Social Experiment: Scientific Standards, Representativeness, and Fairness in Elections

What do you think of the following social experiment?

  1. Establish a set of labels, and assign your subject monetary payoffs for the success of each label in an election.

  2. Inform the subject that they will be participating in a small election along with other subjects, all of whom have different profiles of monetary payoffs for each label. Explain to them the voting system that will be used.

  3. Allow the subject to vote. Collect their ballot.

  4. After collecting (or fabricating) many ballots, produce a random sample of the ballots that contains the subject’s own ballot. Next, randomly reassign new labels to the candidates in the sample, keeping the reassignment secret from the subject, and additionally introduce a suitable amount of random noise in each ballot so that the subject cannot tell which ballot might correspond with their original. This will be an “incognito random sample.”

  5. Inform the subject of how the sample was constructed, keeping the relevant information secret, and allow them to examine the incognito random sample. Finally, allow them to freely choose a winner or ranking of the incognito labels based on the information they were able to gather from the sample, and record the results.

  6. Give them the prize associated with the original label that corresponds to the new winner or ranking they have chosen. Record their payoff as well.

Just an idea I had a while ago, I feel like it would be very interesting information to analyze. The outcomes of the experiment could be recorded, analyzed, and compressed into a sort of statistical compendium of “fair” or “representative” election results, against which perhaps the outcomes of voting systems for elections conforming to certain type parameters could be compared for some kind of standardization. Or, if the compendium were diverse enough, it could outline its own voting method.

Either way, I think voting theory needs more scientific basis and experimentation. Theorizing can only go so far.


Definitely agree, and think we should be doing more here (well, mostly at the new forum when it arrives) along those lines.

One experiment I’d like to do is a vote (monetary payoffs would be awesome), but where previous votes are revealed along with the current tabulation. And let them change their votes a certain number of times, maybe over the course of a week.

That could give us a real assessment of which methods were the most GTS (game theoretically stable). I’d predict that some would instantly stabilize (i.e. no one bothered changing their vote because it didn’t help them) while others didn’t.


You might like these polls that also deal with fairness: Majority rule rejected by a majority


Those results are encouraging. Still I think people would be less keen to support a utilitarian outcome when the parties involved are not “friends” per se, and when they themselves are a member of the block majority. There’s sort of a double standard people hold for themselves versus others—subject X wants everyone else to be utilitarian when it serves his own interests, but he will probably support majoritarianism when it’s convenient. It’s a total prisoner’s dilemma.

This social experiment is supposed to be designed so that the subject is basically forced to put himself in the shoes of a “generic” voter, and to come up with a solution that he feels will probably be sufficiently satisfactory for him knowing that fact and nothing else. Basically it’s forced empathy lol. They have to balance the risks of being a part of the block majority against those inherent to being a part of the more diffuse voter pool, and come up with a result that they think is reasonable or fair or representative enough to serve their interests “reliably” as a part of that pool.

My hypothesis is that fewer people would be utilitarian and/or people would be utilitarian to a lesser degree, but that it would still be a majority of people who prefer generally utilitarian outcomes to majoritarian ones. It becomes less a matter of personal strategy or posturing, and more a matter of spreading risk. In other words, my intuition is that compared to the questions you asked and the results you received, the strategies of the subjects would change in a somewhat significant way.

Unfortunately, the actual success of this kind of system for arriving at “good” results would rely on the subjects’ statistical reasoning abilities. And people are notoriously horrible at that. For example, say a ranked choice system is used, and suppose the subject ranks A first. Then they see the incognito random sample, and see that X is rated first by most people in the sample. They might then decide that it’s “most likely” that X stands for A, and choose X as the winner, causing a majoritarian outcome. But that kind of thinking totally ignores the risk that X does not stand for A. A more utilitarian strategy would probably do a better job of mitigating that risk. I’m not sure people would think that through on the first try or appreciate that risk well enough though lol. Maybe after repeated trials.

I think something like this could make a good game show lol. And something important you did that I neglected is to consider recording their reasoning as well.

As another point, I think if a subject knows how the experiment is going to work, they have very good incentives to vote “honestly”! The only drawback is a degree of necessary randomness.

To make this system work, there needs to be an auditing process though. People need to be able to rely on the system that is producing their random samples to be actually random.