The author argues that faster AI-generated results do not necessarily translate into higher overall productivity. The time, cognitive cost, and sustained attention required to verify outputs may offset efficiency gains. In one randomized trial, experienced software developers took 19% longer after adding verification steps, yet they still subjectively believed they were working faster. The author recommends providing employees with the time, training, and permissions needed for verification while they remain accountable for AI-assisted work, and measuring productivity by the credibility and total cost of the complete human-AI collaboration process.
Source:The Conversation 科技版 · Read the original article
Leave a Reply