There are a number of differences
First, the voter does know their vote record reflects their intent. They can't actually know that with a paperless touch screen.
Second, we do have the ballots and various things can be done to resolve an election, and they can be done in court, by humans. Trust can be established there too.
Now, we don't always need that, but it's extremely important that we have it, or the systems simply will be gamed. I would. What's to really catch me?
You bring up a great secondary point. Why trust computers at all?
In this case, both you and I have a dialog, and it needs to be sane. The machines could be coughing up a lot of garbage, but we both know they don't as the chat is coherent.
WHile this ambiguity has always existed with computers, most use cases get past it in similar fashion. Banking, by the way, resolves down to personally identifiable records, or at the least an account known by someone, and distributed copies of said records too.
All of that must make sense, or there are problems. It's subtle, but we catch a lot of trouble this way.
Voting is an edge case, and it's extreme as well as important, and a significant issue when it's not trusted.
Billions and Trillions of dollars can be impacted by votes. To think we aren't seeing manipulation is crazy.
In the case of SC, we don't actually have the means to catch this being done!
The niche case of voting breaks down to anonymity being coupled with no public record of the vote and this ambiguity of input.
In my State, Oregon, we use a hybrid. Voters do cast paper ballots and have plenty of time to do so. Individually, we can understand our votes are accurate.
They are verified by signature, which preserves the secrecy of the voters. It's damn tough to get data out of that part of the system specific to a voter.
We then publish voting and validity status for everyone too. So individually, and collectively, we can understand the votes are valid and will be counted.
For those who check, they can remedy things too.
Then, counting is done electronically coupled with blind audits done by hand.
This question is of paramount importance: How can you have any confidence in any system, unless you wrote it yourself? And deployed yourself, and completely control 100% of the system it's deployed on.
For most use cases, it's not an issue for reasons I've given. Votes are an edge case we do not bump into very often.
And that status makes this a difficult discussion.
Couple it with the strong incentives to control elections, and it's not something we should take lightly, and given we have plenty of time tested, cost effective, reasonable means to do this, we hardly require that risk be taken.
That it is consistently taken is a bother to me. There is no reason for it.