I just thought I would share my most recent thoughts about voting theory and about the voting systems I’ve had the opportunity to learn about over these past many months. Hopefully I will be able to illustrate any amount of understanding I have been able to develop for myself, and to outline my own journeying about this very intriguing topic so far. I am not an expert, but I think that it may be of interest even to those who are experts to hear what a novice like myself has been able to gather, and what areas I remain skeptical about, and perhaps why.
My interest in voting theory began when it hit me, as I was contemplating the very poor quality of our representation, how absurd the plurality system we have in place now truly is. I felt that there must have been somebody or some group some time in the past who had long ago figured out a better way to organize a democracy so that the public would have the capability to functionally and truly hold their representatives accountable. I started to learn about ordinal systems and Arrow’s theorem, and purchased a textbook on voting theory called “Economics and Computation.” Through that textbook and various online lectures, I was introduced to a swathe of different voting systems, and eventually I found this forum, where I have had the pleasure to discuss some of my own thoughts and to engage in the thoughts of other very insightful voting theory enthusiasts. It became more and more apparent to me that the problem of voting systems is much more intricate and difficult than any superficial conception might suggest.
Through my time on this forum, I have tried to remain skeptical, yet my mind consistently seems to return to three particular paradigms of thought regarding voting systems. Those paradigms can be seen as championing Condorcet-analogue methods, Approval methods, and STAR-based methods, respectively. These are related in that Condorcet-analogues are the Nash Equilibria of standard Approval and STAR. It is my current opinion that STAR or STAR-based methods are and ought to be superior to both Condorcet-analogue and Approval methods.
I understand the appeal of Condorcet methods, but I have always been very skeptical of them. They are for the most part resistant to tactical voting, and yet in my opinion they are too rigid. I believe that agreeing to support a Condorcet method sacrifices too much in return for this resistance to strategy. That is a value judgment, and you may disagree. However, I have seen too many examples for my liking of situations where Condorcet methods yields results that I do not personally find distributionally just. In those instances, the Condorcet methods have proved too majoritarian for my liking.
Similarly, I understand the appeal of Approval voting. It is very simple and resists some strategy (although not all that many of them) while yielding fair results. However, I do not think it is resistant enough to strategy, and along the same line there is too much arbitrariness in how voters ought to cast their ballots. The meaning of a ballot therefore is not even generally clear. It is a system I wish were better than it is, but in my opinion it just doesn’t work out as nicely as vertain other systems, namely STAR.
As for STAR, I believe it combines incentives in an extremely productive way that simultaneously and synergistically extracts relevant information from voters while putting that information to very good use to produce a fair and appealing result. On top of that, it is simple, straightforward, and easy to implement. There are some modifications I would personally prefer to be made to the standard STAR method in order to further improve (as I believe they would) the quality of results, but at the very least I presently believe that STAR is on the right track and well on the way to being what I would call a very superb voting system.
I have already alluded to modifications I would personally make to STAR to improve its quality—the main modification would be to stray away from the formal utilitarian framework that seeks to optimize score sums. Rather, in my opinion, what should be optimized is some metric of distributional justice. I gave examples of my own conception of such a metric in previous post, and through a derivation I could share I have also come to the belief that adhering to a particular sort of metric—one that amounts to exponentially-weighted scores—would not only produce superior results in itself, but would compound with the synergistic extraction of information to enhance STAR beyond what it accomplishes on its own already. That is just a hypothesis that would need to be tested experimentally.
Anyway, that’s all I have to say for right now. I would be interested to hear what anybody has to say, perhaps in defense of Condorcet systems, Approval voting, or any systems that I did not mention.