Ask Slashdot: How Would You Use Computers To Make Elections Better?

shanen writes: Regarding politics, is there anything that Americans agree on? If so, it's probably something negative like "The system is broken," or "The leading candidates are terrible," or even "Your state is a shithole." With all our fancy technology, what's going wrong? Our computers are creating problems, not solutions. For example, gerrymandering relies on fancy computers to rig the maps. Negative campaigning increasingly relies on computers to target the attacks on specific voters. Even international attacks exploit the internet to intrude into elections around the world. Here are three of my suggested solutions, though I can't imagine any of today's politicians would ever support anything along these lines:

(1) Guest voting: If you hate your district, you could vote in a neighboring district. The more they gerrymander, the less predictable the election results.
(2) Results-based weighting: The winning candidates get more voting power in the legislature, reflecting how many people actually voted for them. If you win a boring and uncontested election where few people vote, then part of your vote in the legislature would be transferred to the winners who also had more real votes.
(3) Negative voting: A voter could use an electronic ballot to make it explicit that the vote is negative, not positive. The candidate with the most positive or fewest negative votes still wins, but if the election has too many negative votes, then that "winner" would be penalized, perhaps with a half term rather than a full term.

What wild and crazy ideas do you have for using computers to make elections better, not worse?

Comments

Take the average of the desires of the voters

[jader3rd]

Instead of voting for a candidate, have the electorate vote on a number of issues (combination of recent past issues and issues on the docket). Then take the average, and the candidate of the political party that is closest to the average, wins. Parties can do whatever they want to determine candidates.

Down side is that for those who already feel like voting is like busy homework, this will add to the load.

Re:Simple

[0100010001010011]

Paper ballot with scanners.

Rather than a mono-culture require that no 2 bordering counties may use the same brand scanner.

After the early election results are in share ballots with 2 neighboring counties to use on their machines.

If the machines report different numbers you hand count them.

Bonus companies for how accurate they are. Those that are more than 1% off get no payment for the machines and the company is not allowed to make machines for the next election cycle.

Voting Can Be Improved But Not With Computers

[crunchygranola]

There are is a very active field of "voting theory" research about how voting systems can be improved, but little of it has to do with computers per se, though they can make implementation of the post-vote processing more convenient. That is to say, it is not the "computer" that is improving anything. Various forms of preferential voting have a lot to recommend them, along with variations like "instant run-off".

In general it is a good idea to identify actual problems (e.g. widely unpopular candidates winning due multiple candidates splitting the vote; spoilers being run to siphon of votes from specific candidates, etc.) and propose actual fixes that are subjected to formal analyses, large scale simulations and such to validate that they are improvements.

The suggestions of the OP mostly sound like "let's just try something different" rather than carefully considered improvement proposals.

Re:Simple

[doom]

Paper is pretty secure here, where most anyone and especially members of all parties, can watch the whole process.

Yes, that's pretty much it. There are scams you can use with paper ballots, but they're harder to get it to scale [1]. Arguably, a hybrid system (as is common these days) of paper ballots counted electronically could be better than a pure paper system-- then you can use computer techniques to look for problems, and paper hand counts to check afterwards.

Equally important is that the process is so simple that the average person can easily understand the process.

Yeah, exactly. You might have your spooky crypto-magic uncrackable system deployed perfectly, but it's too complicated for citizen's to understand, you could get demagogues whipping up distrust for the system. Counting the vote accurately doesn't help if no one believes you.

[1] One year, the coast guard found the lids of ballot boxes floating around in the San Francisco bay...

Re:Simple

[Chas]

I simply don't have enough mod points for this post.

Sorry, but computers simply are NOT a trustworthy medium for something like this.

I LIKE computers. And I trust them to work as they're told.

I simply don't trust the assholes who're doing the telling. Nor that someone couldn't subvert them and become the assholes who're doing the telling...

Re:Ranked voting

[swillden]

Computers would make this easier but are not required.

https://en.wikipedia.org/wiki/Ranked_voting http://www.fairvote.org/rcv

I used to be big fan of ranked voting, especially with Condorcet evaluation with Schwarz Sequential Dropping. Then I tried to explain it to a few people and changed my mind. Instant-runoff is a little simpler, but still pretty complicated -- and actually a bit tricky to execute correctly since it's inherently multi-pass (Condorcet is simpler to execute). Simplicity matters because what's just as important as having a fair election, is having a fair election that voters can understand and trust.

I think the best scheme overall is approval voting. The mathematical properties of approval voting are almost as good as the best ranked voting schemes. It's a little more vulnerable to strategic voting (which is when voters might have reason to vote other than their true preferences, as is the norm in plurality-rules schemes), but really not very much. In theory it also doesn't capture quite as much nuance of voter intent since it doesn't allow one to express a preference between two acceptable candidates. But it does allow voters to express another important element of intent which ranked ballots don't allow: acceptability. And it's brain-dead simple to understand.

If you don't know how it works, here you go: An approval voting ballot has all of the candidates listed. You mark all of those that are acceptable to you. The candidate with the most marks wins.

Such a system eliminates the strong two-party bias that plurality-rules systems have (Duverger's Law, that bias is called). In very few cases does it ever make sense to vote other than your true preferences. And it encourages parties to field broadly-acceptable candidates.

Tallying is a single-pass process and counts can be provided by sub-regions for totalling (unlike IRV, where the runoff phases require reinterpretation of the ballots at each runoff). If it's desired, you can even specify a minimum win threshold -- if no candidate gets, say, 50% approval then no one wins and you re-run the election with a new slate of candidates. There's an obvious risk of never getting a winner here, so such a system should probably progressively lower the required approval level to be sure that someone eventually wins, but the flip side is that such a system would mean that the 2016 US presidential election would never have put either Hillary Clinton or Donald Trump on the ballot; both (all) parties would be looking for someone with broader appeal.

However, approval voting can be done with or without computers, so it's not really relevant here. IRV can also be done without computers, though it's kind of tedious without them.