3 ms·

> Not only should the voter be marking their ballot by hand not doing so through a machine whose UI may have bugs, but any scanner must scan the human-readable

> Not only should the voter be marking their ballot by hand not doing so through a machine whose UI may have bugs, but any scanner must scan the human-readable marks, not some machine-readable marks because the voter can't check that both sets of marks match. I don't think I agree with the first part. Marking by hand is probably more error-prone. People are notoriously not great at correctly filling in a bubble, leading to scanning errors. Any UI bug should be caught thanks to the human-readable printout. If I select Biden and my printout says Trump, I'll notice. The second part, I think I agree. It should print out a perfectly bubbled-in human-readable ballot. That would seem to remove both human bubble-filling error and QR-code shenanigans. > why does that not extend to people who make black-box machines that have a ton of chances to change the vote? The one and only added chance to change the vote is to alter the QR code to something other than the human-readable selections, and this is trivial to check by having humans sample a few thousand ballots. The other way to change it would be for the QR reader software to lie about what it read, but you could do that with traditional bubble-ballot reader software too. In both cases, having humans sample the ballots is the solution.

LawTalkingGuy 3y ago
What problem are you trying to solve with e-voting? To make the vote more accurate by cutting down on UI-type errors? Get results sooner? > > Not only should the voter be marking their ballot by hand not doing so through a machine whose UI may have bugs > I don't think I agree with the first part. Marking by hand is probably more error-prone. People are notoriously not great at correctly filling in a bubble, leading to scanning errors. Perhaps, but risk of systemic failure allowing a vote to be stolen is the paramount concern. UI errors (esp on randomized ballots) tend to balance out. The UI concerns are real, but they can both be addressed physically and through ballot design. > the QR code to something other than the human-readable You wouldn't want to have a QR code though, if you simply read the human-readable marks. The QR code is some sort of premature optimization based, I think, on the assumption that scanning is difficult. If you have only one mark then you know it's the canonical mark. > Any UI bug should be caught thanks to the human-readable printout. If I select Biden and my printout says Trump, I'll notice. Real ballots often have twenty or more choices so this is harder in practice. And you're trusting a machine's UI at each step. It's telling you which marks mean what rather than you marking the paper according to instructions on the paper. A paper ballot can't be altered after the fact to change your vote, any machine-generated paper can simply be reprinted. The correct answer, if anything is needed at all, is an under-vote detector. It doesn't tell you how you voted, it just says "no selection is readable for Mayor" or "You only made four marks for City Councillor but are allowed to place up to five votes", and so on.

LawTalkingGuy 3y ago

What problem are you trying to solve with e-voting? To make the vote more accurate by cutting down on UI-type errors? Get results sooner? > > Not only should the voter be marking their ballot by hand not doing so through a machine whose UI may have bugs > I don't think I agree with the first part. Marking by hand is probably more error-prone. People are notoriously not great at correctly filling in a bubble, leading to scanning errors. Perhaps, but risk of systemic failure allowing a vote to be stolen is the paramount concern. UI errors (esp on randomized ballots) tend to balance out. The UI concerns are real, but they can both be addressed physically and through ballot design. > the QR code to something other than the human-readable You wouldn't want to have a QR code though, if you simply read the human-readable marks. The QR code is some sort of premature optimization based, I think, on the assumption that scanning is difficult. If you have only one mark then you know it's the canonical mark. > Any UI bug should be caught thanks to the human-readable printout. If I select Biden and my printout says Trump, I'll notice. Real ballots often have twenty or more choices so this is harder in practice. And you're trusting a machine's UI at each step. It's telling you which marks mean what rather than you marking the paper according to instructions on the paper. A paper ballot can't be altered after the fact to change your vote, any machine-generated paper can simply be reprinted. The correct answer, if anything is needed at all, is an under-vote detector. It doesn't tell you how you voted, it just says "no selection is readable for Mayor" or "You only made four marks for City Councillor but are allowed to place up to five votes", and so on.