Voting Machines and Ballot Types
-
@rob said in more than one thread that there are voting machines that can handle ranking ballots, but no voting machines that can handle rating ballots. This is to argue that voting machines designed for FPtP can handle rating, provided that the number of different possible ratings is small (I'm thinking seven or fewer, and certainly not 100).
Treat, on the unfilled ballot forms, and in the instructions given to the machines, treat each candidate as a "race" and treat each possible score as a "candidate". The machines will then tally how many voters picked a given score for each candidate. The summary data, then, tallied by the machines and the procedure around them having been built up for FPtP, will suffice to allow anyone to complete a Score tally.
Approval-ordered Llull (letter grades) [10], Score // Llull [9], Score, STAR, Approval, other rated Condorcet [8]; equal-ranked Condorcet [4]; strictly-ranked Condorcet [3]; everything else [0].
-
@jack-waugh I feel like I responded to this previously, but maybe not.
My statement that machines and infrastructure doesn't handle score ballots is based on this thread from EndFPTP:
This is user choco_pi, who has impressed me previously with their knowledge, but that's the only degree I can attest to this. I could be wrong. But I'd be more impressed with your proposed solution if you would address their points.
Hopefully it's ok if I post the comment directly here since I can't seem to directly link that exact comment. It is in response to user manageorigin. I emphasized a couple places that most directly address the issue.
But it's said that current machines can handle score voting. Do you have proof on your statement?
It's hard to respond to claims like this because they're just, super wrong. Like it's so disconnected with the facts on the ground it's difficult to know where to begin.
Voting systems are not just a single machine, but the entire chain of hardware, software, instructions, and protocols covering everything from ballot creation, vote casting, ballot scanning, adjucation, reporting, verification, and auditing. For federal and state elections, the entire system A-Z has to be certified according to law, including equipment testing at designated labs. This includes not just security, but a litany of accessibility concerns.
There are only 11 organizations even registered with the EAC as qualified to even submit a system for certification. (Of these, Vidaloop is just a startup seemingly focused on mobile voting, so it is really 10.)
3 of these, ES&S, Dominion, and Hart, make up over 92% of the US market. All have end-to-end RCV ballot support on their current generation machines. Unisyn and Avante do as well.
The software for Clear Ballot and various legacy systems by the aforementioned vendors (such as Sequoia, who I believe Dominion purchased) do not support ranked ballot tabulation directly, but export ballot-level results that can still be tabulated externally. To plug this gap, RCVRC has invested a lot of resources into supplying an open source version of this missing piece and getting it certified in multiple states. (So that one Clear Ballot machine in Sheboygan Country doesn't block the entire state of Wisconsin) Smartmatic and VotingWorks are probably in a similar boat, but it has not yet come up that such an implementation has needed certification. (VotingWorks is itself open-source, so already certified functionality could just be integrated directly anyhow.)
Conversely, 0 of these 11 vendors support scored ballots. Not the ballots, not the voting machine UI, not the tabulation, not the reporting, not the auditing; none of it. You could try do an externally tabulated workaround, but in this case I think you'd find the reality of that compromise to be unacceptably ugly, 10x worse than the RCVTab stopgap. (A ballot with a question asking which candidates you would like to score 0, a seperate question asking which candidates you would score 1, etc.)
Approval sneaks in because you can take frequently take a plurality systems and just turn off overvoting detection. You still have a lot of work to do--interface software has to support it, accessibility instructions have to still be valid, LEO instructions need to be updated--but this is a realistic, immediately achievable set of changes if state law allows it.
Warren Smith yelling that there is no reason vendors can't make voting machines that support score ballots is like yelling that there's no reason car manufacturers can't make cars that run on vegetable oil. I mean... sure. But, they don't, which is the topic at hand.
