This
is the answer. You should never trust any one given voting machine. But it is much, much harder to corrupt large consensus networks, and blockchain based voting lets you maintain pseudonymity and it lets
you verify the results with whatever computer you want.
Some neat features you can bake into a blockchain based voting protocol:
* You can have temporary identifiers in the confirmation stage. IE, you cast your vote, you can use other computers for the first couple minutes to verify that vote, but after that your vote is added to blocks as a signature rather than direct identifier tied to the block, meaning once committed neither you nor anyone else has a concrete correlation between you and your vote.
* Alternatively, you can have a dual ledger of voter and vote, where the two are not correlated, but you can identify who has or has not voted but not what they have voted for. This gives you less verification integrity but guarantees anonymity to prevent voter coercion.
* You can have kill switches built in, generated hash keys at vote time that will cancel a vote out unique to the vote, either in the confirmation phase or even when committed (basically the same as adding something to a balance and then removing it again).
* You can use one single protocol implementation for all public record voting, and then have membership restricted groups that define constituencies through sidechains. That way you can leverage an international network of computers to secure all votes you want secured, rather than have easily 51% attacked small chains for each vote or constituency. It would be distinct from trustless models in that you would need some kind of parent organization to arrange membership in voting groups and to establish vote blocks.