The main idea: representative democracy fails in producing unbiased crowdsourced answers to difficult problems, as do its direct and statistical offshoots. Only unfettered trial by jury can be realistically considered a fair crowdsourcing mechanism.
My previous post tried to analyze the claim that mass (i.e. representative) democracy is the only realistic non-violent alternative to civil conflict and found it in need of severe qualifications, to put it mildly. Here I will try to deal with the other appealing argument made for mass democracy: that it provides an algorithm for crowd-sourcing difficult issues. May democracy be in better luck this time?
1. The claim
1.1 The experimental setup is well-known (1, or 2 for a lighter exposition). A researcher asks random people non-obvious questions, such as estimating the number of beans in a big jar or the weight of a large animal, giving them only a few seconds to look at the object. When several such guesses are averaged, a stunningly close figure emerges. Where any single individual failed, the crowd managed extraordinarily well.
1.2 The implications are clear. Winston Churchill may or may not have said that the best argument against democracy is a five minute talk with the average voter, but millions of such voters will make a far better impression and may inspire another, better known quote.
2. The issues
2.1 Let us first diffuse some of the hype: as this fine article clarifies, crowds do indeed come up with good guesses, but the averaging method is crucial to good results. If taking an average instead of a mean is enough to throw the ‘crowd’ off track with simple quantifiable questions for which the answers float around all of the time, let us imagine what monumental importance would the voting method have when infinitely more complex issues are discussed. Until more aggressive research in this promising field is conduced, we cannot be excused in awarding anything more than a benign question mark to the issue.
2.2 An especially important subset of this issue is presented by the institution of representative assemblies: common sense would indicate that the proper analogy would come between the bean-counting folks and the voters in a referendum, and not with the voting representatives in turn elected by the citizens. Even if we assume that the ‘wisdom of the crowd’ survives the voting process on difficult issues, assuming that it would also ace the test of electing representatives who’d then vote on the issues would mean stretching the implications of simple experiments too far.
2.3 An even greater issue with the analogy between mass democracy and bean-counting is interest skewing the results. To illustrate, let us conduct a thought experiment: let us suppose that a group of random citizens are asked a simple question and allowed a few seconds to guess: what results from multiplying 5’345 with 8’612 (let us just provide a large pool of subjects, as some rare minds are actually capable of carrying out such computations in the shortest time and must not be allowed to skew the results)? As a nice twist, let us clarify to them that they will be paid depending on their guess: a cent for each unit. Could we still imagine that the average of the crowd would be good enough? Could it perchance become noticeably higher? Indeed, we would agree that the inclusion of personal interest suffices to break the crowdsourcing mechanism.
2.4 This is a far better analogy to the actual voting process, both in elected assemblies and general elections, than a series of innocuous guesses would be. Voters have interests, and the voting pattern will clearly be affected by such interests. Trying to defend mass democracy by retorting that the interests of many million voters shall cancel out, leaving us with a fair result, will not do.
2.5 The fact is that almost no one has any direct monetary interest in the reduction the role of the state.
Any monetary interest involved in voting is always slanted toward bigger government: we can readily predict that the result of ‘crowdsourcing’ political decisions will likewise be heavily slanted towards feeding a leviathan state, not solving the actual issue at hand. And let us not even get started on the interests of the ‘legislators’ who rarely ever vote out of conscience at all!
2.6 All in all, the claim that mass democracy amounts to a fair crowdsourcing mechanism for tough political decisions is very unlikely. Until monumental, ground-breaking and counter-intuitive research comes forth to make one suspect otherwise, we have little to expect form electoral results.
3. Referenda and Statistical Democracy
3.1 Attentive readers may come to see that some of these issues could be solved. Can we try to use the logic of crowdsourcing to ‘improve’ democracy just as we used the logic of virtual civil war for that same purpose in our previous post?
3.2 Some have proposed doing entirely away with representative assemblies: surely in our digital age it would cost little – if anything – to hold constant referenda on any issue in some digital format. Why, then, insist on putting a layer of elected representatives between the ‘people’ and their ‘will’? Direct democracy would, after all, seem to work fine for the rich Swiss with their old-fashioned landsgemeinde.
3.3 Or, following yet another proposal, an assembly of elected members cannot represent the will of the people in the way that an assembly of randomly selected citizens would. Supporters of ‘statistical democracy’, who count among their numbers the great H.L. Mencken, insist that the well-known refrain of political machines playing certain issues such as to herd the vote in a given way could no longer work in a randomly selected assembly. And anything that would cut on the tedium, expense and general waste of electoral campaigns would be well-heeded.
3.4 Though both these proposals would come far closer to functioning as proper crowdsourcing mechanisms than current mass democracy, they would both still fail in general. Neither would address the issue of interests skewing the results, though both (and especially referenda) could mitigate the effects as compared to the current situation.
3.5 With both proposals there will be an issue with the long-run interests of the decision-makers. Since neither referenda nor statistical democracy addresses the issue of interest skewing results, we must factor the role that these play in the decision taken. There may an argument to be had that elected representatives often manage to sit for long periods of time, and that they acquire a limited version of the Hoppean interest in the capital value of the country. Should this be the case, than both proposals would cancel any such gains, since citizens are not personally and immediately affected by drops in the capital value of their country.
3.6 Whether in the end capital consumption would actually be higher or lower in these three versions of democracy cannot be stated a priori. On a purely instinctual basis, I will speculate that mass democracy will outperform direct democracy, but that statistical democracy will turn out to be the best of the three.
3.7 And if neither direct nor statistical democracy can be expected to produce true crowdsourced decisions, both also fail at the test of civil strife, for the very same reason that current mass democracy does: we may decide an issue by a referendum or by calling 500 random citizens to the vote, but we will have no way to factor in the intensity of opinions. If a minority insists on its dissent to the point of violence, civil strife will result from all these systems.
3.8 Here it is direct democracy which may be saved, again by making the vote transferable, but in doing so any remaining pretense of the process yielding an unbiased crowdsourced guess would have to be dropped. No such solutions can be found for statistical democracy.
3.9 Thus, not any of the three versions of democracy we analyzed can either satisfactorily crowdsource decisions or ensure civil peace.
4. Trial by jury
4.1 Is there no system which will provide for a reasonable way to crowdwource difficult decisions than? I submit that there is such a system, and an ancient one by that: trial by jury. Juries suffer from fewer of the handicaps that bar democracy, in any form, form producing reasonable, unbiased decisions by the crowdsourcing mechanism.
4.2 First, unlike the complex and multi-faceted questions any legislature is called upon to answer, an actual trial proposes a simpler problem: has there been any punishable wrongdoing in a specific case at hand? Certainly, different legal norms may have momentously different implications down the road for any society, and different individuals may evaluate these implications differently, if most will even bother.
4.3 Still, it is not wholly unreasonable to expect the individual biases in such appraisals to cancel out, yielding a fine sentence. Indeed, any reasonably competent juror could provide a good answer based on nothing but his conscience. Many such jurors could even provide a reasonable approximation of a just rule.
4.4 Also, unlike MPs, jurors have no personal stakes in the issue at hand: they are called upon to express their opinion on a matter which does not touch any of them personally, completely removing both the undue influence of personal interest and the incentive to resort to violence should the result fail to conform to any individual juror’s opinion. Though a determined minority could still refuse to submit to verdicts, allowing jurors to refuse to serve (subject to penalties, see paragraph 4.13 below) could lower that risk, and ultimately such conflicts would persuade many of the need for the dissenting minority to secede. All in all, trial by jury does provide us with a reasonable – but not perfect – mechanism for crowdsourcing decisions.
4.5 An added benefit of trials by jury would be to speed up the ‘education’ of the community. If a community sports moral norms which are detrimental to its development, Hayekian logic would expect this community to loose members (subject to complications). If such a community adopts trial by jury, the process is considerably sped up. Norms materialize directly in the form of sentences, and not indirectly by the long and tortuous road toward legislation.
4.6 There will also be an instinctive understanding among community members that the norms they cherish do have real world consequences: had many typical underdeveloped countries been run by juries, their citizens would long since have grasped that the general lack of foreign investment within their borders is caused by the norms they themselves approved of. If we consider both these factors, we could easily expect trial by jury to swiftly reward efficient communities and punish less efficient ones. Indeed, Hayekian Cultural Evolution in its distilled form!
4.7 Finally, just like the transferable vote minimizes the importance of voting qualifications, trial by jury in turn minimizes the importance of the rest of the government. it matters less that strong cabinets and/or parliaments fall prey to rent-seeking groups of interest if their every decision may be struck down (or, as the parlance goes, nullified) when it matters: at the individual level. Trial by jury could thus minimize the adverse effects of otherwise sub-optimal political structures.
4.8 All this being said, trial by jury as practiced today in a few (lucky) countries if far from optimal. The version trialed by the ancient Athenians* would suit our purposes far better.
4.9 To start counting the shortcomings of what trial by jury has evolved into, 12 is too small a number to be statistically representative of the general opinion. Larger juries, to the tune of 50 or more jurors, would be needed to produce credible guesses.
4.10 Also, much of the ‘magic’ of crowdsourcing is lost – indeed to the point of becoming counter-productive – if the individual participants are allowed to influence each other, notably by debating their guesses. So, having jurors not only reach a verdict together, but even sitting together would not do. No individual juror must have any contacts with any other person until he produces his individual verdict.
4.11 But the larger issue is that trial by jury cannot coexist with any other legal rule or meta-rule: one cannot have both trial by jury and legislation or precedent. The jurors must be allowed to express their individual conscience when judging a case and be influenced by nothing else, otherwise the results of the process will be skewed. A proper jury must function in the Athenian way: let the plaintiff and the defendant specify in advance what sentence they deem proper, and have each juror individually vote on either of these proposals. If the resulting sentence contradicts any statutory law or precedent, then so be it.
4.12 Trial by jury has a long way to go if it is to realize once again its potential as a good crowdsourcing mechanism. Always seeking to ease the pain of such transitions, I propose that any community interested in the adoption of true trial by jury should do so in a gradual, voluntary fashion. Let it award trials by jury to any defendant who has priory agreed to pay a given periodic fee. The fee should suffice to cover the expenses of calling in 50 random citizens from their day lives and would allow for an initially small subset of cases to be settled in this way.
4.13 Should trial by jury prove itself to be as appealing as I propose, than the fee would probably seem trivial when compared to the virtual immunity from the State’s legislative fiat that it would grant. An added benefit of a voluntary parallel system would be to do away with the requirement for compulsory jury duty without skewing the results by self-selection: any participant refusing to serve when called for jury duty would then be ineligible for such a trial should he himself become a defendant.
5. Real solutions
5.1 One must not forget that crowdsourcing is not the optimal solution to a problem. The optimal solution will be produced in a system where there’s an incentive to find the truth. In the field of Law, such a system has been often described by libertarians. For our own take, refer to Clayton’s series of posts.
5.2 It is only when such a system cannot be implemented that second-bests are to be found. Crowdsourcing provides a decent plan B, especially compared to the system we often fall back to: providing an incentive to give a specific answer, not to find the truth. If incentives cannot be made to work in our favor, it stands to reason that their influence should be curtailed as far as possible.
5.3 To return to our initial experimental setup for an example, the best way to estimate the weight of an elephant in not to average the guesses of a hundred random people, but to actually weight the beast. Should this not be possible, it will be silly to ask a hundred people to guess any pay them based on their guess.
5.4 As I myself am fond of repeating ad nauseam, whenever the actual solution lies in the future Free Society, a good approximation is to be found in a system of City-states, where an incentive exists to discover the ‘right’ (i.e. efficient) laws. What these laws will be we cannot say in advance, competition being the whole point of the system.
5.5 Let us just point out that the solutions that these competing entities (tomorrow City-states, private arbiters the day after that) will find could also be meta-solutions: systems designed to find a good answer instead of the answers themselves. It would thus be not at all surprising if competing City-states or arbitration firms could discover that adversarial cases where no conciliation is possible could better be solved by trial by jury. Crowdsourcing is often a good plan B, but may sometimes be a good plan A too.
5.6 To conclude, in these posts I have tried to analyze the alleged strengths of mass democracy. Arguendo, I renounced the first and even second lines of argument that libertarians use in such debates. Even allowing our demophiles ample spaces, even assuming the need for the State, and even assuming he need for states larger than cities, even under such conditions we have found mass democracy to fail. It does not prevent civil unrest and it does not provide unbiased crowdsourced decisions. Other systems provide for far better solutions to concerns which, in any case, only exist because political competition is still so weak in our day and age.
* Kakugo provides the following concise description of the Athenian jury system, as it existed during the Peloponnesian Wars: “Athens enrolled 6000 jurors at the beginning of each year. They were assigned at random (by ballot) to each different case. According to the sums at stake the jury was made up from 200 to 1500 members, except when a new kind of case was to be discussed. In that case all 6000 jurors were summoned. Jurors were paid 2 obolos per day for their troubles starting with Pericles. Cleon increased the sum to 3 obolos. No judges were present (or indeed existed in the modern sense) and magistrates had non-legal functions, like declaring a session open and closed and paying the jurors, and for most of Athens’ history the rank of magistrate could only be held once in a lifetime. The only thing remotely resembling a traditional legal system was the defendant and the accuser could hire professional speechwriters (logographoi) but they had to deliver the speeches themselves. Did it work? That’s the 100 trillion Zimbabwe dollars question. I suspect Mencken may have been inspired by this system.”
See Roderick Long’s essay for a full exposition of the Athenian system.