20 April 2006

Vote Centers In Denver In August

Denver's election system will transition from the precinct system to the vote center system, where any voter in the City and County of Denver may vote at any of 47 vote centers rather than only at their local precinct, for the August primary in 2006. It hadn't been clear, until today, whether or not this would be implimented in the primary, or only in the general election. The move is a good one, as it allows the kinks to be worked out of the system when the stakes are lower and there are fewer voters involved.

One by one, we are removing problems from the electoral system. Electronic voting eliminated the "overvoting" problem (when someone accidentally votes for more than one candidate in the same race), the problem of ambigious markings on a ballot, and mathematical mistakes in tallying up the results.

Vote centers eliminate the wrong precinct, right county problem, which is one of the most common impediments to voting in the precinct system; almost every precinct has someone who makes that mistake. Ultimately, perhaps we can move from a system of countywide vote centers, to statewide vote centers, so that a voter from Denver can cast his or her vote in Aspen, for example.

We aren't at the end of the road yet. It is still too easy to undervote in the current system (i.e. accidentally not cast a vote in a race), which is a problem that could be easily be 97% cured if voting machines would flash a "do you really mean it" warning if you hit the "cast vote" button without voting on every issue.

Another big impediment to voting is the voter registration process, which should permit people who register to vote on the day of the election to vote, but does not. I'd also like to see the bar on voting for people on parole who are not in custody lifted. A good statewide voter registration database could facilitate this change.

This most intense debate right now is a fear that electronic voting machines can be a device for untraceable fraud or a wrong result due to an undiscoverable software mistake. This fear isn't baseless. But, I don't share the intense democracy is at stake fervor that many people who care about this issue do.

Tampering with an electronic voting machine is a sophisticated fraud, particularly when, as in the Denver system, the machines are not connected to a network during the course of the election. At the end of the day, each machine prints out its results and those paper printouts of machine by machine summarys are tabulated and reported to the election commission.

The current system provides quite a good check against fraud that impacts some, but not all machines at a single polling place (and in a vote center system, there will always be multiple machines at every polling place). Since voters are assigned at random to each machine, and the number of votes cast in each machine and at each polling place (especially with vote centers) is quite large, one can discern using a college freshman statistics class level Chi-square test how likely it is that a disparity in results between one machine and another at a single polling place would arise from random chance. When results are particularly disparate, election officials (including partisan election judges) are given the notice that they need to promptly investigate more carefully for any possible source of error, either in the source code, or otherwise (e.g. a sticky button or a burnt out vote indicator light).

The main checks against a systemic fraud impacting all of the machines at a particular polling place, which is something those on the ground would be powerless to detect in the abstract, are the security under which the voting machines are kept between being programmed and the election (for example, most can not be turned on without a key), and the integrity of the people who have been charged with running the election.

Most people who are concerned about the issue are not primarily concerned about either of these things (after all, if these issues are not in order, a paper ballot system can be corrupted as well, they are instead primarily concerned with the integrity of the people who program the voting machines in the first place, because they are private companies that often have political ties and interests themselves, and the better intentioned people who run the elections rarely have the technical competence to detect this irregularities directly.

Also, there is very little incentive for the system to have systemic fraud in the part of the source code for the machines not particular to a given election. Until the individual candidate details are entered into the system via some sort of data entry interface, the distant corporate officials who wrote the basic software would need to take some very blatant measures to bias the code to favor the candidates that they want to win (such as searching for party affiliation code words and having code that changes votes slightly in favor of certain party affiliations), because their involvement is typically not necessary to take the pre-election step of specifying who will be on the ballot in what order. The easiest way to solve this is to use open source software. But, we haven't gotten to that point yet. In the meantime, we have to keep in mind that this kind of blatant vote theft in the basic software is easily to discern by someone with access to it, could be leaked by anyone with that information with anonymity, and once discovered makes it easy to prove criminal wrongdoing.

Also, the difficulty of directly directing fraud in non-public source code doesn't mean that systemic fraud can't be strongly indicated by results. If results in one election are in line with historical trends for a neighborhood, the likelihood of systemic fraud is very low. If results are grossly out of line with a neighborhood, the likelihood of systemic fraud is quite high. A real vendor level fraud has to be big enough to make a difference, yet small enough to remain plausible. And, in any such fraud, the stakes in terms of lost business and criminal penalties, if it is discovered (and many people with only incidental suspicions may feel a civic duty to report this kind of fraud in a way they might not if mere grey areas concerning money were involved) are high.

Yes, it is possible to use electronic voting machines to cheat. But, a wholistic view of the process as a whole and the means for detecting and correcting deviance in that process, when compared to the often relatively good faith problems with other systems that can have systemic bias effects, gives me more comfort than many who make this a top priority issue.

Also, the reality is that while paper ballots may work fine in Britain or Canada, where only a single candidate race for a single public office is one the ballot at any one time, in the important national MP elections, they are far less workable in Colorado (and most jurisdictions), which customarily has dozens of matters upon which a voter must make a decision at the same time. The experience from those small counties in Colorado that do use paper ballots is that it is a very slow and cumbersome process to count them and that like any human process, hand counting itself has real accuracy issues.

No comments: