As algorithms increasingly assist and influence human decision-making, we’re finding more examples of high-stakes bias. Computer programs can only be as “fair” as the programmers who select the training data.