New Research Reveals How Fair Ranking Votes Can Really Be

We vote all the time—but not just for one winner. Think of a team rating pizza places, an awards committee ranking finalists, or social media crowdsourcing a list of “best movies.” All these systems turn people’s ordered preferences into a single overall ranking. The question: how good can that ranking possibly be, when voters only tell you their order, not how much they actually care?

This paper tackles that question using an idea called “metric distortion.” Imagine each voter has a true “distance” to every candidate—how much they truly like them. But the voter only gives you a list from favorite to least favorite. The system then creates a full ranking. Metric distortion measures the worst-case gap between the ranking the system produces and the best possible ranking you could make if you knew all the true distances.

The researchers found some clear limits. If you know how much importance each voter puts on first place versus second place, you can build a ranking that is at most 3 times worse than the ideal—and no method can beat that. That’s like a GPS route guaranteed to be at most 3 times longer than the perfect path. If voters have different, unknown priorities, there is no guarantee at all unless you add fair assumptions—and then the best you can promise is directly tied to the number of candidates.

This is theoretical math, not a consumer app. But it matters because real systems—scholarships, hiring pipelines, recommendation lists—silently struggle with this problem. Knowing the hard limits helps designers stop chasing impossible perfection and instead communicate honestly about accuracy, or ask better questions to improve outcomes.