MIT preference model improves choices by comparing options in threes

MIT researchers have proposed a new way to model preferences by comparing options in groups of three rather than only in pairs.

That shift matters because human choices often depend on context. A third option can reveal trade-offs that a binary comparison hides, giving models a richer signal about what people actually value.

Preference modeling sits behind recommendation systems, personalization, and alignment research, so better structure here can improve how AI systems learn from human judgments.