Goals of Voting Systems

This week I was approached by two different groups seeking to build better voting systems, something I talk about here in my new democracy topic. The discussions quickly got into all the various goals we have for voting systems, and I did some more thinking I want to express here, but I want to start by talking about the goals. Then shortly I will talk about the one goal both systems wanted to abandon, namely the inability to prove how you voted.

Many of the goals we talk about are actually sub-goals of the core high-level goals I will outline here. The challenge comes because no system yet proposed doesn't have to trade off one goal for another. This forces us to examine these goals and see which ones we care about more.

The main goals, as I break them out are: Accuracy, Independence, Enfranchisement, Confidence and Cost. I seek input on refining these goals, though I realize there will be some overlap.

Accuracy and Precision

We want our voting system to accurately report the will of the voters. Unlike most other systems to measure will, we also want perfect precision -- we believe that every vote must count, and as most systems are defined we believe it appropriate that a single vote can flip an entire election result if the votes are even.

Security

The most talked about sub-goal of accuracy is security. While often thought of as its own goal, security is simply a means to accuracy. We don't want it to be possible for bad actors to deliberately add, remove or change votes and in particular the vote total. We don't want outsiders to be able to do it, and we don't want trusted people such as election officials or providers of voting equipment to be able to do it.

Authentication of voters

Those not eligible to vote should not be able to vote. Each voter should vote exactly once.

Reliability

Systems should record votes as cast. Systems should not break down, impeding or delaying voting. Systems must be up to the job of handling the full complexity of a ballot.


Voter independence

We wish to record a voter's true will. While people are free to influence voters with persuasion and argument, we do not wish voters to let external pressure change their vote, including pressure from family, friends, employers, churches, thugs, vote-buyers or even political parties. Every vote should be cast because the voter thinks it is the right choice.

Confidential ballot

One-half of what we often call "secret ballot," we feel that a ballot that nobody else can observe helps promote voter independence. If people can see how you vote, you may vote to please them, rather than select the candidate you feel is best.

Unprovable ballot

The other half of "secret ballot" is the inability to demonstrate to others how you voted. You might try to demonstrate that because of coercion or bribes, or due to peer pressure, or simply out of pride. Being able to prove how you voted is very useful for vote buying. While revelation is usually optional, preserving a confidential ballot, if pressure is involved then a provable ballot is not a confidential ballot.

This is the sub-goal that some are considering abandoning.

Unbiased

The voting system itself must not bias the results towards any choice.


Enfranchisement

We wish all eligible voters to be able to vote without undue difficulty. For no voter should it be impossible to vote, and any high difficulty in voting should be mitigated as much as reasonably possible.

All eligible voters must be able to be authenticated (registered) to vote in the appropriate elections.

Turnout

Somewhat more controversially, we seek high participation in the system. I say controversial because some would argue that it is satisfactory if everybody who truly cares above a certain level votes, as long as that level is roughly equal among the population, while those who don't care enough to overcome certain inconveniences often don't vote.

Ease of voting

The easier it is to vote, the more who will vote. In some cases this may even affect ability to vote, so I have not classed it as simply a sub-goal of turnout.

Vote at Home/Vote by Phone

Allowing vote at home supports ease of voting, and in some cases ability to vote.

Voter assistance

Particularly if ballots become complex, systems can make it easier for voters to understand ballots and the process. (They can also help voters understand the security, aiding the Confidence goal below.) Systems can also detect errors in voting before a ballot is finalized and offer, in an unbiased way, chances to correct those errors. Preferential ballots (where candidates are ranked in order, such as in Australia) are notoriously difficult for some voters to understand and get right, and function best with voter assistance.

One simple form of voter assistance is the ability for the voter to read the ballot and any instructions in a language in which the voter is fluent, or in a non-printed form if the voter is not literate or is visually impaired.

Inherent in this goal is systems that don't go the other way and confuse the voters, or make voters miss things.


Confidence

The result must not only be accurate, the public must be confident in its accuracy and precision. Strictly, a system that is accurate is still valid if there is no public confidence, but in practice this does not work.

Auditability

It is generally considered essential to confidence that there be some means to audit the results of whatever primary system tallies the votes, both for close elections, and in random spot-checks to prevent fraud. If practical, there is no reason every vote should not be counted in multiple, independent ways to assure security and accuracy.

This problem is difficultly recursive, however. There must be confidence in the methods of accountability as well.

Simplicity

The voters will not be confident in the results if they do not understand how the system works, and how the security and auditability work. "Trust the experts" may not be a great deal better than "trust the officials."


Cost

There are constraints on how much money, as well as voter time, can be expended on a system.

Speed

Possibly an independent goal, but it is generally desired to have results within a day or two of the completion of voting, and there are in some cases rules demanding final results within some longer period, such as 2 months, even in very close and contested elections.


Goals Not Addressed

There are other goals which exist outside the voting system itself. For example, we seek district boundaries to be drawn fairly. (No gerrymandering.) Some seek to avoid unfair advantages for incumbents, or large biasing of the system via money. Some seek to avoid voter-chosen undervote (voting in one election but not in another.)

Strategies that seek to meet multiple goals

Voting Machines

Electronic Voting machines are the hot topic in voting. People first came to them hoping for perfect precision and accuracy, which were not found in most other systems. However, their big supporters today are those looking for ease of voting and voter assistance. Machines can allow the disabled to vote on their own, explain complex ballots and point out errors.

However at the same time most implementations of machines failed at other goals, providing no auditability, costing lots of money and in some cases confusing voters rather than helping them. Failure of machines also has led to great difficulty of voting, long lines and even disenfranchisement. For some they inspire confidence, but many view voting machines as a dangerous exercise in "trust the experts."

Vote by Mail

The entire state of Oregon has gone to vote by mail. This is very easy, and some argue it enhances turnout greatly. However, it sacrifices the unprovable ballot completely and has several security issues. More to come on this.

Basic Paper

Where elections are simple, very simple systems are possible. In Canada voters simply take a slip of paper behind a screen, mark an X, tear off a serial number and place their ballot in a box, the serial number slip going in another. They are quickly counted and tallied. It is very easy to see how the system works, and how auditing works in recounts. Scrutineers from the various parties oversee the process to avoid corruption by election officials.

In some nations, voter tagging is done with a very low tech purple ink system. Those who vote get their thumb dipped in colour, they can't vote again as it does not wear off for days. In some nations ballots are placed in transparent boxes (folded so votes can't be seen) and these boxes remain under competing watchful eyes at all time.

Comments

In considering the merits of various existing and proposed election systems, I've found the following two concepts highly useful:

1. Community of Trust: For a specified election system, who are the people I have to trust? Where machines are concerned, who are the people behind the machines that must be trusted? Are there any ways of reliably verifying these people do what they're trusted to do and don't do what they're trusted not to do? What, exactly, am I trusting these people to do or not do? We should seek to eliminate unverifiable trust in individuals who could potentially be the corruptible "weak link" in the chain to ballot processing. In some cases, our trust is divided among a larger group such that all members of the group would have to collude in order to manipulate election results. In these cases, we should seek to maximize the size of the community because the more people who would have to collude to manipulate results, the less likely manipulation is to occur. Similarly, where trust is given to a group, it is wise to have representatives from multiple competing interests within the group to act as checks and balances on each others' participation.

2. WWKD: "What Would K. Do?" (Where K is a corrupt high-level elections official or Secretary of State). This is similar to game theory. To expose any flaws in an election system, I pretend that I'm a powerful and corrupt election official with access to equipment and records and hiring/firing authority over my subordinates and, further, that I want to manipulate an election to favor a specific candidate in a race conducted using a specific election system without getting caught, or at least without leaving any evidence of said manipulation. I advocate that proponents of new elections systems conduct "war games" such as this to expose possible methods of election fraud. In some scenarios, the actors in the game could be vendors, first line elections workers or IT professionals within the elections agency.

These are sub-goals (though I am enumerating some of them) which is important to understand. No sub-goal is a "must" because there might be another way to implement the main goal. The main goal is that voters are confident that the system has met the other goals, particularly accuracy and precision. There are a number of possible ways to attain that trust.

Perversely, the trust goal may actually be faked so long as the other goals are attained. The trust goal is one of feeling. While some people must truly understand the correct reason why the system can be trusted, it is quite possible that many voters might trust the system for incorrect reasons. As long as everybody is not incorrect, and this universal error does not permit accuracy to fail, the system is still returning the right result.

The confidence goal means that people will treat the government as legitimate, and thus will defend it and support it and in the extreme die for it. This seems Machiavelian of course, and even comes with a risk that if people discover their trust was incorrect, they could act differently. (Though it is not so bad if you learn the system is still valid, though you didn't quite correctly understand why.)

Of course, a system everybody can understand sufficiently to have confidence in it remains the top expression of this goal. It is one of the real burdens to DRE voting machines.

You might be interested in the "defense tree" I've posted at http://zesty.ca/voting/reqs.html. The higher levels of the tree could apply to voting systems in general, whereas the lower levels are more specific to my implementation.

I would however ask about the turnout sub-goal I list. This one requires some thought, and is the driving force behind registration drives, register on election day, motor-voter, australian must-vote laws and mail-from-home ballot.

The tough question is, how important is voter turnout? A good system allows everybody to vote if they make the effort. However, you can control the "effort" level and affect an election quite a bit. In some cases the effort bar is raised so high that people call it true disenfranchisement. In other cases, people question the value of setting the effort bar so low that people who barely care still vote, as this affects an election quite a bit too.

How do we make that decision? Which is better, an election with 45% turnout and secret ballot, or one with 70% turnout and vote-by-mail? Or the 96% turnout (up from 47%) in Australia? This is not a question with a simple answer, but every election system ends up making a choice along this spectrum.

Add new comment