[sopooneo]

But why not scan-tron? It seems like that is the best of both worlds. Start with a simple physical paper that everyone can understand and audit as the primary record, but then have each voter feed their own directly into a scan machine that can tally and send counts in quickly?

That's how we do it around me, and it seems like a strictly superior approach. What am I missing?

[EvilTerran]

Scantrons are a big improvement over electronic-only counting, but they can struggle with imperfectly-filled-in ballots... another good alternative, which avoids that particular problem, is to have a machine where you make your selections by button-press - but it prints them onto a physical card, lets you verify the printout through a window, then drops it into a built-in ballotbox.

The machine itself can keep count, or the cards could be designed for scantron-esque machine counting - regardless, in case of a disputed result, the cards can be counted the traditional way (by hand, with observers from each party present, etc).

Not my idea, BTW, but I don't recall where I read it - nor whether it was a description of something actually in use or merely a proposal.

[worldsayshi]

Or just some kind of simple mechanical stamping or cutting machine that assures that each vote is registered in the same way but in an easily understandable and transparent format.

[EvilTerran]

> cutting machine

Do you remember "hanging chads"?

But yeah, that's the crux of the idea: have a machine take the voter's choices, to effectively eliminate accidentally-spoiled ballots (the design of the "hanging chads" machines was sorely lacking on this point); but then have it produce a physical record, visually checked by the voter, to enable auditing & recounts.

Counting the hard copies would be the definitive source of truth, just like traditional paper ballots - any automated score-keeping would just be a bonus for early result reporting (although might also stand in for the manual count in "safe seats" where no-one cares to dispute the expected result).

[EdHominem]

The key to the system you describe is having the vote-punching machine right next to the vote-checking machine. That way if there's a hanging-chads issue you can catch it before it spoils more than a single vote. And the person just slides the vote down the 'discard' chute and gets a fresh one until they're happy with the results.

Have the checker mark the holes it detects with red ink or something, to make it clear to the user that the system detects their votes properly, and to provide a fallback. In the event another machine fails to count it, the user's intent is double-marked.

And then have the same style of vote-counting, where people manually scrutinize the votes, and have each party's representatives slide the votes into their counting machine. If the machines lose sync, you stop and figure it out at the point of the specific vote that fails to scan.

[ecnahc515]

The scantron machines are expensive and it's not really feasible to have at every polling location. Lots of polling locations are at churches a d such anyways. Additionally, scantrons can be fairly slow depending on the machine.

[mjcl]

They do in some counties. I'm a pollworker in Alameda County and every precinct uses an optical scanner (made by Sequoia). At the end of the night the memory pack is removed and taken back to a return center to be counted. The paper ballots are boxed and sealed in case there are any problems with the electronic tally.

Speed-wise, it only takes the machine about 1 second per ballot card.

[makeapoint]

Just to be clear, you are arguing that that scantron machines are expensive and not feasible at every polling location so instead we should provide dozens of more expensive electronic voting machines at each polling location?