Blockchains with zk proofs fix this tho.

That's literally how Dominion voting machines work.

And hand recounts are only for verification, you shouldn't rely on them because they're far less accurate.

> And hand recounts are only for verification, you shouldn't rely on them because they're far less accurate.

Voting machines can be altered. Go to any big hacker conference and you can learn how to hack the common models in thirty seconds. Humans may be fallible, but I'll always trust those fallible humans over anything automated.

And hence - use random hand counted samples for verification.

But nobody has proven that the dominion machines changed a single thing, in fact, the hand recount in Georgia showed as much. Which is why the "election fraud" claims then moved on to more outrageous claims.

The errors in hand recounts are rare because there are lots of mutually suspicious people looking at each ballot.

And when they do happen, they are uncorrelated. A single software error can easily change the result of the election; a thousand human errors cannot.

> This means initial results will be available quickly, followed by actual results a day or two later

Actual results from paper ballots in French elections are in around an hour or two after voting closes. Yes, it relies on volunteers counting and on people watching but it’s much more effective than anything that would involve machine. You can literally watch your ballot box all day if you’re suspicious.

France's voting system is bulletproof when it comes to vote integrity. However, it has one significant drawback: it is hard to ask more complex questions than voting for x/y/z or yes/no questions.

One could say that it is a good thing (since it is hard to have a good public debate on many topics at once, voting on many topics at once means you collect people's preconceived opinions rather than people's informed judgement), but many US states ask many questions at once when people vote, so they would have to significantly reduce the share of direct democracy in their system.

> However, I don't think the concept of voting machines is entirely incompatible as long as we keep the paper trail authoritative.

The problem is impossible to solve if you want to respect the following requirements:

- No one can know who you voted for

- Anyone can oversee the process and understand why fraud can't happen

There simply is no way to prove voter anonymity with an electronic machine without enough technical knowledge to audit the system.

The ones I saw seemed fine? You fill out a paper ballot, and you submit it to the worker at the end who scans it and drops the ballot in the box in one motion.

You have a paper trail, it anonymous, and it's pretty easy to understand.

I guess if you modified the firmware on the scanner to disclose the vote counters continuously, and the worker at the end knew me by sight (the worker I showed my I'd to is on the other end of the room) you could find my vote. But you could also hide a camera in the booth, which is easier.

> But you could also hide a camera in the booth, which is easier.

It is also easier to check whether there is a camera in the booth.

Is it? Cameras can be really small.

You can see it this way: if you can time people voting, you can hide the camera anywhere in the voting room. Checking the booth is always going to be easier than checking the room.

There is also the issue of voters selling their vote. It's pretty easy for them to wear a tracker that tells you when they were in the booth. On the other hand, with paper ballots, the buyer has no way to check the vote, since market ballots are null.

The first problem is a non-issue. You don't need the voting machine to do the authentication if you let a human control the ballots that get stamped/marked/whatever.

The second problem also isn't a problem, because the machine doesn't need to be right. The paper ballot is leading, the machine is just an indication. The ballots are still counted by hand afterwards using the normal process. That means that as long as citizens are able to review the manual counting, they don't need to know or care how the voting machine works.

And yes, that does probably negate 90% of the advantages of voting machines.

> The first problem is a non-issue. You don't need the voting machine to do the authentication if you let a human control the ballots that get stamped/marked/whatever.

The machine can register what was vote was cast at 13:42. From there, the whole idea of anonymity disapears.

> The second problem also isn't a problem

The impossibility to have the second criterion is a problem because it prevents the first criterion from being verifiable by anyone.