Weapons of Math Destruction: 8 key ideas
The core ideas of Weapons of Math Destruction, each in under a minute, written in our own words. Want the full depth? Get the book.
- IDEA 1 OF 8
Algorithms can harm people
O'Neil, a mathematician, argues that many algorithms used to make decisions about people are unfair and harmful.
She calls the worst of them weapons of math destruction. Their harm often falls on the poor and the powerless.
- IDEA 2 OF 8
Three features make them dangerous
O'Neil defines harmful models by three features: they are opaque, they operate at large scale and they cause damage.
People affected often cannot see or challenge them. Together, these features make mistakes hard to find and fix.
- IDEA 3 OF 8
Bad models feed on themselves
O'Neil explains that harmful algorithms create feedback loops.
Their predictions shape outcomes, which then appear to confirm the predictions. A model that unfairly labels a person can make their life harder, which then makes the label seem true.
- IDEA 4 OF 8
Teachers were wrongly judged
O'Neil describes how teacher evaluation models in America led to good teachers being fired.
The models were based on flawed statistics. Their jobs were lost over scores that experts say were too noisy to trust.
- IDEA 5 OF 8
Policing models target the poor
O'Neil describes how predictive policing sends officers to poor neighbourhoods, leading to more arrests there and reinforcing the pattern.
More arrests in those areas then feed back into the data, so the model keeps pointing to the same places.
- IDEA 6 OF 8
Scores shape opportunity
O'Neil describes how credit scores and similar scores affect jobs, loans and insurance.
Hidden scores can trap people. A low score can limit choices in one area, which then lowers scores in others.
- IDEA 7 OF 8
Hiring tools filter unfairly
O'Neil describes how automated hiring tools can screen out candidates unfairly.
Personality tests may discriminate. A person may be rejected by a machine before any human reads their application, and nobody explains why.
- IDEA 8 OF 8
Algorithms need oversight
O'Neil argues for auditing algorithms and holding their designers accountable.
She calls for fairness to be built in. In her view, people harmed by a model must be able to question it and get a fair answer.
Get one idea like this every daySwipe through ideas, go deeper on any of them, and save the ones that matter. Free.
Read Weapons of Math Destruction in the app