要在很短的面试窗口里判断陌生人的能力与品格，却要做出可能影响数十年的雇用决定，这一流程还会被紧张反应与排练式问答扭曲；在人工智能工具扩散后，甚至无法完全确定回答者是否为人类。面试官也可能因为没吃午饭且当天还剩另外五场面试而进一步降低判断质量。

2022年明尼苏达大学Paul Sackett等人的一项元分析发现，结构化面试在各类招聘方法中具有最高的预测效度，领先于测评中心或心理测验。关键在于“结构化”：对所有候选人使用标准化、与岗位相关的问题并按统一规则评分，而非结构化面试的预测效度不到结构化的一半。

即便是规范的结构化面试，其分数对真实工作绩效的解释比例也不到20%。研究还表明，将人格测验、工作样本等多种工具组合起来，或许能把可预测比例提升到30%–40%，因此更现实的做法是叠加评估并在入职后用培训、实习与内部流动等机制提高成功率。

A short interview window is used to judge a stranger’s ability and character, yet it can drive commitments that last decades, and the process is distorted by nerves and rehearsed answers; as AI tools spread, it is not even certain a human is answering. Decision quality can also drop when interviewers are hungry and still have another five interviews to run that day.

A 2022 meta-analysis by Paul Sackett and co-authors at the University of Minnesota finds structured job interviews have the highest predictive value of any recruitment method, ahead of assessment centres or psychometric tests. “Structured” means standardized, job-related questions asked of every candidate and scored by an agreed system, while an unstructured interview has less than half the predictive validity of a structured one.

Even textbook structured-interview scores account for less than 20% of actual job performance. Research suggests layering measures such as personality tests and work samples may raise predictability to about 30–40%, so realism, multiple assessments, and post-hire onboarding, training, internships, and internal applicants are emphasized as ways to tilt outcomes toward success.

Source: How to conduct a job interview

Subtitle: The worst way to select people, except for all the others

Dateline: 12月 18, 2025 05:44 上午