Sortition Design for the Future

Dilemmas of Sortition Design

We should carefully consider human factors and design elements that contribute to, or hinder, good legislative decision-making, especially for randomly selected legislative bodies. We should not only be concerned with the legitimacy or fairness of decision-making, but also the quality of the outputs. Landemore makes an interesting argument in her book, Democratic Reason: Politics, Collective Intelligence, and the Rule of the Many, that democratic group decisions have a greater probability of being “good” than the decisions of smaller oligarchic groups, even if both groups have the best of intentions.

However, democracy does not assure justice or “good” decisions. It may open the door to such decisions with optimal institutional design, but all it can promise is freedom from oppression by a powerful minority. An informed majority can still make “mistakes”. A majority, or even a unanimous decision that causes the death of millions, might arise through a democratic process. The citizens of Athens democratically decided to launch wars of aggression. The use of the democratic process does not change the moral or ethical predisposition of the participants. Also, an intolerant, illiberal culture, which places inordinate weight on the opinions of certain individuals (whether political, religious, or otherwise), may not be able to benefit from the advantages of democratic design.

Setting aside definitions of morally or ethically bad decisions, we can focus on avoiding “bad” decisions that the makers of the decision themselves would agree were bad if they had had more information, understanding, or through after-the-fact hindsight. Being minimally “democratic” is desirable, but not sufficient for avoiding such bad decisions.

Rather than focusing on design elements that are hoped to generate good decisions, many political theorists have deemed it more practical to focus on design elements intended to avoid bad decision-making. Madison’s hope that a vast republic with institutional checks and balances could prevent any one faction from taking control to pursue their narrow self-interest was essentially a defensive concept (though the quick rise of national political parties that spanned the three branches of government effectively stymied his hope). Devising ways of preventing bad decision-making more than assuring good decision-making was also the approach of other theorists, such as Jeremy Bentham. The goal can be summarized by the title of one of Bentham’s nineteenth-century works and that of modern theorist Jon Elster of Columbia University—Securities Against Misrule. How can we minimize the problematic proclivities of human psychology (such as confirmation bias, groupthink, and self-aggrandizement) and their tendency to poison good decision-making? What institutional design features and procedures minimize the likelihood of “bad” decisions, and maximize the likelihood of “good” ones? Cognitive biases are notoriously difficult to avoid, but recent research suggests that certain procedures and bias reduction training may indeed help (Sellier, Morewedge, and Scopelliti 2024). Such procedures and training are in fundamental conflict with politicians’ power-seeking priorities, but could readily be deployed within mini-publics. We should examine ancient insights as well as modern discoveries of social psychology in order to avoid the pitfalls that have crippled so many democratic efforts over the centuries. What follows is a substantial, though not exhaustive, list of factors that need consideration. We will quickly discover that there are dilemmas where solving one design problem will make another problem worse. In short, no single design can be optimal across the board. A single body will necessitate inevitable trade-offs. However, as I summarized at the end of Chapter 11 and described in greater detail below, there is a design that can resolve these dilemmas and avoid trade-offs. I refer to it as the multi-body sortition model.

Five Dilemmas of Sortition Design

When people first imagine a sortition democracy, it is common to imagine a one-for-one replacement of an elected chamber, with one selected by lot. This would be a big mistake. All single-body sortition proposals face five dilemmas— five pairs of opposing objectives—which can’t be reconciled with only one type of body.

1. There is a conflict between maximizing accurate and descriptive representa-

tiveness, versus maximizing interest and commitment among members of a deliberative body. In Deliberative Democracy in America: A Proposal for a Popular Branch of Government, Ethan Leib seeks to maximize descriptive representativeness and avoid the bias of “participatory distortion” by

insisting on mandatory service as in a jury or draft (Leib 2004). Some other theorists and advocates put more priority on assuring interest and motivation. Their designs tend more toward volunteerism, or a lottery of the willing. Accurate representativeness also requires a large sample—a group that is then too large to engage in active deliberation effectively.

2. There is a conflict between spreading participation widely, maximizing resist-

ance to corruption on the one hand, through a large number of short terms of office, versus maximizing participants’ expertise or familiarity with the issues under consideration through far fewer, longer terms. Longer terms increase expertise, but also increase the risk of corruption. Further, certain sorts of people would be willing to serve long terms, but others would consider it an unacceptable imposition and would be unwilling to “do the work”.

3. There is a conflict between giving every citizen the right to speak (encourag-

ing self-selection)—offering agenda items, information and arguments for the deliberative process, versus the danger that the self-selection of those most motivated to speak will promote domination by special interests and steer outcomes away from the common good. Those who seek to speak are not usually typical or representative of the general population.

4. There is a conflict between wanting a diverse body that engages in

problem-solving through active deliberation, fostering collective intelligence, versus independent personal assessment and private deliberation that taps the wisdom of crowds and avoids information cascades, but which shuts out the sharing of private knowledge. There is persuasive research showing the value of cognitive diversity for problem-solving, but also the value of independent, private assessment of information (Page 2007; Landemore 2012a; Lorenz et al. 2011; Surowiecki 2005). Group deliberations can also suffer from deference to high-status members or group solidarity, leading either to groupthink or polarization (Sunstein 2003).

5. Lastly, there is a conflict between maximizing democratic power by allowing

a deliberative body to set its own agenda, draft its own bills, and vote on them, versus avoiding the bundling of issues, with the resulting vote- swapping, as well as arbitrary decisions arising from the persuasive powers of a few unrepresentative charismatic members (Sutherland 2008). These five dilemmas are not exhaustive, but are crucial.

Before diving into the solutions to these dilemmas, I am proposing below, I want to touch on a couple of other design considerations.

Motivation to Participate

Any proposed system that requires an increase in the amount of citizen participation, as this one does, must respond to the question of whether there would be sufficient motivation among broad swaths of the population to participate. After all, only a minority of citizens is willing to vote in most American elections, and that requires relatively minimal time and effort. New England communities that still have town meetings also see only a fraction of their citizens attending. The competing uses for personal time in modern society, and the “unattractiveness” of politics to most people, raise serious questions about the viability of any such democratic undertaking, especially one centered on deliberation and broad participation (Warren 1996). Even in Classical Athens, democracy was an activity only of that (quite large) portion of the citizenry who chose to participate by volunteering for the lottery pool, or attending the ecclesia. The goal with this plan is to go one better than Athens and include the general population, rather than merely those who are eager to participate.

The (testable) assumption I make here is that most citizens would readily participate for a set period of time, with appropriate compensation, in a process in which they believed their input really mattered (unlike mass elections). This democratic process would bear almost no relationship to “politics” as we know it today. While there are high levels of satisfaction reported by participants in various deliberative processes, such as the British Columbia Citizens’ Assembly, or the Danish Consensus Conferences (Fischer 2009), this may be misleading, since participants were randomly selected from among those who already responded that they were interested in being one of those “selected”. Satisfaction with serving on a court jury is more indicative and suggests that sortition for policy making does have the potential to overcome the “rational ignorance” problem of mass elections. Just as jurors in court systems may complain about the nuisance of serving, they almost universally take the job seriously. Indeed, many jurors go away with a heightened sense of citizenship (Matthews, Hancock, and Briggs 2004). The multi-body system described below also seeks to accommodate varying levels of willingness to commit personal time to self-governance. The largest portion of participants would commit a very limited time—say, no more than a week (serving on a municipal-level Policy Jury). Those serving on a national-level Review Panel might serve a year or more.

Ongoing Design Revision

In a mature sortitional democracy, the group or organization that designs deliberative processes should also be democratically controlled. Otherwise, the designers may corruptly create a system that tilts policy outcomes in their preferred direction. Ideally, this meta-legislative design task would be undertaken, or at least overseen, by a separate mini-public created specifically for this function, rather than as an initial step of a policy deliberation. Isolated from any particular legislation, people who would be on opposite sides of a debate on a “policy issue” can agree on what procedures, rules and expert witness selection process should be used by future mini-publics, which they won’t be a part of. A randomly selected meta-legislative body could consult experts and use the latest research on group decision-making and cognitive biases to engineer an optimal decision-making process, without the electoral imperative that forecloses this possibility for an elected chamber. Therefore, the design I presented at the end of Chapter 11 and further explicated below in relation to the dilemmas discussed above, is only a starting point, which I expect to be improved over time.

Solving Dilemmas with Multi-Body Sortition

This proposal uses a variety of bodies, each with unique characteristics (such as selection method and term of office) that are optimized for the task each body handles, and to overcome the dilemmas discussed above. The proposal below is comprehensive—a sort of “reference design” such as that used by architects. In real-world applications, it is likely that certain bodies described might be used, and others not. For example, a city government might utilize the Rules Council, Review Panel and Policy Juries within one specific policy area, but keep the Interest Panel, Agenda Council, and Oversight Council functions within the existing city council.

Core Bodies

Agenda Council

An Agenda Council will be created periodically (or as a permanent body, with rotating members) to set the agenda for topics, policies, or laws deemed important to address in the coming period of time. This body might consist of 150 members. A serious problem with the current system of politicians setting the agenda is the priority of mobilizing voters by selecting hot-button issues that make it easy to vilify other politicians. Politicians also have a motivation to elevate issues that big-spending lobbyists want on the agenda, regardless of public benefit. It is also electorally inconvenient to raise issues that are highly complex, or don’t have easy sound-bite policy options, unless the reality on the ground or crisis has forced this agenda item forward. A random sample of citizens has an interest and motivation to hear from a variety of opposing experts, and take a long-term view (rather than prioritize the next election cycle), and so is likely to generate an agenda more beneficial for society. The Agenda Council would issue a call for proposals on each of the agenda items selected.

It might be appropriate to also have an alternate petition route to add topics to the ensuing period’s agenda. Of course, those petitioning couldn’t determine the direction of policy, since the Review Panels and Policy Juries discussed next may conclude the best public policy is the opposite of what the petitioners hoped for. Since petitioners are by definition self-selected, they raise topics, but do not determine outcomes.

Interest Panels

Good ideas for specific policies are spread throughout society, and this is a compelling reason to open the gates wide for draft proposals on the agenda topics. Self-selection at the level of the Interest Panel allows experts who would be unelectable (due to their appearance, lack of wealth, personality, or other traits) to contribute to governance. It also means that special interests and self-deluded incompetents could participate. While a panel of a dozen would be likely to identify and discount the ideas of “wackos”, it is likely that some Interest Panels would put forth terrible legislative proposals. This is one reason why it is desirable to have a large number of Interest Panels, and why self-selected Interest Panels should not make ultimate decisions. This is real democratic participation, as opposed to partisan marketing, which is the focus of electoral participation. Some of them might deadlock or disintegrate due to quorum failures. In most cases, however, multiple Interest Panels would produce draft proposals for the next level allotted body—a Review Panel—to consider.

These Interest Panels could be structured in a variety of ways. Each panel would be small enough to facilitate active participation (perhaps a dozen members). A group of like-minded people could self-organize (perhaps with a common concern). Other panels could have randomly matched volunteers to create more diverse groups.

There would be as many panels as needed to allow anyone who wished to participate to do so. Interest Panels at the local level might meet in person on weekends or evenings, but, where Internet access is common throughout society, many would likely use online collaboration tools that allow members to communicate and work on proposals sequentially, rather than needing to coordinate meeting times. These panels would be generating raw material, rather than making any final policy decisions, so their lack of representativeness is not important. However, because they know that their work product will have to pass muster in front of subsequent representative bodies, they have an incentive to craft proposals that serve the common good, rather than just some special interest.

Review Panels

There would be a Separate Review Panel for each policy domain (for example, one for transportation, one for healthcare, etc.). The process of the Review Panel would be significantly different from that of an elected legislature.¹ The Review Panels fulfill the central function of traditional legislative committees, excluding agenda-setting and the initiation of legislation. They also have no role in the final passage of legislation. They might operate using a process similar to that of the British Columbia Citizens’ Assembly, with a learning phase, a deliberation phase, and a drafting phase. They would receive the raw inputs from the Interest Panels, compare, combine, reject and amend various draft proposals to prepare a final piece of legislation to pass on to a Policy Jury for final adoption or rejection.

A democratic lottery process similar to that used by citizens’ assemblies today, likely using a two-round lottery would be appropriate. The first round would find people willing to put in the required time and effort. However, they would not be told which policy domain they would be addressing, to avoid selecting participants with a biased special interest. This would be followed by a stratified random sampling to constitute a roughly representative body. Because their workload would be substantial, a significant portion of the population would probably decline to serve if called, so stratified sampling would be necessary to approximate representativeness. However, since they also would not be making a final yes/no decision on legislation, their rough representativeness and small size (150 is too small to have an adequately high probability of accurate representativeness) facilitate active give-and-take deliberation. These bodies would benefit from great diversity, while also being mostly free of special-interest self-selection bias and corruption. Their task is to prepare a final piece of legislation for a Policy Jury to consider.

Each Review Panel would also have professional staff. For reference, in the United States, the various congressional committees have a total of around 6,000 staff. However, the individual elected members of Congress each have personal staff for such things as constituent outreach, district office staff, schedulers, speech writers, and others, which total another 12,500 or so. While none of these are officially campaign-related, most of them exist to enhance re-election prospects. Nearly all of these personal staff would not exist in a sortition system.

For a state or national design, these panels would be full-time, with overlapping terms of perhaps three years (to gain familiarity with the subject matter). For a municipal implementation, this would likely be part-time work, and it might be appropriate to have meetings on weekends or evenings so as not to interfere with normal employment. This body would be far more descriptively representative than the Interest Panels.

The Review Panel would perform traditional legislative activities: holding hearings, inviting and listening to expert witnesses, utilizing professional staff for research and drafting, and amending or combining elements from the proposals submitted by Interest Panels to produce a final bill. The procedures used by the Review Panel require careful design so as to maximize problem-solving potential and minimize both groupthink and internal polarization (Sunstein 2003).

These psychological tendencies are powerful, but can be minimized with good design (Manin 2005; Fishkin 2009). However, since the risk of bad results from information cascades and groupthink, etc. cannot be completely eliminated, and the fact that the Review Panels are not fully representative of the population, these panels would not make final decisions about the adoption of the legislation they crafted.

Policy Juries

A key feature of this design is that we don’t leave the final decisions on legislation to Interest Panels nor to Review Panels. Because of the risk of groupthink, pride of authorship, or simply inaccurate representativeness (due to smaller sample size and self-selection bias), final decisions are made by separate bodies, which I call Policy Juries. This separation of functions would also reduce the likelihood of extreme polarized majority positions coming out of the Review Panel, since members of both the Interest Panels and Review Panels would understand that their final products must be able to pass muster in front of a large, fully representative Policy Jury. Finding middle ground, or win–win options, and addressing the needs of minority perspectives might enhance their chances of ultimate adoption.

Each Policy Jury would vote on one piece of legislation, as the Legislative Panels (nomothetai) of Athens did. To avoid self-selection distortions, jury service should ideally be quasi-mandatory (though with reasonable hardship deferrals allowed). It is unlikely that early implementations of this multi-body design would have legal authority to mandate service, so substantial alternative means of encouraging participation would need to be employed. People have few, if any, comparable opportunities for civic engagement in politician-centered systems, so we can only speculate about potential participation rates. Short service durations (perhaps a week or two) combined with compensation and some symbolic status honors might achieve adequate participation, and thus descriptive representativeness. A state or national Policy Jury should probably have between 500 and 1,500 members to achieve a representative sample. A municipal Policy Jury would probably be significantly smaller, simply for financial reasons, and the fact that the issues would generally be significantly less consequential.

The essential task and the procedures of the Policy Jury are intrinsically different from those of the Interest Panels and Review Panels. Like the Athenian Legislative Panels (nomothetai) and People’s Courts (dikasteria), a Policy Jury would listen to pro and con presentations on the proposed legislation, and without further debate, vote to adopt or reject the law by secret ballot. This procedure is intended to benefit from the wisdom of crowds described in James Surowiecki’s book while avoiding the groupthink and polarization dynamics that can arise when participants engage in discussion and advocacy (Surowiecki 2003).

This procedure elevates the skill of democratic listening—an often neglected skill that is distinct and apart from debate. Some researchers and political theorists note that it is unfortunately far more common to focus on the value of speaking, or “voice”, than listening when examining democratic behaviors (Dobson 2012). When debating (or even conversing), as we hear another speaker, our mind often gets preoccupied with making a plan for our response or counter-arguments, not openly absorbing or seeking to understand others’ perspectives. Debate is about winning—not learning. Narrowing the jurors’ task to listening to presentations and voting (with the uncommon inducement of exercising actual decision-making authority) will also increase willingness to participate, especially among shy and socially awkward members of society, and thus the jury’s descriptive representativeness. Because give-and-take deliberation among members is not involved, this entire stage might be conducted remotely using the internet.²

The secret ballot is important, as it helps avoid the social pressures and deference to high-status members of society that can interfere with a juror’s ability to make their own independent assessment and vote as they think best. The secret vote also reduces the risk of corruption and vote buying. Jury tampering protections (including announced sting operations of pretend bribers) would still be appropriate, but potential corrupters are less tempted to try buying votes when there is no way of knowing if the votes were delivered.

Preparation of the pro-and-con presentations could be overseen by the members of the Review Panel who had been on opposite sides of the decision within that body. A Rules Council might be wise to also allow jurors to ask clarifying questions, and perhaps even seek out independent expert witnesses. The Policy Jury is the failsafe check that assures new laws would be deemed appropriate and desirable by an informed population. If a bill passes by a very small majority, it might be wise to automatically convene a second (perhaps larger) Policy Jury and require two passing votes to adopt the law.

Coordination Council

Because Review Panels would be siloed by policy domain, yet some issues would inevitably have implications outside their core domain (especially in terms of budget), a Coordination Council, similar to the Agenda Council, would propose modifications to harmonize passed policies. To avoid a concentration of power, the proposals of the Coordination Council would also need to pass muster in front of a Policy Jury.

Meta-Legislative Bodies

Rules Council

A lottery-selected body (with new members regularly rotated in) should be constantly monitoring how the system is functioning, to propose improvements. Changes might have to do with the lottery process, the size or duration of bodies, the procedure for selecting expert witnesses, etc. These changes would also go to a Policy Jury to ensure proposed changes would not inappropriately increase the power of the Rules Council itself.

The reason that rules should be promulgated by a distinct sortition body is both practical and fundamental. It would be impractical to have each (often short-duration) body craft its own rules. Factors influencing procedures and rules can be as complicated and time-consuming as many bills themselves. But the fundamental reason for separation is somewhat similar to the thought experiment proposed by John Rawls in his book A Theory of Justice (Rawls 1999). He imagined an Original Position concept where justice is sought behind a “veil of ignorance”, where the interlocutors do not yet know what position they will have in society (whether male or female, rich or poor, blind, or sighted, etc.). In this scenario, they would rationally seek a fair society because of this ignorance of how it would affect them. A sortition body that won’t itself be adopting laws is best suited to decide on rules for other sortition bodies, as they avoid corrupt motivations. Within any group, alliances and factions can develop.

Once a majority faction becomes self-aware (as in every elected legislature), they have an incentive to make their work easier by, for example, limiting the speech of their “opponents”, and adopting all sorts of other corrupt rule changes. Yet, these exact same people, if they were deciding on rules for some other sortition body, would agree on the importance of allowing the minorities to contribute their ideas to deliberation. It is the having of power to adopt laws that can corrupt—not merely whether the representatives are selected by election or lottery.

Oversight Council

While the Rules Council decides the rules and procedures, monitoring a large number of mini-publics would overtax this body. There would need to be at least one, but more likely, a large number of Oversight Councils, responsible for enforcing the rules. Take, for example, the question of who presents the pro and con arguments to the Policy Juries, and how they decide exactly what content to include, in order to give a “balanced” presentation? Even the relative charisma, eloquence, appearance or social status of different presenters may be significant. One possible approach the Rules Council might adopt is to have the same staff members present both pro and con arguments crafted by members of the Interest and Review Panels who fall on either side of the pro-con divide. However, this leaves the Policy Jury in danger of being steered by the administrative bureaucracy promoting its own interests—a common problem with elected bodies. Many state and national legislatures try to solve this problem by allowing each legislator to hire their own staff. However, this individualized staff often ends up spending an inordinate amount of time on re-election concerns, such as public relations and constituent service, rather than policy.

Instead, for a sortition system, I propose the creation of Oversight Councils, chosen by lot, which deal exclusively with staff performance and fairness, rather than the policy issues themselves. In addition to evaluating the general competence of the staff, they would rule on complaints about biased or unfair presentations given by staff. They should also have the power to hire and fire staff serving the other sortition bodies.

At a state or national level, it might be appropriate to have several separate meta-legislative councils. I have described a Rules Council establishing rules and procedures for a good deliberative process, and an Oversight Council overseeing the staff of all of the sortition bodies. A full sortition system might also have one that monitors the performance of the executive in implementing the laws adopted by the Policy Juries. For a municipality, a single meta-legislative council seems appropriate, with the limited task of assuring that sortition-formed bodies have a good process and staff that provide balanced presentations— independent of any existing government or bureaucracy.

Keys to Good Design

At a bare minimum, democracy requires that agenda-setting and the final decisions are in the hands of genuinely representative bodies. However, replacing elected legislators with randomly selected representatives also opens the door for innovative and superior means of proposing and evaluating draft legislation. In traditional elected legislatures, the same set of elite elected officials is empowered to establish the agenda, prepare specific bills and amendments. They are also responsible for engaging in (or sidestepping) fact-finding, advocacy for or against bills, and finally judging legislation to adopt or reject it. These tasks should be divided among a variety of other groups of people. These groups should generally be randomly selected (or possibly self-selected in the case of Interest Panels) to optimize diversity and representativeness, while avoiding self-selection distortion and corruption. Interestingly, both the randomly selected Australian Citizens’ Parliament and the Belgian G-1000, two of the earliest sortition implementations, provided an additional means for interested individuals who were not part of the randomly selected body to participate in a manner online, by providing proposals or assessments and “raw material” for the formal deliberators. This process of providing a come-one-come-all opportunity for interested citizens to feed into the decision-making process, but countering the resulting self-selection bias by then submitting proposals to the more descriptively representative body, is a model worthy of emulation.

While it seems inevitable that the use of sortition in the near term will be part of a hybrid system that also includes politicians, it is important to understand that maintaining an electoral component is not essential, nor desirable, for democracy, and that it would be possible to have a democracy with no mass elections at all. In this chapter, I set forth a multi-body design of democracy that does not use elections. Societies have been on the beaten path, relying on elections for so long that such a sharp break in how we think about formulating a democracy is extremely challenging. I will examine some transition strategies in the next (final) chapter. But the design I set out here is intended to help people appreciate that elections should not be the default tool when setting up a democracy. We need to understand that a better way is possible before people will willingly abandon the familiar ways.