New research says the field of AI has long relied on a “single truth” paradigm, where each input is expected to have one “right” label. The researchers argue that this approach breaks down in areas where judgments are subjective, including ethics, harmful intent and the character of social interaction.
The source says that even when there is a single ground-truth, it may not be possible to measure it. It also says that AI systems moving into more subjective areas need benchmarks that reflect the complexity and different perspectives behind human disagreement.
Instead of focusing on the “forest”, the research recommends embracing the “tree” so practitioners can build more reliable tests. The source says this could help create benchmarks that are more reproducible and less expensive to design.
The research also says that understanding why humans disagree is just as important as knowing where they agree, and that it provides mathematical tools to capture both.
Source: research.google.
Companies can share verified announcements through Newz9’s international press release submission page.
