Algorithms are increasingly used to shape a whole range of social policy issues in the States. This podcast looks at some of the risks attached to their use. Much has been made of their application to stock market and derivatives trading notably in Michael Lewis’s book “Flash Boys”. Cathy O’Neil does talk about that but also looks at how algorithms have been constructed to predict recidivism rates as an aid to sentencing and as ways of assessing teacher competence to manage performance and ultimately to dismiss those that the algorithms show are failing.
What Cathy does, is show how the structure of these algorithms relies on assumptions which replicate racist practice in relation to sentencing and provide virtually useless information about the performance of teachers. It addresses the distinction between prediction and causation which in the age of big data is in danger of being lost, at some cost.
If you have never heard a Econ Talk podcast I recommend them generally. The host, Russ Roberts is unfailingly courteous to his guests but takes a very orthodox liberal view of economics and the wonder of markets. For example he thinks that the state has played a bad role in relation to education in the United States and it would be better provided by the private sector. You might occasionally get irritated but it is good to try to avoid confirmation bias and he does get some fascinating speakers on issues economic. Having said that it is a wide definition of economics that is adopted.