New Voting Math Makes Sure Small Groups Aren't Ignored

Most fairness rules for voting were built for a single election: one vote, one winner, everyone goes home. But real life rarely works that way. A city council approves a budget every year. A board hands out grants quarter after quarter. A website picks which creators to promote week after week. Over dozens of these repeated decisions, a group can lose every single time even though it makes up a big chunk of the voters. That is the problem this paper tackles.

The paper introduces a rule called "temporal FJR+" (think of it as a fairness promise that stretches across time). It says a group of like-minded voters should win roughly its share of the decisions, and it's flexible about how — different members of the group can be satisfied by different rounds. The author proves this rule can be computed and verified in "polynomial time," which is math-speak for: the computer doesn't slow to a crawl as the number of voters grows. Verifying it is even simpler — just count approvals below each satisfaction level.

Here's the counterintuitive part. The standard method, Proportional Approval Voting, can get stuck in a bad outcome where a group is unfairly shut out. The fix isn't to add more rules — it's to subtract a small amount tied to the least-satisfied voter. That tiny penalty makes every unfair outcome easy to spot and easy to fix with a single swap, which is why the whole process runs fast.

The catch: this is theory, not yet tested in a real election, and fairness rules compete with each other — this one isn't strictly better than all alternatives. Still, as software and AI systems increasingly decide who gets money, attention, or slots on a schedule, having a fairness rule that can be quickly checked by an outside auditor matters more than ever.