The internal state of the machine is not observable and not auditable. Anything stored in a computer can be falsified and the voter cannot know that it recorded their vote correctly. Unless the printing you are referring to is that it prints a paper ballot for each voter that they can check and then drop in a ballot box, then there is a step in the process that scrutineers can’t have eyes on. Even if the software is open source and the scrutineers got to watch it being compiled and installed on the machines, there are so many unobservables. Computers are literally billions of times more complicated and harder to audit than pencil and paper.

The only electronic voting system I would even begin to entertain trusting is one that basically just fills the ballot out for you to make sure it is legible, and that you can validate the paper ballot. Such a system could then also keep its own count, giving a much quicker unofficial count before the paper ballots are counted, but I don’t really think it’s worth it. It is still an extremely complex machine that it would be easy to hide some sort of surveillance/wireless connection in, and anyone who can inspect the running total counts can also use that to determine how each person that goes through the booth voted.

There is value in the brutal simplicity of pencil and paper, folding your ballot, and putting it in a box. It’s a system anyone can understand and that the human eye can audit.