Earlier quoted context omitted.
Disagree. It's enough for the average voter to trust that some other people - independent experts - are able to verify the vote. Not everyone needs to be an expert at anything. I wrote more about this trust aspect in the appendix of my thesis on voting: https://attejuvonen.fi/thesis
Yes, but then all that's needed to attack the voting system is to trot out your own experts that voice disagreement. Without the means to assess the system for themselves, voters will lose trust in it. Especially in this day and age, when trust in institutions and expects in general is extremely low. (Heck, this attack already works to some extent with the current, extremely transparent system of ballots)
Which is precisely what Trump did in 2020.
I doubt people made up their minds on whether the 2020 vote was sound based on the mechanics of how the votes were counted. The counting procedure with it's interlocking checks is rather complex, and differs between states. They made up their minds based on what they trusted more - the Trump version of the facts or the testimonies of the people counting the votes and those administering it.
It will be exactly the same with computerised voting. Ideally the software will be open source with reproducible builds. Just as with the present voting system most won't be able the check the actual mechanics themselves, but they likely know someone who knows someone who knows someone who can.
By the by, it wasn't done that way with computerised voting and probably still isn't in many places. I vaguely recall the story of a voting machine breaking down, a technician waving his magic wand over it after voting had closed, and a whole pile of votes fell out. It made the people in charge of the voting process distinctly uncomfortable.