I work with digitisation in the public sector of Denmark. We’ve digitised our elections, but we’ve digitised the part that makes sense, the registration you do before you’re handed you ballot.
In the old days, we used to have big books where you’d get crossed off after you were identified. This naturally takes a lot of time, so today we print a little bar code on the piece of paper that we mail every adult citizen at every election. This means that we can scan you instead of manually crossing you off in a book.
We still have queues at prime time, but they are 10-15 minutes instead of two hours.
The actual voting is done with paper, so that there is a paper trail.
This is the only thing that makes sense. Especially when you look at the business side of things. We reduce the hassle for citizens (our customers of sorts) and we maintain security. Sure we could provide results faster if we counted votes digitally, and you could frankly also provide a paper trail if the machine printed you vote, but does speed of counting really matter? Financially digital vote counting would be insanely more expensive, because public IT systems are insanely expensive and paying staff a little extra to count votes isn’t.
I mean, the registration system is really expensive as well, but at least it benefits the citizens, so that is a reasonable sacrifice to us. But digital voting? That’s as you put it, insane.
It’s not a democratic process if you don’t have the physical votes and a system which makes sure they aren’t tampered with.
In Australia we have the staff still ruling us off in the electoral role. That usually takes a minute or less.
The entire voting process (including queuing) depends upon the popularity of the individual voting booth and time of day, but is usually less than 10 minutes. This may be because there are an adequate number of booths and trained staff.But it is also because of compulsory voting.
The highly likely attendance numbers per booth and their distribution across the day are known and can be planned for, unlike some other more random systems.
As a Dane, our palementary elections gets about 86-90% participations, and not having voting mandatory means you get an effective signal for how the population feels by how well they attend.
We are due for a new election before july, so I will probably just save the queue and vote by mail, if I elect to vote at all.
There's another way to use computers to help with voting.
Have Ballots with a unique identifier. People come to a polling station, get a ballot, fill in their vote.
The ballot goes through a scanner to tally the vote, and then goes into a standard vote bin.
At the end of voting, you cross-check a random sample (both ways) and check the total number of votes matches between the scanner and bin.
If all goes well, scanner results get electronically combined. If the sampling shows an error, count by hand.
One extra addition. Your ballot is filled out by a separate printer. This ensures proper readability at the scanner, and allows placing the unique ID after someone gave you the ballot (to keep your vote secret). Any tampering with non-unique IDs is detectable by the random sampling.
IDs on ballots don't make sense. You cannot know your ID without breaking a requirement for good free voting systems: It shall not be possible to prove to others how you voted. This is to prevent forcing or purchasing votes.
If you place the ID after the ballot is handed out (by a printer that is also used to fill in the ballot). Then this systems still doesn't allow proving of votes.
The ID here is meant to identify a ballot, not a voter. It should probably be something like a UUID. The aim of this system is to allow cross-checking between the scanner and the physical ballots.
I got that, but you can still kind of prove it. Your know your ID + your-vote. This is likely the only valid ID+vote combination you can know before results are counted. That's when I'd "ask" you and late verify it.
If you want to verify the machine is working, just put the ballot in the standard bin and add those IDs in the counting phase. That seems fine in principle and make it easy to check the tech is working as intended. You'd end up with having list of all individual votes available, maybe even to the public. I'd be worried about people throwing statistical algorithms at that. You better also find a near perfect method to randomize order...
That's a great question. The answer depends a little bit on how many issues are on the ballot, and a hell of a lot more on whether you're asking the media, or those with a direct stake in the outcome, or those voters. Your system is optimized to serving voters.
