Workshop Report Now Available

Last fall, we held a workshop at Trinity College Dublin on “The Lottery as a Democratic Institution.” The workshop was organized by me, Gil Delannoi (Sciences Po), and Oliver Dowlen (Queen Mary, University of London), and sponsored by Sciences Po, the Policy Institute, and Trinity’s Arts and Social Sciences Benefactions Fund. Our report from the workshop has now been published by the Policy Institute. It can be found online here:

23 Responses

  1. Hi Peter,

    Thanks for posting the report.

    I disagree, of course, with the main (or one of the main) claims of the report. I think that the primary justification for sortition is that it is a tool for representing the interests of the population in public policy (i.e., that it is democratic – unlike elections).

    Despite the objections put forward in the report, I think that establishing this capability, via descriptive representation, is straightforward. In a nutshell, the group of delegates (regardless of selection method) promotes its own interests, when the delegates are selected via sortition, then the interests of the delegates are aligned with those of the rest of the population due to the statistical properties of the selection procedure. (See the presentation and comments thread here.)

    I hope to post a more detailed response in the near future.


  2. Yoram,

    I haven’t had time to read the report but, if my memory serves me correctly, the issue of the lot as a tool of descriptive representation was a contested topic during the seminar. The fact that the principal authors of the report have tended to view this as a “weak” use of sortition, or sought to subsume it under the negative (prophylactic) function of the lot might be the reason this is disputed in the report (assuming that is your claim). The (singular) “Lottery Thesis” has been a characteristic of this seminar series and the fact that plurality is now acknowledged is a distinct sign of progress.


  3. The main conclusion from the summary of findings:

    “We argue that random selection makes its strongest contribution
    when it selects citizens to function as impartial guardians of the political system. This means selecting citizens at random, not to make policy or enact laws, but to protect the integrity of the political process—by making and enforcing legislative ethics standards,
    for example. Random selection’s strongest contribution is to the prevention of corruption and/or domination; the fact that it enables descriptive representation, while undeniably true, is less important to politics.”

    “We”, in this quote, means “Oliver Dowlen and Peter Stone”. I’m not sure what Gil’s views are and would dispute any implication that this was a consensus view of the seminar participants. This is a good illustration of how the (speech acts) of some animals are a lot more equal than others, a general factor of group dynamics that has to be taken into account when designing allotted bodies. Whether this is an illustration of a lack of “descriptive representation” or “corruption/domination” is a moot point. Given that there was no vote, the “Summary of Findings” certainly has no democratic legitimacy. If we seek to lecture others on the failures of democracy we might want to put our own house in order first.


  4. Keith: From the report, page 6, note 2: “The authors would like to stress the report’s status as a discussion document, designed to stimulate the exploration of the ideas expressed in the seminar series which generated it. It does not represent any common or defined position taken by the authors as a group.” The report explicitly denies that it reflects a common worked-out position where every position is endorsed by all 3 authors. It certainly makes no pretense that it reflects a common position worked out by the workshop’s participants. The workshop wasn’t a democracy, Keith, and your suggestion that the Dowlen/Stone mafia is brutally repressing you is, to be blunt, rather obnoxious.

    Yoram: Let me give you a very quick response; perhaps I can post further on this later. It seems to me that there are two arguments out there trying to justify sortition using descriptive representation.

    First argument: 1) there is such a thing as the common good, 2) this common good is some sort of function of the good of the individuals in the population, 3) if the entire population was able to make a decision in a certain way, this common good would naturally result, 4) an AC is statistically indistinguishable, in terms of composition, from the population from which it is selected, 5) therefore, the AC will reach the same decision as the entire population, i.e., realizing the common good.

    I take this is something like your argument, Yoram. Short response: I’ve read enough social choice theory to be very suspicious of any claim as to how to identify the common good of a group as an aggregation of individual members. It’s really hard to do, especially if you want the AC to do more than just vote–i.e., if you want them to collect information, deliberate, etc.

    Second argument: 1) we want government by consent, 2) an AC looks like the population as a whole, 3) therefore, we can take the AC as representing the entire population, so that the consent of the AC counts as the consent of the entire population.

    Something like this seems to be Keith’s argument. (Again, I’m going fast here.) Short response: I don’t believe in government by consent of the governed. It’s never happened, and it never will. It’s a foolish and dangerously misleading idea to suggest that government needs your permission to exist. (This is NOT to say that democracy is a bad idea, merely that it needs to be justified in terms other than government by consent.) All efforts to show that government by election “counts” as government by popular consent have proven badly flawed, and the current effort to show that government by AC “counts” even better, while arguably not worse, is just as flawed.

    I think that defending sortition by appeal to, for example, its ability to prevent corruption and domination avoids both of these pitfalls. First, we don’t have to deal with the problem of aggregating the interests of the entire population in order to talk about these things; all we have to do is presume that there exists some kind of good policy, and that things like corruption and domination get in the way of realizing it. And we don’t need to touch the idea of consent at all.

    That’s a very short and probably unsatisfying answer, but I do think it’s fair to say that Oliver and I have a very different take on the virtues of sortition than you (Yoram and Keith) do. (Isn’t it amazing how much you two agree!?) I hope this helps explain why.


  5. A final report would normally accurately reflect what actually took place at the conference/workshop. The main conclusion of this report is in fact the authors’ original theses (as reflected in their two books), and in this respect the verdict is more akin to a show trial. I wouldn’t have a problem with this if it were just a co-authored paper, rather than the proceedings of a public conference. As to whether or not a workshop is a democracy, if you preface the adjective “deliberative”, that would be the general norm for academic discourse.


  6. PS The default position for the chairs of academic panels is the self-denying ordinance; if they do participate themselves it’s generally rather apologetically — especially if they have strong views (or a book to promote) on the topic in question. The main conclusion of this report is about as far removed from this norm as is imaginable.


  7. Peter,

    Moving on to the substantive issue (descriptive representation), my focus on consent was purely for the Manin paper (in which he argued that consent was the reason for the triumph of election). I’m more disposed to the first argument, although I’m hesitant about the Rousseauian language that you employ. The social choice theory problem refers to the difficulty in aggregating preferences across a number of issues/domains, and doesn’t apply to up/down voting on a single legislative bill. I agree with you that it’s difficult if you want the AC to do more than just vote–i.e., if you want them to collect information, deliberate, etc., that’s why I rule these activities out for an allotted forum.

    But I’m glad you’ve moved on from attacking straw men such as descriptive representation being akin to “someone like me selling my flat without my permission.” So, notwithstanding the report conclusions, there has been some progress.


  8. PS: The claim that representation is a “weak” use of sortition is only true if one define a “lottery principle” in terms of arationality (the blind break). If you choose a different lottery principle, for example to generate a RATIO then this is also a unique use of the lot, as stratified sampling will not work as there is no way of knowing in advance what strata to sample for in order to produce a) the necessary cognitive diversity for the epistemic benefits and b) a statistically “representative” sample. Apart from gender, age, etc, who is to say what the relevant factors are? If this is true, then descriptive representation is an extremely strong use of the lot. Needlesss to say Dowlen and Stone deny this, as they have already defined the Lottery Principle as arationality — this is the problem if your starting point is theoretical analysis (Chapter 1 in Dowlen’s book), as opposed to saying that sortition is just a man-made tool, that can be used in a variety of different ways.


  9. Keith, I don’t have time to respond to your substantive remarks. But two comments must be made. First, your characterization of the event is wildly inaccurate. We were not, and never pretended to be, neutral organizers/chairs of a conference for which we sent out an open call for papers. We prepared a draft document, and asked people to participate in a discussion of it. We then revised the draft document into a report based on the discussion of it. The idea that we were supposed to abstain from active participation in such a conversation is beyond absurd. The views expressed in the report are ours; if you’d like to prepare your own report denouncing ours, perhaps I can interest the Policy Institute in it.

    Second, if your goal is to make the world of sortition advocates as difficult and divisive as possible, by all means keep using wild overblown language like “show trial.” But if you’d like to discuss the substance of the issues at stake, I must respectfully suggest that you adopt a different tone.


  10. I did adopt a different tone for my comments on the substantive issues, but now you “don’t have time to respond”. Look forward to your response once you can fit it in to your busy schedule.


  11. FWIW, I understood the report to represent the opinions of Peter and Dowlen. In general I don’t expect academic workshops to generate some sort of a consensus or majority opinion reports – is there such a tradition?


  12. Peter,

    I am not sure whether my argument falls under the first or second type of your taxonomy. I am not making any explicit “common good” assumptions, but of course you may consider my assumptions to be implicitly implying some sort of a common good concept. E.g., I assume that a small group can represent its own interests under certain circumstances – does this mean that they are identifying their common good? But if so, then any sort of normative discussion of public policy would require a concept of common good. In particular, asserting that sortition is a tool against corruption and domination would be very difficult until you define what constitutes good government. Without that, how would you even know if a certain government is corrupt or based on domination, and why would it matter anyway? And while we are rejecting any notion of “common good”, we might want to doubt the “personal good” concept and the ability of individuals to represent their own interests – these are not any more solidly founded than the small-group-level ideas.

    Regarding consent: I agree that formal consent is a silly concept that should be avoided together with the entire “natural rights” framework. However, it is certainly expected that in a democratic system most people would consider the system to be well functioning and representing their interests, and thus supporting it, or “consenting” to it. This expectation can be operationalized by setting up an allotted monitoring body that expresses satisfaction or dissatisfaction with public policy and the procedures of government. (I presume this would fall within the “integrity of the political system” role that you endorse.)

    Anyway, I would be interested in your critique of my argument as presented above in its own terms.


  13. Yoram,

    Corruption/Domination normally indicates the vulnerability of the system to manipulation by the powerful. Even if the individuals that secure government office under a corrupt/dominating system ruled for the general good (as they saw it) and with spotless personal integrity, the system would still be corrupt/dominating. The blind break inhibits any such manipulation but, in the process, also inhibits the selection of individuals on account of the common good (or, for that matter, any other reason). This is why the blind break can only ever be a secondary normative principle, defending the existing political system (whatever that may be) against manipulation.

    My argument for descriptive representation is also a secondary principle as it enables a) sound epistemic judgment and b) political equality. I suspect that you are arguing for sortition as a primary principle (you use corruption/domination in an unusual way), and this is perhaps why we continue to talk past each other. Rousseau argued that individuals in a corrupt society would vote in their own interests, rather than seeking to uncover the general will (the common good). Would that better describe your own view of corruption as, presumably, sortition would be of no value in such a society? In my view, such a system would not be corrupt, although it might well send us all to hell in a handcart.

    In sum, is corruption a structural principal (a view that I share with the Blind Breakers) or a substantive moral value (as in the Incorruptible Robespierre)?


  14. Peter, one further response to your:

    >I’ve read enough social choice theory to be very suspicious of any claim as to how to identify the common good of a group as an aggregation of individual members.

    Electoral democracy is just as vulnerable to the preference-aggregation problem (unless you follow Jeffrey Green into the Schumpeterian black hole). In fact the only democratic decision process that is NOT subject to the impossibility theorem is when a statistically-representative body votes on individual legislative bills.


  15. “1) there is such a thing as the common good, 2) this common good is some sort of function of the good of the individuals in the population”

    What kind of sidetrack is this? No, you don’t have to assume the existence of a common good function. Sure, sometimes we will be able to say one outcome is better for the whole of us – I hope you agree with “the existence of a common good” in that sense, else what’s the point of trying to get good government? – but if we can’t agree on what the common good is, it would at least be desirable that we are equally empowered/disempowered to seek our own interests, or our own ideas of the common good.


  16. Harald,

    Agree. Sortition as a selection mechanism is agnostic as to whether an allotted assembly would decide on the basis of the general will (the common good) or the will of all (aggregated preferences). In practice it’s likely to be a combination of the two, but there is no way of knowing (and it doesn’t really matter). We await Peter’s explanation as to how preference elections are not vulnerable to the social choice objection that he raises.


  17. While we wait, Keith: This …

    “In fact the only democratic decision process that is NOT subject to the impossibility theorem is when a statistically-representative body votes on individual legislative bills.”

    … is wrong. Even a perfectly representative assembly (however you define representativity) is vulnerable to preference cycles. You can conceal their presence by just allowing yes and no voting, but the cycle still is there – it’s a property of people’s preferences, not the voting system.

    If you use just yes-and-no voting, cycles will be resolved differently depending on the order you vote on the issues, that doesn’t sound very desirable to me.


  18. IIRC, a random dictator does satisfy Arrow’s conditions.

    But why would that matter, in any case? If one can assume preference cycles in a population, why can’t one assume preference cycles for an individual? Or vice versa – if we assume that each individual can always order their preferences in some black box fashion, why not assume that a group can always do so as well in some black box fashion?


  19. Thanks Harald, my preference has always been for ad hoc allotted bodies for each legislative bill, each returning an up/down verdict (as with trial juries). My understanding of the impossibility theorem was the difficulty of aggregating preferences over a number of different domains. Say if the general population were fiscally liberal but socially conservative, those two positions are impossible to represent in a two-party elective system, but the same (or multiple) allotted bodies could vote liberally over fiscal bills and conservatively over social bills. Wouldn’t that get round the social choice objection?


  20. Peter Stone,

    I would like to thank you and your colleagues for your efforts in trying to create a form of government that is more responsive to the common good.

    I agree with Chouard that terms like “electoral democracy” and “representative democracy” are oxymoronic. When the many select the few, the outcome is oligarchy. I also disagree with the statement, “Democracy is defined in terms of self-government, which is taken to mean government by the consent of the governed.” Where there is self-government, there is a citizen state. Citizens govern on their own behalf. They do not consent to have others govern for them. Using a term like “liberal democracy” only spreads confusion. Liberal democracy refers to an ideology, not a form of government.

    One of the chief advantages of sortition is that it establishes political equality. This only applies however when randomly selected citizens serve as officers or legislators, i.e. when they have real political power. Citizens serving in an oversight capacity is not a bad idea, but why the compromise? Let them be the government rather than oversee it.

    The oversight committee is empowered to veto legislation for three reasons. Number two specifies that it may review and return “non-financial legislation that it believes will not achieve the objectives the government claims.” This provision, in my view, defeats the entire exercise. If citizens are not empowered to veto financial legislation, the source of the corruption and social decline for which we seek surcease, they are not empowered in any meaningful sense.

    Which brings us to the issue of expertise and training. After all, what are the key decisions to be made by legislators, randomly selected or elected? The key decisions are choices between war and peace, choices between alternative ways of spending money. Should we build bombers or schools? Should we subsidize oil drilling or the manufacture of solar panels? These are the critical decisions and we amateurs are the most qualified to make them. There is no training necessary. The legislative process is arcane so that those in power can hold on to what they have and maybe get more.

    If there is concern about the consequences of selecting office holders by lottery, why not start out on a small, local level and see how it works. The town of Rockport, Massachusetts is government by a board of five selectmen. Why not have them chosen by lottery, serving one-year terms? What harm could come?

    One could use vetted lotteries to select Supreme Court judges. A large jury, randomly selected, could do the vetting. Similarly, sortition could be used to eliminate the spoils system. A large pool of vetted candidates could be established. Openings would be filled by sortition.

    I believe there is a growing awareness that the constitutional oligarchies that have ruled the western world for the last two hundred years have outlived their usefulness. We need a new form of government. In “Paradise Lost, Paradise Regained: The True Meaning of Democracy” I make the case for democracy as a viable alternative.

    Arthur D. Robbins


  21. Yoram, formally speaking a random dictator goes against Arrow’s “non-dictatorship” criterion, but it does highlight a problem with that criterion (it only considers equality after the collective decision, not before).

    Keith, no. The objection isn’t when there are independent domains, but when there are more than two possible options in a single domain. For instance, say you have only money to build one power plant, and it can either be wind, natural gas or nuclear. It is then possible for a majority to prefer wind over gas, a different majority to prefer nuclear over wind, and a third majority to prefer gas over nuclear.

    In that case it’s easy to see that if you allow only two options on a vote, the order you ask questions in will determine the outcome. A voting system with multiple alternatives does not get around the problem either – although of course it can be designed to give a deterministic answer in situations such as these (not all answers are equally good either).


  22. Harald

    Many thanks for the correction. Clearly this is a problem that applies to any form of democracy that attempts to mirror preferences in the wider population, and Peter is wrong to attribute it purely to sortive democracy. The only form of democracy that would be immune from it is the Schumpetian variant.

    But when multiple domains are taken into account sortive democracy aggregates preferences far better than the elective variant. For example citizens who are socially conservative but fiscally liberal (Catholics etc) have serious problems in two-party electoral systems as “progressive” parties are generally liberal in both domains. Such people would be much better served by descriptive representation as their proxies would vote liberally on fiscal issues and conservatively on social policy issues, thus better reflecting their preferences.


  23. […] the face of such difficulties, Stone and his collaborator Oliver Dowlen propose (1, 2) to abandon the representativity justification for sortition. Instead they offer a different […]


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: