Many critiques of our increasingly algorithmic world focus on where algorithms fail. One boils down to simple math: Some sample sizes are so small (of terror attacks, for instance) that algorithms can hardly reliably predict what they purport to predict. Another boils down to psychology: As anyone who has ever become lost by blindly following Google Maps can attest, people tend to perilously ignore common sense and trust the machine. A third boils down to checks and balances: Algorithms are often proprietary black boxes, closed to outside scrutiny. A fourth critique points out that algorithms, far from being objective, often just encode and scale up human biases: If a predictive policing system learns that most of a city’s arrests have historically been in a certain majority-Black neighborhood, the computer may decide to deploy more officers to that neighborhood, perpetuating a racist pattern of arrests and violence. Garbage in, garbage out.