Wednesday, September 14, 2016

Math is Not Racist, But Big Data Cannot Predict Individuals

Under the provocative title "Math is Racist," CNN Money reports on the work of mathematician Cathy O'Neil.  She is distressed with the way mathematical models about groups are then used to hurt individuals.  For example, the fact that the relatives of prisoners are themselves more likely to commit crimes can be used in sentencing a specific relative of a prisoner.

O'Neil is right that this is wrong, and is an injustice spurred by the increasing availability of Big Data.

The problem is one we face in sociology all the time.  I have for years made my students repeat Rule #1 of Sociology (according to me):  We make generalizations about groups, which do not necessarily apply to each individual in the group.  So, for example, it is true that the relatives of prisoners are more likely to commit crimes.  Finding that fact has been a great and important achievement of sociology.  But it would be wrong to conclude that this particular relative of a prisoner is, therefore, more likely to commit crimes.  Sociology cannot provide that fine-grained a solution.

Or, to use the simplified rule that I now teach students, sociologists are the people who understand the difference between most and all.

No comments: