Any group of people must, from time to time, make choices as a group, meaning that the choice will affect all of the members of the group and it can only be made “once”, in the sense that part of the group can’t make the choice one way while the remaining members make it the other way. Thus what to eat for lunch is not a collective choice, everyone is able to make that choice individually without concerning themselves with the choice made by anyone else. In contrast deciding what color to paint the lighthouse, or whether to build a lighthouse, is something that can only be decided once (as “compromise” color schemes are hideous) and, given that it affects everyone, must be decided in a way that reflects everyone’s preferences.
In order to deal with collective choices let us first divide them into two kinds, those that are simply a matter of preference and those for which there is, in an objective sense, a right or wrong choice. Which color to paint the lighthouse is simply a matter of preference, the only effects of this choice are how happy people are with the way it ends up being painted. In contrast, whether to build a lighthouse is a matter of objective right and wrong, either building a lighthouse will be good for the town or it won’t; people on both sides of the issue are not divided on the basis of whether they like lighthouses, but on the basis of whether they think it is good for the town (at least ideally, in actual situations there will always be some for whom this choice is of immediate advantage or disadvantage, such as the people who will be asked to construct it, and they will probably not have the town’s best interest in mind). Given the differences between these two kinds of choices it may very well be the case that the same decision procedure is not suited for both, and so I will examine them separately and then determine afterwards whether a unified procedure can properly address both kinds of choices.
First I will examine the case where the choice is purely a matter of preference, because that is the easier of the two to deal with. Our decision procedure (for both kinds of choices) should be designed to make the choice that is best for the group. Since these choices have no significant consequences besides how happy the members of the group are with them all we need to consider when designing this decision procedure is how to ensure that the group is maximally satisfied. Thus simple voting (one voter one vote) is a reasonably decent way of deciding the issue. Simple voting is, however, slightly deficient for two reasons. First of all it doesn’t reflect that people don’t all care about every issue equally, and so in some situations it may be better to give a significant minority that really cares about the issue what it wants rather than a largely apathetic majority. Simple voting also makes it possible that certain small groups won’t get what they want on any issues, simply because they always happen to be in the minority, and thus are effectively alienated, which is not something that is good for the group as a whole. To overcome these problems we may turn instead to a system that combines voting with bidding. Specifically each voter would have a supply of votes (replenished at regular intervals) that they can use to bid on various issues. The total votes bid are tallied and the side with the most total votes wins. The people who voted for the winning side lose the votes that they bid on the issue, while the losers get their votes back. Obviously in this system how much the voters care about the issues is taken into account because voters are free to divide their votes unequally, and in fact are expected to do so. Thus a passionate minority can get what they want when competing with an apathetic majority because they will bid more of their votes on that issue than they do on others, while the apathetic majority is concerned primarily with other issues and will allocate their votes to them instead. And a minority that loses on a regular basis will soon find themselves with an overabundance of votes, allowing them to bid more and more votes on the issues they care about until eventually they win.
And thus we can set the easy cases aside. Unfortunately the hard cases, situations where there is an objectively right and wrong way to make the choice, are not so easily dealt with; certainly voting is not the optimal decision procedure. If we vote on such decisions what we get will simply what the group thinks will make them happiest, but what they think will make them happiest may not actually be what will really make them happiest, or the group may irrationally prioritize short-term happiness over long term effects. A similar situation arises with happiness and knowledge in individuals. In general people act in whatever way they think will make them happiest. And certainly believing some things (such as that there is a treasure chest in their backyards) will make them happier than others. But believing things because they make us happy is a bad way to be happy, because often those beliefs will lead us to do things that result in unhappiness, canceling out the happiness that we achieved by believing them. To solve this problem we resolve to ourselves to believe only what is true and then, operating on the basis of those facts, satisfy our desires in every other area except what we would like to believe. Deciding what to do as a group when there are matters of fact involved concerning which outcome will genuinely be best for the group on the basis of collective preference is similar in many ways to deciding what to believe on the basis of which beliefs make us happiest. Which is not to say that every member of the group is voting on these issues solely on the basis of preference, but some of them may be, and many more may be misinformed (possibly because of the few who do have a vested interest), and because of this the vote may favor something other than the optimal choice. Of course it may be that this is simply the best we can do, but let us first consider other possibilities.
Given the analogy to the problems facing individual decision making it may seem that we need to impose some kind of reason on our collective desires, so that this reason will decide what is the objectively best way to satisfy our expressed collective desires. Of course we already have sources of reason as part of society, each individual is able to reason, some better than others. And so it may seem that a way to solve this problem is simply to get some individual or individuals to decide how to best satisfy our collective desires. This is essentially the traditional solution to the problem, namely deciding that someone should direct the group and then giving them the power to do so. But this solution is notoriously problematic. First of all it is hard to get the group to select the best leader (the person who is the best reasoner about how to satisfy the group’s desires). And, more importantly, it is hard to make this leader actually reason about how to best satisfy the group’s desires rather than their own desires or the desires of the people who put them into power. Because of these problems it appears to me that relying on a leader to reason about the group’s desires may be only slightly better than simply voting directly on the issues.
Since the problematic nature of that solution stems from trying to put individual reasoning over collective desires we might be inclined to improve it by putting collective reasoning over collective desires. But what collective reasoning consists in is not immediately obvious. I can, however, tell you what it isn’t; collective reasoning isn’t allowing the group to decide on the facts of the matter by voting, that is simply allowing collective desires to manifest themselves about what they would like to be the case. Since collective reasoning is not something that we actually engage in on a regular basis perhaps the best I can do is provide a speculative account as to what collective reasoning might be (which will have to be evaluated by someone else as to how well it reasons). To exploit the intelligence of the whole group in a way that doesn’t end up illegitimately reflecting individual preferences it would seem that we need to proceed in something like the way individual reasoning does, by breaking down the problem into smaller tasks where preferences don’t play a role and allowing the results of those tasks to inform our choice. Thus reasoning itself involves essentially three steps: breaking the problem up into smaller tasks, completing those tasks, and then using those results to solve the problem. And reasoning in this way can be done by a group. A number of people could be asked to complete each of the steps and their results are averaged, and thus the group decides what is to be done as a whole, as each individual solves only part of the problem. For example, when deciding whether to build a lighthouse first a number of people would be asked to divide the problem up into smaller tasks. Then a second group would decide which of those divisions into tasks was the best. Then each task would be given to a group to complete, and so on. Obviously at some points in this process a consensus may not be forthcoming. For example, people may be genuinely divided on which effects of the lighthouse’s presence (economic, aesthetic, etc) matter. Such cases are thus revealed to be a matter of individual preference (or at least that there is no better way to decide them) and can then be put to a vote in order for the process to proceed, and so this method of decision-making subsumes that which we use to decide matters of preference. To put such a system into operation would probably require computer automation in order to be efficient (and to avoid corruption). And thus, in action, it would probably be rather spooky, since the system, entirely unintelligent in its own right, would decide what to do for the whole group, apparently out of “nowhere”. Whether this way of dealing with collective choices is actually better than voting on them or allowing a single individual to lead is a matter that probably has to be decided by an empirical investigation or mathematical modeling. Certainly writing the software to test the theory would be an interesting challenge, and may be something that I may turn to in the future.