Let’s face it.
It is impossible to engineer an electoral process that will have accuracy to .01%
No matter who takes the White House we are well outside our margin for error.
These irregularities take place everywhere every election.
We need a joint statement of unity endorsed by both candidates about how far we take this thing, and how we can resolve it quickly for the good of the country (and the stock market.)
Agreed.
The process works the way it’s supposed to work. I’m sick of hearing reports of “irregularities” that happen in each and every election in each and every district.
The Bush campaign can demand recounts in Wisconsin, Iowa, New Mexico, and Oregon and win the damn election without Florida. Do we really want to prolong the process to include recounts in every state?
I disagree. It is well within our technical abilities to record an election result with almost arbitrary accuracy. There is nothing in the nature of the problem which demands inaccuracy. The choices are discrete and easily identifiable. The only limiting factors are time and money.
Time is far less of a constraint than it used to be. We have a built in cushion resulting from the fact that the timing of the election and the inauguration were set long before the era of instantaneous communication.
Money will always be a limiting factor, but we are nowhere near the ceiling of what might reasonably be spent on a matter of serious import to the republic.
There are numerous examples of large, complex information gathering systems in this world that fundtion with a higher degree of accuracy than the .01% threshold you mention. I do not think it unreasonble for the citizens of this country to expect that their direct voice in the governing of this nation be afforded at least that much respect.
I’m a little pissed about the whole thing.
I want a major overhaul of the system after this is over.
I want statewide standerdized ballots. I want ID required to vote. I want HUGE jail sentences for those convicted of vote fraud. I want dead people removed off all registration rolls. I want people to have to re-register if they haven’t voted for since the last presidential election.
Any other ideas?
If we do not have a honest election, then why vote?
No. The process is supposed to record the actual count of valid ballots. To date, in Florida, it has not done that.
If there are questions about the reliability of the vote which could actually affect the final result, then yes. Why place expediency over accuracy when we have time to make the accurate decision?
As to whether the Bush campaign would gain votes from recounts in Oregon, New Mexico, Iowa or Wisconsin, that is pure speculation. A recount, I believe, is underway in New Mexico. If it turns up similar discrepencies to those in some Florida counties, a hand recount might be warranted there, too. If not, then the recount will simply confirm the present result. How is that a bad thing?
We have 70 days until the next President will be inaugurated. Why do some people feel the need to “settle” the election before the votes have been reliably tallied?
Because they don’t want their guy to lose.
By the way, Spiritus, I wanted to tell you how I admire your calm reason and excellent communication skills.
Bravo.
stoid
Spiritus:
I don’t understand how you propose to get an accuracy of .01% or better in a system that has 100 million+ potential voters, with each of them subject to human error.
But I’m willing to listen 
Good points, Spiritus and Freedom. I was thinking of the system as it stands when I wrote my reply.
While it’s theoretically possible to institute a standardized system with a minimal margin of error, that system simply does not exist at the present time. I’ll admit to a bit of naivete in that I had no idea the polling is so . . . primitive in so many places.
The problem as I see it with endless recounts is that it opens a can of worms we may not want to face. There are always discounted ballots and voter errors. It seems there is rampant confusion as well, conflated or not by post-election sniping. While the ultimate goal of the most correct count possible is obviously paramount, we may be looking at a situation in which with the system currently in place, we have what we can only believe to be the most correct count.
One outcome of the lawsuit(s) will probably be some form of standardized voting methodology. I’d argue it should be national in scope, with built-in disallowance of invalid ballots while the voter can still correct their choice, and near-instantaneous totals. Hell, everything else is computerized - why ain’t this?
Have you noticed when other countries have elections they are ‘tainted’, how soon we see in our papers stories of us sending in people to count and recount or to observe for election ‘abnormalities’. Obviously we want to make sure the elected official is the one our country wanted in the first place. Would we let election officials from Haiti, Panama or Bosnia come in and monitor our election? A simple touch screen would be easy to implement and not cost as much as a rebuilt Cole warship. Would older and under-educated be scared off by that? Probably no more than by giving them a piece of paper and a sharp tool.
I posted on this very subject on another thread. To me (not an American, BTW), the way to go would be ATM-style voting machines, with a touch screen displaying large icons for each contestant, and a ‘IS SO-AND-SO YOUR CHOICE?’ prompt before producing a physical record of some sort, to be deposited in a ballot box. (I just do not like the idea of an all-electronic vote, with no means to physically count the votes.)
These should be provided to every polling station for elections. It would cost a pretty penny, but compared to the money being spent in the elections by various parties, it is probably a pittance.
Bill
Stoid I dissagree, I think people just don’t care enough to want to hear about this any more. This patrioism is very short lived.
I just want to state that I am 100% against electronic voting.
At least with paper ballots and machine voting, the average person can serve as a challenger and understand if he is observing fraud. With electronic voting, we are completely at the mercy of whoever programs the machines.
The same thing goes with internet voting. I see the election as a balance to the government. When we relinquish oversight of this process by the average citizen, then we have abdicated our responsibility.
From a systems standpoint, it would be extremely difficult to engineer an election process that had an error rate less than 3 or 4 sigmas. Even in highly controlled environments like production lines it’s a major feat to get your error rates below this number.
If you have 100 million voters, an error rate in the hundreds or thousands is pure noise, and probably inevitable.
If Yaweh himself knew exactly how everyone wanted to vote and compared it with the actual tallies in each state, I’d be surprised if any state was accurate within a thousand votes.
Well, to begin with you misunderstand the problem.
We have no difficulties in adding numbers with arbitrary precision, even numbers which total 100 million. The places where accuracy needs to be ensured are in the marking of the ballot and the evaluation of the mark.
Now, an electronic mark can be evaluated with incredible precision. It is done trillions of times every day. Having the ballot “marked” in an electronic medium removes, for all practical purposes, error in evaluation.
That leaves only errors in marking. It is a simple matter to program an input screen to accept one and only one entry (none of the above would have to be a valid choice). This eliminates errors of double entry or no entry. Because the entry is electronic, you eliminate the “noise” of improper hole alignment, incomplete punches, cardboard chaff, etc.
For record keeping, a hard copy of each ballot could be printed at the voters acceptance of his marks and retained by election officials.
Because the system is electronic, it can also easily be tested for accuracy in both evaluation and computation of data.
There are very few theoretical limits to the accuracy (determined as a measure of recording a ballot exactly as marked) in this system.
blush
Gawrsh. Thanks. You must be catching me on one of my “good” days. 
[sub]hmm, now where was that anti-infinity link again . . .[/sub
Except for human error, in:
The design of the software
The human factors of the interface
Data Entry errors
Machine failures
Hell, when you’re talking about error rates of 1 in 100,000,000, you might as well toss in stray cosmic rays and power failures.
The biggest ‘slop’ areas in this system are the human factor engineering and data entry errors.
By Human factors, I mean introducing bias through the look-and-feel of the interface. A simple example: Many seniors are unreasonably afraid of electronic devices like this. I know several who have never used an ATM and never will because they don’t understand ‘those machines’. So an electronic system will bias against the candidate that gets the most senior support.
There could be many other interface errors:
[ul]
[li]A font that’s too small, biasing against people with poor vision '[/li][li]Font colors that make it hard for colorblind people to read the ballot[/li][li]Biases by position - the candidates on the outside of the button row will suffer fewer data entry errors than those that have a button on each side, because the buttons in the middle have less physical space.[/li][li]Mechanical Problems (dirty switches, broken contacts, faulty displays, etc)[/li][/ul]
And these are just the ones off the top of my head. In a complicated system like a state-wide computerized election I guarantee that there would all sorts of other potential sources for error. Race hazards in networks, packet collisions, security hacks, data overruns in bad software, you name it.
Sam, perhaps you are defining your (hypothetical) error rate based upon the intent of teh voter. In that case, there will always be an element of human error and carelessness.
However, I have been addressing (based upon exchanges in other threads), errors in accurately and reliably recording validly marked ballots. It will not stop someone from marking “Buchanan” when they meant “Gore”. But it will correct both large numbers of discarded ballots due to over/under punching and the demonstrated unreliably of the present machine count process in several Florida counties.
Software design for data collection is extremely simple and easy to test when dealing with discrete data sets.
Interface elements can bias selection, as you have noted, but have no effect upon the accuracy of measuring and tabulating the selection made. I have proposed that citizens have a right to expect their ballots to be recorded accurately, not that they have a right to have their inadvertent selections “corrected”. That said, every effort should be made to reduce bias from the interface.
Data entry error – see above.
Machine failures. Always a possibility, but very few machine failures make data unrecoverable. the hard copy generated, even in teh most extreme case, would allow any affected ballots to be recovered.
We were talking about error rates of .01%
All of these are easily corrected for and/or minimized. A private network or shudder physical data transfer removes many from consideration. It requires little more than a second-level Comp-sci course and sufficient incentive to write code secure from data overrun errors. As for hacking: Any system might be hacked. Hacked undetectably is a different matter. There is no reason to allow open network access to the machines in teh first place. Also, as I noted before, test data for discrete data sets is remarkably easay to construct and maintain. Machines can easily be certified for accuracy both before and after any election.
You underestimate the problem of complexity. Of course any one of those issues can be resolved, IF YOU KNOW THEY EXIST. Unfortunately, in very complex systems it’s hard to know where all the sources of errors are, and then there are unforseen interactions, etc. If there are hundreds of ballot machines, it doesn’t take too many glitches, intermittent switches, etc. before the error rate creeps up.
The company I work for has spent 2 billion dollars this year in an attempt to raise the accuracy of its systems to six sigmas. They’ve only made a dent in that, and only in the most well-defined procedures (i.e. assembly lines and industrial automation). I just finished my ‘six sigma’ training, and got a good appreciation for how difficult that task is. I don’t believe it’s ever attainable in a system that relies on manual interaction with millions of people.
Another issue here is one of cost. How much money are you willing to spend to engineer such a solution, when current accuracy rates seem to be good enough to decide the election in every case but a strange fluke like this one? WOuldn’t it be a lot cheaper to simply amend the constitution to say that any state’s (or combination of states) election that is closer than, say, 1% in difference AND those electoral votes would change the presidential election, then there must be a new national election after a 30-day wait? It’d probably never come up again, but at least there would be a constitutional procedure in place.
Again, Sam, we seem to be speaking about different problems. Your assempbly line analogy and your repeated reference to the element of human interaction both point to a focus on teh problem of accurately registering the vote which a person intended to make upon entering the system. I agree that there is an irreducible element of error in this which cannot be elimintated.
However, I am talking about minimizing error in the accounting of ballots actually marked. this is a far simpler problem. In particualr, by elminating a mechanical interface in the production of a marked ballot, we eliminate a major source of noise. Also, we guarantee a discrete and easily differentiated data set.
Bit switches on modern computer systems are quite rare, and parity checking algorithms reduce the danger even further. Calibration and testing of results is a well understood problem for discrete data sets: boundary conditions are easily identifiable and accuracy is assured through predefined test data. Network communications algorithms have given us a very powerful toolset for detecting and correcting sequencing errors, bit switches, etc. during data transmission. Some simple care in problem design will additionally ensure that no single bit switch, even if undetected, can change a vote from one valid recipient to another.
Of course, in any systemm it is necessary to understand where errors can originate in order to reduce them. This is, however, a very restricted problem set. The entry parameters are small, the data manipulations are minimal (trivial, even). The primary element of difficulty is one of scale. And the scale in this case brings with it no additional physical constraints/problems. As such, any modular solution will be able to scale upward with minimal risk of introducing error. (errors existing in the module will, of course, be magnified accordingly).
Remember, there is no requirement that the multiple independent machines have any communications or interactions with each other. The only increase in complexity as the problem scales upward occurs in the collection of results into a central repository. This is basically a problem of verification and addition. Binary addition is tight. Verification of data transfer is well understood and has been solved to high degrees of certainty in a number of contexts.
Price? There is always a reasonable limit, but this is a problem affecting the core principal of our republic. I think it deserves a better answer than we are presently using.