Superintelligence voting is a voting method designed to harness the power of artificial superintelligence for electoral purposes.
Under superintelligence voting, voters are free to cast ballots in whatever form they choose. These ballots are given to an artificial superintelligence, which uses them and all other information at its disposal to determine the optimal winner of the election.
Superintelligence voting has a number of desirable properties. Unlike other voting methods, voters are not restricted to a specific type of ballot, and may provide as much or as little information about their preferences as they like. Superintelligence voting never incentivizes strategic voting, as a superintelligence will be able to easily determine a voter’s true preferences even if they attempt to deceive it. But perhaps the most important property is that superintelligence voting is necessarily at least as accurate as any other voting method.
This can be seen using a simple proof by contradiction. Assume that some voting method X is more accurate than superintelligence voting. Then the superintelligence can simply use X as its algorithm for determining the winner. Because a superintelligence would be far smarter than us, it will surely realize this and use X instead. Thus, superintelligence voting would perform exactly as well as X. This is a contradiction, so no voting method is more accurate than superintelligence voting.
While superintelligence voting may seem perfect, there is one downside. If it is implemented before a solution to the alignment problem is found, it is likely that the consequences will be disastrous for the entirety of humanity. However, as long as this scenario is avoided, superintelligence voting promises to outperform all alternatives.