Australia doesn't use STV, which people here are advocating. (Except in a couple of states.)

It uses AV, which has instant runoffs but not multi-member seats. (They also have some unusual twists like "above-the-line voting", where the first-preference party, rather than the voter, decides who the vote gets the vote when the first-preference party gets knocked out.)

AV isn't a remotely proportional system, whereas STV is more-or-less proportional, provided the seats have enough members.

In this case it sounds like they've already got a multi-member constituency, so they just need to add the candidate ranking / instant-runoff aspect and they'll be done.

The + or - thing (Forward or Backwards, Heads or Tails, or whatever metaphor the interaction designer recommends) would also work fine with a single choice amount multiple candidates without using an ordering mechanism, as long as you don't mind giving people the option to vote against somebody instead of for them. (In other words, unpopular candidates could end up with a negative total.) No particular reason why you shouldn't do this - it just hasn't been done much before, probably just because the pre-internet logistics were too difficult.

Well, there are ways of making the interface help you figure out the real vote thing. For example, you could call the Forwards Votes "Heads" and the Backwards Votes "Tails" and have a screen with the controls in the middle, and the meanings for "Heads" on one side and the meanings for "Tails" on the other side. This problem should be fairly trivial for a decent UI designer.

As far as multiple candidates go, most sensible electoral systems let you put the candidates in order, which would work absolutely fine. You'd order them on the screen, with them appearing in one order on the left and the other order on the right, and only you know whether it's the left or the right version that matters.

It may be true that some people will still be coerced even though they have the means to defeat the coertion, but that's true with secret ballots as well. (And it may not be true either; Coercing people may tend to backfire if the system gives people a good way to be bloody-minded.)

Obviously some people will forget, and have to go back to the booth; I think if you're only going to ask people to vote once every 4 years, you're probably best keeping computers out of it. But with a decent online system, you have the option of consulting people far more often than you do currently, since you're no longer hampered by a system designed to work around 18th Century logistics.

You can solve this problem with Backwards Votes. Here's how it works:

You have to register once (only once) in a controlled location, where somebody makes sure you're not videoing it and there's nobody watching over your shoulder.

When you register, you decide (or the system randomly decides and tells you) whether your vote should be a Forward Vote or an Backwards Vote. Only you know whether your vote is Forward Vote or a Backwards Vote, and it will never be shown to you when you go to vote online, or anywhere outside the secure booth where you registered.

When you vote online, you have the option of casting a vote for a candidate or against them. A vote for someone increases their vote by one, and a vote against them decreases their vote by one. But if you have a Backwards Vote, a vote for would decrease their total, while a vote against would increase it.

If someone tries to bribe or intimidate you to vote a certain way, they have no way of knowing whether they are forcing you to vote for their favourite candidate or against them. The same goes for malware on your client or between the client and the server: Since the information about whether your vote is Forward or Backward is only stored on the central server, the attacker only has a 50% chance of getting the outcome they want, cancelled out by a 50% chance of shooting themselves in the foot.