There is evidence that even very simple algorithms can outperform expert judgement on simple prediction problems. For example, algorithms have proved more (a) accurate than humans in predicting whether a prisoner released on parole will go on to commit another crime, or in predicting whether a potential candidate will perform well in a job in future. In over 100 studies across many different domains, half of all cases show simple formulas make (b) better significant predictions than human experts, and the remainder (except a very small handful), show a tie between the two. When there are a lot of different factors involved and a situation is very uncertain, simple formulas can win out by focusing on the most important factors and being consistent, while human judgement is too easily influenced by particularly salient and perhaps (c) irrelevant considerations. A similar idea is supported by further evidence that ‘checklists’ can improve the quality of expert decisions in a range of domains by ensuring that important steps or considerations aren’t missed when people are feeling (d) relaxed. For example, treating patients in intensive care can require hundreds of small actions per day, and one small error could cost a life. Using checklists to ensure that no crucial steps are missed has proved to be remarkably (e) effective in a range of medical contexts, from preventing live infections to reducing pneumonia.
* parole: 가석방 ** salient: 두드러진 *** pneumonia: 폐렴