Scientists Built a Voting Formula for What Society Should Allow

Every day, someone has to decide where the line sits. Should this AI model be allowed to give medical advice? Should that speech be banned? Should a candidate win an election with 40 percent of the vote? Right now these calls are made by committees, courts, and regulators — slowly, inconsistently, and often with loud complaints from whoever loses. A group of six researchers has published a paper asking a bigger question: is there a fair, repeatable formula for deciding what society accepts?

Their answer rests on three plain ideas. First, sufficient support — if enough people back something, it should be allowed. Second, minority protection — small groups shouldn't be steamrolled just because they're outnumbered. Third, dominance — if one option is clearly better than another by everyone's standards, the worse one should be blocked. For each idea, the team built a precise mathematical rule that delivers exactly that outcome, no fudging.

The hard part is combining all three at once, because they tug against each other. Too much minority protection and almost nothing gets approved; too little and unpopular groups lose every time. So the researchers built a sliding scale: how much blocking power a minority gets depends on how much genuine approval the thing they're blocking has. Something broadly loved is hard to stop; something barely tolerated is easy to stop.

They then ran tests on five real datasets, including political elections and ethical AI evaluations, and found that methods breaking the rules in theory also broke them in practice. Important caveat: this is a 41-page academic paper, not software you can install. Nobody is governing AI with it tomorrow. But it's a serious attempt to make the messy business of drawing society's lines more transparent — and easier to argue about with evidence instead of shouting.