What is Voting For?

Photo by Arnaud Jaegers on Unsplash

Voting is used on blockchain-enabled platforms in a variety of contexts, from protocol governance to token curated registries. How can we assess whether a voting system is well-designed?

There are two conditions under which implementing a voting system is useful:

  1. When individual members of a group have different preferences over an outcome relevant to them — let’s call this voting for preference.
  2. When individuals share preferences, but there is some uncertainty over the choice that will lead to the correct outcome — let’s call this voting on information.

Often times, situations which incorporate voting are a combination of both.

For example, when it comes to democratic government we can both have disagreements over what outcomes are desirable (e.g. maximizing GDP vs. minimizing inequality) and we can have disagreements about which policies are likely to lead to the outcomes we want.

The design of the rules that govern a voting system can and should depend on three factors:

  1. The context in which the voting will take place.
  2. The reasons for implementing voting — preference vs. information.
  3. The overarching goals for what the voting system will accomplish.

Imagine a group of friends out on a Saturday afternoon. While they are out, they spontaneously decide go to dinner together. In order to decide where to go to dinner, they hold a vote. The option that receives the most votes is where they will eat.

They are selecting an option in a way that takes into account the preferences of each of the group members; voting on preference. While most of them may think their friend Bob has very weird taste in food, they still count his vote equal to the others’, because they believe that in this context no one’s preferences should matter more than the preferences of another. Additionally, they do not call up friends who are not joining to ask for their votes, because their choice does not impact those friends, and therefore, it is not important to take those friends’ preferences into account for this decision.

The group chooses a majority rule vote because it is fast, simple, and seems fair to everyone. They may use different procedures to determine which options will be voted upon and they may even choose to have multiple rounds of voting if enough options are suggested.

While each design choice will have an impact on where the group ultimately chooses to go for dinner, that final decision isn’t necessarily important.

The end goal is to use a system that they all agree is fair, thus combining their distinct preferences into a single choice and thereby encouraging cohesion and coordination on outcomes that serve the group.

If this group of friends all shared the same preferences, as long as the set of options was well known, then introducing voting would not improve the final decision. The outcome of the vote and the outcome of one group member’s choice, in that case, would be the same, allowing for this decision to be made by any individual member of that group without a vote. However, if uncertainty exists over the quality of the restaurant options, choices of different individuals may be different. Even if the group had identical preferences, a limit or difference in available or obtained information would result in different beliefs about which option will be the best.

This is where voting on information comes in.

Juries are a clear example of a voting system in which the only relevant factor is information.

We all want guilty people to be punished and innocent people to go free; that is the goal of a criminal court system.

The reason a jury is used to make a determination is due to uncertainty regarding truth. Juries are an attempt to utilize the wisdom of the group.

The trial is the means by which the jury is given as much information as possible in order to ascertain the factual matter of guilt. As we do not believe that the crowd can enlighten us if they do not have the required information first, we do not ask a jury to vote on guilt before having a trial and deliberations.

In the US, rather than a majority verdict or even a unanimous verdict to acquit, we require a unanimous verdict in order to convict the accused. This is because we would rather make the mistake of letting guilty people go free than convicting someone innocent. So jury voting rules are designed taking into account the tradeoffs of different design choices.

Jury membership is limited and selected on specific dimensions:

  • A juror must be at least 18 years old.
  • A juror must be a US citizen.
  • A juror can’t work in law enforcement.
  • The prosecution and the defense both have a say in which individuals will be chosen for their jury.

These rules are chosen based on assumptions and ideas about who can be informed and unbiased in an attempt to keep preferences out of the decision. This is not to say that this system is perfectly designed or that it achieves its high objectives. However, this system does get closer to achieving those goals than a system lacking well defined rules or a system in which the rules are chosen without those goals in mind.

The aforementioned elements are design choices: choices that have been made in order to guide a decision making process with a particular goal in mind. For decisions of significance and consequence, design elements (some examples listed in the questions below) require significant consideration. Those designing a voting system must identify what type of vote is being implemented, who should participate in the vote and what criteria must be fulfilled for a decision to be made. And the design must be supported by structures that enable good outcomes, such as information systems: structures or processes used to provide relevant participants with the needed materials pertaining to an upcoming vote.

Some questions to ask when you are designing or assessing a voting system are:

  • Does this group share a goal or do they have different preferences?
  • Is there uncertainty or ambiguity regarding the options?
  • Who should get to vote and who should not?
  • How should the options be proposed and included in the ballot?
  • What threshold should be required for a decision?
  • What information, if any, do voters need, and how can the design ensure they get it?