Comments on this post from Jersey Joe of Shelby County offer more evidence why unverifiable electronic voting machines suck and what problems we could face if we use them in 2010:
You are correct. The 2005 voluntary standards of the EAC are not established as the gold standard guidelines for anything. In fact, in November of 2008, the EAC published an evaluation of a system and they used the 2002 standards for that evaluation. There is no reason that Tennessee cannot use the 2002 guidelines.
It is also true that there are the next generation set of guidelines being worked on now. In the comments on the new guidelines, some people question the value of the EAC guidelines, saying the testing does not correlate to real world experience. There is also skepticism expressed by one state implying fraud in the utilization of Diebold machines, the kind we use here in Shelby County. (http://www.eac.gov/program-areas/voting-systems/docs/certification-system-comments-by-section.xls/attachment_download/file)
There is little doubt that Secretary of State Hargett is willfully obstructing the will of the General Assembly. Is there anyone in the Republican majority who cares more for their constitutional oath than their Party and who will demand that Hargett follow the law as passed by the legislature?
The question is why does Hargett want to so brazenly disobey the law? Why has the question of having elections with verifiable ballots that can be recounted become a partisan issue? Accurate recounts should be in everyone’s interest. But not here in Tennessee in 2009. Why? Are the rumors about these systems true? Did the Republicans really win all those surprise House districts last year honestly? All the Voter Confidence Act does in the end is give us the ability to habd count ballots independent of technology. Much of the current technology is totally opaque and so there is no way to determine if the machine accurately captured the voter’s intent. Since all machines have a known error rate, in a very close election, the machine may be incapable of differentiating the true vote counts. This will be a problem with the optical scan machines as well.
When these machines give us a number, they are really giving us a range of numbers based on the known errors of the system. Based on the known error rate, we can construct “confidence intervals†around the result which can be for any degree of precision we choose. In science, the standard is usually 95%. What that means, is if the machine has a CI of plus or minus 5%, then 95% of the time the “real†number will be within that range around the number given in the count. Let’s say 200 balots are counted and the 95% confidence interval is 3. If the machine count comes out 100, then we would add and subtract 3 from 100. We can then say that if we put the ballots through the machine again, we would get an answer between 97 and 103 at least 95% of the time. But lets say that two candidates are running and candidate A gets 99 votes and candidate B gets 101 votes. Candidate B has won—maybe. If we run the count again, candidate A will get a number between 96 and 102; candidate B will get a number between 98 and 104. There is roughly a 43% chance that candidate B will lose or be tied in that second count. Who wins? The machine count, in this case, is not capable of answering that question. It is beyond the physical capability of the equipment.
In that situation, we must have an independent method to ascertain the intent of the voter to do a valid recount. Merely running ballots through the same machine(s) is meaningless. Let us imagine that Congressman Cohen beats Mayor Herenton by 50 votes in next August’s Democratic primary election. Is Secretary of State Hargett eager to stand before this community in the calm, quiet of election night and explain to us all why the Election Commission can’t really recount the votes and therefore, Congressman Cohen has won.
Secrertary Hargett is leading us down just that road. What is the gain that makes it worth taking that kind of risk? That is the question we should be asking of Mr Hargett and Senator Norris and Senator Todd and Senator Stanley and anyone else who advocates no recount voting systems.
If you’re a candidate in the next statewide election and we’re still using the unverifiable machines, you better make sure your margin of victory is a large one.
[...] 2008 Voter Confidence Act that would give Tennesseans paper ballots as well as the ability to do a meaningful recount because he thinks it’s would be a waste of [...]
I don’t know what the stated error rate is on Intel chips, but my computer sure does strange things from time to time. I don’t know what Diebold claims are for its systems. In 2005, the California Secretary of State refused to certify the AccuVote TSX (used in Shelby County) because it had a 10% error rate. This included things such as screens freezing and printer jams (Shelby does not have printers). This report seems to use error rate and failure rate interchangeably. I was referring to the error rate of the machine incorrectly recording or not recording the vote properly according to the intent of the voter.
This may be a little easier to understand in optical scan balloting. The scanner must read the mark made by the voter in relation to a defined field on the ballot. Voters can make stray marks which can be misread by the machines. Many people may be familiar with the strange results you can get when scanning documents into a computer. Symbols, letters, can be misread. Often, the rate of this misreading is a known factor. Remember the infamous punch card ballots of Florida 2000? I have been told by Shelby County election officials that the percent of ballots that will produce hanging chads can be predicted based on the quality of the punch cards used. Most machines, even (or especially) computers, are not perfect. They have a predictable rate of malfunction. If this is known, then based on the situation and sample size, it is possible to compute a range of probable values for any result. If these ranges don’t overlap, then the results are statistically distinct from each other. But if the ranges do overlap, then the machine cannot differentiate between the 2 cases tested. I other words, if asked to say whether A is bigger than B or B bigger than A, the machine cannot consistently give the same answer. In an election, that is bad. The only way around it is to use a completely different tabulation method- usually manually counting the ballots.
Despite Paul’s skepticism, machine errors happen frequently. A whole host of problems have been documented for the AccuVote TSX (see http://www.votersunite.org/info/dieboldinthenews.pdf). I agree with him in the sense of irony that we would be retreating to manually recounted ballots. The image of the sheriff taking the locked ballot box into the back room to count it by himself is a part of Southern lore. Sadly, computerized balloting has not fulfilled the initial expectations many had. I haven’t even touched on the serious security concerns and other software problems that many have raised. I share his disquietude at this seeming retreat to the past, but it appears to be the best solution at this time. The National Science Foundation’s ACCURATE project (http://accurate-voting.org) and, I assume, commercial interests are working toward better tech solutions. But we still must run fair and honest elections until then.
There is a general consensus around optical scan machines with recountable paper ballots. The machine will scan and count. A manual recount will only be needed if the race is close. This does not mean that supporters of the Voter Confidence Act believe manual counting of ballots is the best thing since chocolate chip cookies. But having that ability to do a recount and be able to check the voter’s intent is vital to confirm the legitimacy of the results.
Are you people insane? Machine Error Rate/ What is the Error rate of the computers you and i use everyday? what is the error rate of the ATMs(made by Diebold) that are used around the world every day? What is the error rate of a simple calculator? I am not a Technology wiz but my understanding is that computers are created on a Bi-nary system (1 or 0)(yes or no) With touchscreen computers a person either made a vote or they did not, no room for human “interpretation”
Paper Ballots leave room for problems such as a nefarious attempt by someone to steal the election or pure incompetence from those who take part in the election processes. The latter is more common than the first no matter what people on both sides say. In both cases however paper ballots invite problems before the voting starts, during voting and after the polls close. As in the 40s with Boss Crump you could have people handing out already completed ballots before people get to the polls, EO’s could just make mistakes and send ballots to the wrong polls, After the election on mistake or on purpose ballots can be lost.
Stick with machines take the human element out, and do not subject us to endless recounts like Florida 2000 or MN Senate 2008