How to vote on things by not voting


The overwhelming of the vote of the World Science Fiction Society to send the 2023 convention to China created controversy and not for the first time. They had a similar problem with the Hugo award nomination process which is even easier to overwhelm with a much smaller concerted group. They solved this problem by making the nomination rules much more complex, with an algorithm to attempt to de-rank candidates that appear in "slates" on many ballots. The attackers mostly gave up before the rule came into effect -- it takes 2 years -- and it didn't change end results much so a majority of people are generally satisfied. At least until it happened again with the site voting.

Changing the voting rules may or may not work, but it's the wrong approach. For this, and many other things in democracy, voting is actually the wrong answer. Democracy, which literally means "power to the people" is about knowing what the will of the people is so that it can be enacted.

Voting, however, when you get less than massive turnout, doesn't measure the will of the people. It measures a combination of the choices of the people and their enthusiasm about those choices. Often it measures the latter more strongly. In the USA, which gets 60% voter turnout in Presidential elections and much less in others, the reality has become stark. You don't win elections by convincing people to join your side or switch sides. You win elections by doing the best job at rallying your existing supporters to come to the polls. Almost all the effort in election campaigns is spent on GOTV -- "get out the vote." All those negative ads you see? They aren't there to convince people to switch sides. They are there to scare your weak supporters into feeling motivated to come out.

I don't think that's a good way to choose leaders. That's not a simple question. There is merit to the idea that enthusiasm is a worthwhile thing to measure. It is fair to wonder if elections should be decided by people who barely care or didn't even look at the issues. Some countries have mandatory voting and get turnouts over 90% (and some get those high turnouts with optional voting, too.) Low turnout is an essential tool in doing voter suppression. A party likes low turnout if it's better at rallying the troops.

The idea of voting is deep set in our democratic bones. But in reality there is a science of measuring public opinion through sampling rather than voting. We often correctly deride the results of various polls done during elections, but these polls tend to have worse biases than voting does. As I like to joke, "100% of the people in our phone poll said they answer unknown phone calls and are home right now."

Well done sampling, where participation is compulsory if you are sampled, and the sampling is good, can provide a much more accurate measure of the views of the group than voting can unless you get very high voter turnout.

This won't happen in most places because usually one party is benefiting from the problems with the current system, so it will block any effort to change it.

So I'll return to the World Science Fiction society, which wants to give awards to the works its members think are most deserving, and host its annual party in the city its members prefer. (Some of what I write below will be germane only to that situation, but some is general.) It uses voting to do this, in fact voted nominations as well as a voted final ballot. And there is a general feeling that both are often a popularity contest. They use Australian style STV ranked choice voting, which helps improve the measurement of voter choice a bit. In that system, something that is everybody's 2nd choice but not the leader in widely split 1st choice votes will win.

At the same time, it's easy to get flawed situations where what is clearly the inferior choice (in the 30,000ft view) will lose. This is true of all voting systems. And as we've learned, voting systems can be gamed when there is low turnout, or even when there isn't when votes can be purchased. (In the WSFS system, memberships and votes can be purchased, though doing it in an organized way, with central funding and coordination, would be considered inappropriate.)

Sampling has many advantages. In fact it's superior in just about every way except that it violates the idea that "every voice is counted." Indeed, every voice is deliberately not counted. In low-turnout voting, many voices are not counted, but it's their choice and so we accept it. Even though we know it introduces bias in the result.

In particular, sampling is much more resilient to coordinated attacks. In the 2015 award case, the nomination system began by letting any member submit 5 names on a nomination ballot. The top 5 most named works became the final ballot. In a diverse field like "best short story of the year" there are hundreds of possible stories, and the 5 winners will each only have a fairly modest number of nominations. There might be 5,000 members, but fewer than than 1,000 will send in a nominating ballot and less popular categories might see only a few hundred. The top candidates may get less than 100 nominations.

This means if you can get a small group together and nominate as a bloc, you can easily overwhelm the nominations, and this is what took place. It's harder to do this in the actual vote, though not entirely impossible. That did not happen.

A sampled system would be immune to attacks from small groups. A very large group might get some chance of success in smaller categories, but find it very expensive to totally overwhelm the system.

Naturally there are some challenges in sampling compared to voting. The fact that 90% of members don't nominate in some categories at all is in part due to enthusiasm, and in part because they have not been reading the works in question and don't have a view. One could sample the entire population, which is best, or one could sample those who indicated they are willing to be polled. In some systems, there can be a penalty for not answering the poll (even if the answer is just, "I have nothing" since people prefer not to admit that.) Using only those who indicate they are willing to be polled is more vulnerable to attack because attackers will always indicate that, but it is still more robust than voting based nomination. There are a few other tricks, like allowing people to "pass" and adding a new random member to the sample. This biases for enthusiasm but not nearly as much as voting does.

Sampling might also encourage more participation. "Hi, you've been randomly selected. We hope you'll make the effort to tell us your views" could encourage people to look at things and think about them more than just the general call.

This is even stronger during the final vote on the slate of 5 (now 6) finalists. For these awards, there is now a practice -- which I started in 1993 -- of making electronic versions of almost all of them available to all members. Those who are sampled can be encouraged to take the effort to read them all, especially those that are not very time consuming. The attackers, however have the advantage that they already have planned their vote without bothering to have to read things.

Sampling is so robust it might even withstand the purchase of almost 2,000 memberships by outsiders. (I will define an outsider as somebody who has never previously been a member who purchases only a supporting membership. I would not include the 30 Chinese who purchased virtual memberships that let them watch remotely.) With around 5,000 insider members and 2,000 outsiders, it would be possible, though much more difficult for the outsiders to win without substantial support among the insiders.

But yes, this means asking for, and paying strong attention to the opinions of members who don't have a strong enough opinion to have bothered to vote in a regular voting system. (The WSFS site vote system also requires voters to pay a fee, around $50, which becomes a supporting membership and contribution to whoever wins. That would not be practical in a sampled poll, and an alternative form of raising those contributions would be needed, such as a $10 fee from all members. This voting fee is so large not because the cost of providing services to supporting members is high, but to explicitly give a kickstart to the new effort in each city.)

Sampling is a superior measure of the will of the eligible voters. If you believe the goal is to best measure that, it is the way to go. If you believe that the result of voting -- the will of the motivated voters -- is the goal, then you would not want sampling at all. And on either side, people are likely to be bothered at the elimination of the direct voter-vote connection, the literal exercise of "power to the people." It feels strange to think that it's a good idea to not query your personal wishes in trying to accurately learn the wishes of the group.

This debate has come up a lot around the U.S. census. The constitution of that country says the census must be done by enumerating. Long ago the scientists who produce the census realized that sampling would produce significantly more accurate numbers. To count everyone, you must deliberately not count some people. Even with this understanding, one side does better with enumeration than it would under sampling, and is able to block it because of the wording in the constitution. This has a non-trivial effect, and de-ranks and marginalizes communities who have a harder time participating in enumeration or voting.

Of course, the best of both worlds is 100% turnout voting or complete enumeration. In today's electronic world, that's not nearly so difficult as it might have been imagined to be. Once you accept electronic voting, compulsory electronic voting is not a big step, even if it's just to say "I abstain," which can remain an available choice. One problem is that for elections that really matter (ie. not book awards) there are security risks to doing electronic voting. With things like WSFS, the rule can just be laid down, "You participate in the voting, or you don't come." Or even "You can buy a non-voting membership for $100 more than the voting one, but if you buy a voting one, you have to vote."

Overt sampling

There have been other proposals for unusual forms of sampling. For example, for an election, picking 1% of voters in a country at random and telling them in advance that it is now their duty to study hard, watch the debates, get fully informed and vote. With a large enough sample this will tell you how the whole populace would have voted if they did all that study. You even pay them or give them time off to do the work. One problem is that if it is revealed who they are, the campaigns will go nuts trying to reach them. It's even considered possibly a good idea to let them -- it vastly reduces the cost of the campaign, which provides immediate campaign finance reform and eliminates the need for most fundraising.

You may not even have to limit how much the campaigns harass the picked voters because they don't want to annoy them. However, ordinary citizens would harass them so it almost means a week of sequestration is necessary.

This may come closest to what some would view as the best goal, "If the whole populace were fully informed and studied the election closely, how would they decide?"

Add new comment