Finnish Court Accepts E-Voting Result With 2% Lost

Nailor writes "The Helsinki Administrative court accepted the municipal voting result in an election in which 2% of votes cast were not counted at all. We discussed this situation at the time. The court noted that the e-voting machinery has a feature, that should be considered as an issue. However, it also noted that 'a little over two percent failure rate can not be considered as such as a proof that the voting official would have acted erroneously.' Does this mean 98% of votes is enough to figure out how the other 2% voted? Electronic Frontier Finland has a press release about the court decision (Google translation; Finnish original)."

Comments

Failed to Finnish

[dov_0]

If it is a first past the post system this is a problem. If it is majority rules, then as long as there is enough of a majority.

2% failure rate is a bit much though?

Re:Failed to Finnish

[Jurily]

Finnish municipal elections are always by the D'Hondt method, so the result can be strongly affected by a few additional votes.

Doesn't really matter. If you let them vote, count all the fucking votes. It's that simple.

I have my own problems with any voting system skewing the results in favor of the two candidates most likely to win ("Don't vote on the little guy, your vote will be lost!"), but this is ridiculous.

Did they offer any reimbursement for the people whose vote they didn't count? I'd be pissed off if they did that to me. I'd also start screaming around about someone cheating, and likely sue as well.

Re:Failed to Finnish

[Anonymous Coward]

But it's not that simple. This e-voting debate has always been tainted by a complete lack of scientific logic, which your post typifies.

No measurement is perfectly accurate. Counting votes by hand is not a perfectly accurate method. We expect computers to be perfect but the measuring system isn't perfect so the results aren't either.

It doesn't matter. Our voting system has always had a margin of error, and always will. Thankfully one court in the world understands this and is bold enough to say 2% is an acceptable error.

If the election is not close, it clearly has no effect on the end result. It can only matter if the election is close.

And here's the thing - if the election is close enough, then no system will accurately measure the result. We have a situation, well understood in the world of science, where the noise is louder than the signal. The result of any binary discriminator in this situation is effectively random.

Closely-called elections have always been randomly decided, and they will always be randomly decided.

Let's just accept this and get on with the real debate which is, what is an acceptable margin of error? Do you agree with 2% or not? What would you like the error to be? How much are you willing to spend to reduce the error?

You'll note that in this properly cast debate, anyone saying that only 0.000000% is acceptable counts as an extremist who won't be listened to.

Re:Failed to Finnish

[Jurily]

If the election is not close, it clearly has no effect on the end result. It can only matter if the election is close.

It's the principle that matters. You're basically saying they should randomly not let people in the booth saying "You're the error margin, good bye."
How would you feel?

You'll note that in this properly cast debate, anyone saying that only 0.000000% is acceptable counts as an extremist who won't be listened to.

Make a law saying anyone who didn't get their vote counted doesn't have to pay taxes until the next election. I'm sure they'll get it sorted out in no time.

Re:Failed to Finnish

[Ornedan]

2% is bloody well not an acceptable error rate when manual counting error rate is 0.5% on average. As such, the court saying 2% is acceptable is utter bullshit.

Re:Failed to Finnish

[Anonymous Coward]

> Let's just accept this and get on with the real debate which is, what is an acceptable margin of error? Do you agree with 2% or not?

Given that 1% or even 0.5% is the limit where a party can ask for recompensation for their expenses (speaking about Germany), 2% is definitely far, far too much, and obviously anything above 0.1% is too much.
But that is not the real issue: the real issue is that nobody knows or can proof why which votes were lost, and the electronic voting systems make it completely impossible to find out even if 60% of votes were lost. You can then also not determinine if there was any system or regularity in which the votes were lost. This means that just speaking statistically, the margin for error must be far lower for electronic systems in order to allow the same accuracy as traditional voting.
If someone had had the possibility to let the green party loose 2% at will it is quite likely it would never have grown - a party that now regularly gets around 10% of votes.
If you say 2% loss is acceptable with a voting system that makes reliable tracking impossible (since different to pen and paper, electrons do not leave any traces we can track these days), that means you essentially o.k. any arbitrary vote manipulation as long as it does not exceed 2%.
That is not much if all you have is a party with 60% and one with 40% but if you have e.g. 1 with 6%, 1 with 5%, 4 with 3% maybe 3 more with about 1% that is basically giving the permission to mess up the system as you like (particularly if above/below 5% decides whether you get seats in parliament or not).

Re:Failed to Finnish

[Jurily]

Given that 1% or even 0.5% is the limit where a party can ask for recompensation for their expenses (speaking about Germany)

1% in Hungary.

But that is not the real issue: the real issue is that nobody knows or can proof why which votes were lost, and the electronic voting systems make it completely impossible to find out even if 60% of votes were lost.

No, it's not impossibe. We're talking about computers here, they can be audited.

Also, I'd like to see, how

votes[canditate]+=1;

has an error margin of 2%.

What's the margin of victory?

[AuMatar]

If one person had 70% and the other 30%, the 2% won't matter and they should accept the election while fixing the problem for the next. If it's 51/49 victory, then its an issue now.

Non-electronic spoilage rate

[profplump]

2% doesn't mean anything unless you know the spoilage rate for non-electronic voting. In the US 2000 elections, 1.94% of the ballots cast were spoiled, and most of those were not electronic. I don't know if Finland usually has similar spoilage rates, but if they do I don't see why this is any more or less a problem than the old method.

Re:Non-electronic spoilage rate

[getuid()]

I don't see why this is any more or less a problem than the old method.

I do. It's because of why the spoilage occured.

In the case of paper ballots, it's easy to imagine where the spoilage comes from: thousands of helpers handle millions of pieces of paper. Either they mis-count, they accidentally destroy bills... whatever. It's easy to imagine that there is some damage, and the most likely scenario is not a systematic error or a manipulation.

In the voting machine case, it's different. There's a machine counting results. It's not supposed to miscount. There's also the machines adding the numbers and doing the math -- it's not supposed to be wrong. If it is wrong, then it's propably not because of 1000 small errors adding uncorrelated small pieces to the spoilage (like it would be in the case of the manual counting), but it's most probably because of one single error or manipulation in the system. Being one single malfunction, it is not at all likely anymore that it has nothing to do with manipulation on purpose -- on the contrary, this scenario is very possible, and more or less likely.

An analogy with weapons, to clarify (not to justify) my points: if a bow+arrow wouldn't be able to hit a 1 foot target at 1000 yards distance, you wouldn't complain. That's the limits what a bow can do, after all... But if a high-tech sniper rifle missed the same target from the same distance, you'd have all rights to complain! The gun is supposed to work orders of magnitude better, there's no room for such a big error there.

Re:Non-electronic spoilage rate

[volpe]

So, your opponent got THREE TIMES as many votes (9) as you did (3), and you consider that a small margin?

2%

[clarkkent09]

I guess this is a question of whether it is possible to have a "perfect" user interface such that 100% of people who use it will get it right. Given the number of nincompoops out there, that is a pretty difficult problem. What is the percentage of mistakes with paper ballots? I bet there are 2% who manage to screw that up too.

I looked at the demo of the voting machine user interface and it seems perfectly sensible. You put your voting card in and press the number of the candidate you are voting for. A message comes up with large friendly letters telling you something like "This is the candidate you are voting for: <candidate details>. Press OK to cast your vote. [OK] [CANCEL] Apparently, at this point 2% of the voters simply pulled their card out of the machine and walked out of the booth without pressing OK. If they didn't have the confirmation screen, then the same people would press the wrong number and vote for the wrong candidate, and then complain that they weren't given a chance to correct it.

It all depends...

[fuzzyfuzzyfungus]

Except in extremely close races, a smallish percentage of lost/spoiled/uncounted votes isn't an issue as long as the lost votes are a representative subset of all the votes. If it is a selective subset, then you have a serious, serious problem.

Same thing as polling. If the people you do poll are a representative sample, you don't actually need all that many of them to get the correct answer. If you get an unrepresentative sample, then your results are worthless. It should be noted, of course, that with elections, unlike polling, you are still obligated to put forth your best effort to count everybody's vote(though, depending on the technology, imperfection is inevitable to some degree).