The best reason to use pen and paper is to ensure that the process is auditable and legible to the most number of regular citizens, not only for security and integrity, but for perception and public trust, which are arguably the most important backbones of democracy. (We may have low confidence in some aspects of our system, like gerrymandering and voter roll purges, but we almost take for granted that once one is in the booth, there's a high reliability that one's vote will be counted correctly.)
Well that said, I don't think its the "best" solution per se. It is wasteful -- a lot of paper must be produced for this single task. It is not completely free of loopholes, but they are burdensome and can be verified by a second count. And I can understand why countries like e.g. India or Indonesia use computers in this situation, millions upon millions of people cast their vote over a limited time period -- it's a logistical nightmare.
But it is far better than trusting a computer, that by definition "can do everything computable". And falsifying an election is definitly computable.
Tampering with paper elections is extremely easy. You can throw out entire ballots, fill in choices where left blank, or force the ballot into an invalid state.
Not to mention the task of counting ballots using people necessarily requires splitting up the task and trusting people.
Electronic voting systems are supposed to address many of these problems, especially as processing the election becomes laborious. We don't need to give up on electronic counting solutions. We just need to make them impervious to tampering and/or verifiable.
>But it is far better than trusting a computer, that by definition "can do everything computable". And falsifying an election is definitly computable.
Tallying up votes by hand is also computation, you're just switching silicon for people.
Hard to do this while you're being watched by observers though...
And if you need to recruit many people to make your tampering scheme work, then it's much harder to keep it a secret.
This is still the most optimal current system as it much harder to corrupt at scale. The proprietary systems we have can easily be corrupted at the time of aggregation. There have been enough examples of people demonstrating simple tally hacks that there is no way to know if any election is genuinely counted accurately.
That's not to say that its impossible. I think block chain has some promising potential here but databases that are proprietary are completely untrustworthy today.