Researchers suggest limiting the possible results from an algorithm, particularly when non-experts are at the controls. Source: Wired
