AI 비디오 인터뷰 평가

In an asynchronous, or one-way, video interview, no one is on the other end. The candidate sees a question, gets a short preparation time and records an answer, often with limited retakes. AI scoring then runs on the recording. Speech recognition converts it to text. Language models or classifiers analyze the content for signs linked to job competencies, such as describing a specific situation and its result. Some systems also use vocal features such as pace or pauses. Models trained on human ratings or on answers from high performers produce scores, and recruiters focus on the top-ranked candidates. Early systems also analyzed facial expressions, and that became the biggest controversy. In 2019 the Electronic Privacy Information Center filed a complaint with the US Federal Trade Commission about HireVue, and in early 2021 HireVue said it had stopped using facial analysis. Also in 2021, an investigation by the German broadcaster Bayerischer Rundfunk reported that one video-assessment tool changed personality scores when the same person wore glasses or a headscarf or sat in front of a bookshelf. Speech recognition raises its own concerns. A 2020 study found that major commercial systems made substantially more errors transcribing Black speakers than white speakers, and errors in the transcript carry through into content scores. The law is catching up. Illinois' Artificial Intelligence Video Interview Act, in force since 2020, requires notice, an explanation and consent. Maryland requires consent before facial recognition is used in interviews. New York City's Local Law 144 requires bias audits and notices for automated employment decision tools. The EU AI Act classes AI used in recruitment as high-risk. This is different from resume screening, which scores documents. Video assessment scores a performance. For candidates, the practical advice follows from how the systems work: answer the actual question with a clear structure, make sure the audio is clean, and ask for accommodations if you need them.