4 ms·
I have an "engineering reason" for you. Look at this as a once-in-a-lifetime chance to redesign the user-interface and API of a major product with a huge and di
by Terr_ 8y ago
I have an "engineering reason" for you. Look at this as a once-in-a-lifetime chance to redesign the user-interface and API of a major product with a huge and diverse set of extremely conservative users.
1. You need to support transitional back-ends for certain users (e.g. instant runoffs)
2. You strongly want to avoid a "new system" that cannot be replaced and has no room for expansion.
___
Any voting system able to handle ranked-preference voting can trivially be "dumbed down" for a certain set of customers who want approval-voting -- it's almost as simple as mapping integers to booleans.
In contrast, if everything is built assuming inputs are boolean values for each item in a set, your system is crippled. You either have to hope that everybody loves approval-voting, or admit defeat and switch it back to the old (i.e. current) stupid system of "pick one".
- AstralStorm 8y agoOr you could go for an even richer system of score voting with integers out of a limited set. You can convert ordinal ranks to scores directly meaning the set exactly same as number of candidates. By that, each median score represents median ordering in a ranked ballot. You may lose certain properties that are relevant in case of direct ties but nobody really cares about that in big constituencies. Heck, you can also keep counts of each score in case the dominant is needed for a tie breaker. Approval voting is a score system with a binary score. Rules are simple: in case of rank scores, candidate with lowest median rank wins, if there is a tie candidate with higher count of low scores wins. This converts IRV or other ranked system to Bucklin vote. In case of approval scores, replace lowest with highest.