithms and mathematical data.
O'Neill cites examples where algorithms can cause serious harm. This happens when evaluating employees. For example, in 2011 in a school in Washington County have been dismissed more than 200 teachers after their
weed out algorithmEven though they had excellent recommendations from their parents and peers.In addition, the algorithms are often the reason for removal of biased verdicts. News organization ProPublica recently conducted an investigation and foundThat the algorithms that determine the risk of recidivism, work objectively. At the same crimes sentences often taken out black Americans.
We are all subject to biases, and we bring them into the algorithms that decide which data needs to be taken into account.
Algorithms are simply repeating our past mistakes, automate the existing order. So we can not blindly trust them, we need to test them to be objective: to rethink the definition of a successful outcome, error, are not insured by any algorithm. How often they occur and who is affected? What is the cost of such errors?
Professionals working with the data, should not be the arbiters of justice. It's time to stop blindly believe big data.