Future of work
Amazon explains the science behind its AI interviews
- Published
- Source
- AWS Machine Learning Blog
Summary
Amazon has published how Amazon Connect Talent, its service for AI-led structured behavioral interviews, is designed. Interviews are built around job competencies with adaptive follow-up questions, the interviewer only gathers evidence while a separate process scores each competency from 1 to 5, and video, voice tone, and appearance are excluded from scoring. In an evaluation by an independent industrial-organizational psychology firm commissioned by Amazon, 98 percent of generated questions targeted the intended competencies and candidate scoring matched expert decisions 89.3 percent of the time (Cohen's kappa 0.79). The evaluations were commissioned by Amazon itself, a limit the post leaves in place.
Why it matters for our work
It shows how AI can enter hiring, a task built on human judgment, through principles like separating evidence gathering from scoring. For organizations hiring at scale, it is a case to study for expanding interview opportunities while keeping evaluation consistent.
Translated from the Korean original. Summaries may be translated and edited. Commentary reflects our perspective; forecasts remain the source’s views.