Achieving Accurate Representation

From "The Trouble With Elections: Everything We Thought We Knew About Democracy is Wrong," Chapter 10.4

citizens’ assembly invitation issued following first round lottery

Since quasi-mandatory service (like jury duty) is not a legal option at present, how is representativeness and stratified sampling currently being done for citizens’ assemblies in the real world? The standard approach has been to make an initial invitation by mail, phone or door knocking to a randomly selected list of people (or phone numbers, or addresses). In some instances (depending on the budget) follow-up contacts are made to encourage those initially contacted to step up, answering their questions and offering remedies to any impediments to serving. The random selection lottery process should be impartial and transparent, so that anyone who wishes can verify that it was done honestly. 

From a pool of those who responded that they would be willing to serve, a second lottery is conducted, but using stratified sampling. Those who agree to serve provide various demographic and other details about themselves to allow the stratified sampling process to match the diversity of the target population (a municipality, a nation, or in some cases the entire world). Quotas are set according to the demographics of the population according to age, education, sex, income, geographic location, etc. to make the final body as descriptively representative of the population as possible. It is even common to use polling data on the topic of the assembly to assure the group's attitude matches the population as well. For example, if the assembly is going to be examining a response to climate change, respondents are asked whether they think climate change is a big concern, somewhat of a concern, not much of a concern, or not a concern at all, so that the panel aligns with the population’s attitude on the topic, and is not full of deeply concerned activists, for example. Various procedures have been used to make the final lottery selection, often using a computer to generate a large number of random groupings with different potential participants, with each grouping satisfying the stratified demographic quotas. The final public lottery (perhaps using lottery balls, multiple dice, or other visible random process) picks one of these possible mixes of willing participants as the final body.

Whether stratified sampling adequately compensates for decliners depends on whether those who decline have consistently different interests, attitudes and policy preferences than those within the same demographic groups who do agree to participate. In short, would the outcome of a deliberative process be substantially different if decliners could be induced, or legally compelled, to participate (as with jury duty), as compared to having substitutes from the same demographic categories? Clearly, one requirement is the need to eliminate every possible obstacle to service, such as preventing any loss of income, providing needed childcare, etc. Inducements need to be considered as well, such as generous payment or the honor of helping one’s community, etc.

So, what are some reasons decliners decline? The research by Vincent Jacquet, interviewing decliners from real world citizens’ assemblies, is a starting point for helping to assess whether stratification can work to fill in where decliners leave holes. Jacquet found that responses¹ could be lumped into six kinds of reasons. The more common reasons for declining were: scheduling conflicts;  preference for spending time in the private sphere, especially with family and at work; and low expectations of the assembly's impact on the political system. The less common reasons were: feelings of low personal  political competence; dislike of group situations; and alienation or distrust of politics generally. 

A mature functioning sortition democracy in which participation generated clear public impact, would mostly eliminate concerns related to lack of impact, and to some extent alienation from the political. With paid time off from jobs, child care, etc., issues related to lack of time could also be ameliorated. Certain functions of a sortition democracy could easily be conducted with remote virtual participation to alleviate discomfort with group settings.  

However, a fundamental factor the research didn’t tackle was whether pay and other inducements would change decliners into participants. None of the three assemblies studied offered remuneration for participation (though many others have). This research also couldn’t assess the crucial question of whether those who declined had significant similarities that would leave a certain segment of society out of the sortitional decision-making process, or if on the contrary, non-participants were essentially also random, such that their lack of participation was statistically irrelevant.

But what if the non-participation is an intentional preference, rather than due to obstacles or social inequalities? A huge question is whether it is more oppressive to have quasi-mandatory service, or accept that some kinds of people intentionally want to go unrepresented. When I was a politician knocking on doors campaigning for re-election, I recall a resident who was a member of the Jehovah’s Witness faith. Members of this denomination decline to vote or participate in any political process at all as too worldly. The irony, of course, was having the shoe on the other foot – with polite disinterest in what each of us is “selling” to the other, when politicians and Jehovah’s Witnesses proselytizers knock on each other's doors. I wouldn’t join their religion, and they wouldn’t join my politics. How should democracy handle representation when some people simply don’t want to participate in any form of  democracy?

The accuracy of representation is also deeply impacted by sample size, as well as serving vs. declining. Large samples increase the probability of accurate representativeness, but have the downsides of increased rational ignorance (as each member’s input becomes less significant) and reduced ease of deliberation. The optimal size for a lawmaking body has long been debated. In Federalist #10 Madison noted that

“however small the republic may be, the representatives must be raised to a certain number, in order to guard against the cabals of a few; and that, however large it may be, they must be limited to a certain number, in order to guard against the confusion of a multitude.”

Active deliberation is enhanced by smaller size, such that each member can communicate with all others.² Most citizens’ assemblies have tackled this by including chunks of time using random, rotating break-out groups of ten or fewer members at multiple tables, to allow face-to-face deliberation. It is my strong conviction that there is no such thing as an all-purpose optimal size for sortition bodies. It depends on the task. A mini-public that is crafting (but not authorized to adopt) a proposal might benefit from smaller size to enhance active deliberation, while a final decision-making jury that is only going to hear pro and con arguments, and ask clarifying questions before voting, could be substantially larger, to enhance the accuracy of its representativeness. This is a focus of my multi-body sortition concept I have written about elsewhere.

A key feature of my multi-body design is the constant effort to improve the workings of sortition. A sortition democracy should provide for a mini-public charged exclusively to that task, with no public policy distractions. This brings up another feature of sortitional democracy – its inherent self-correcting nature. By constantly bringing in new participants, and avoiding power-hoarding by entrenched elites, a bad decision by one mini-public can be readily fixed by a subsequent mini-public. This is especially true since the new members have no pride of authorship or vested interest in defending prior poor decisions (unlike elected elites). This self-correcting aspect also provides latitude on the issue of sample size needed for accurate representation, since repeated sortitional bodies will tend to regress to the mean and zero in on optimal decisions. It is only when some irreversible decision is to be made that larger samples, or even supermajority voting thresholds, become vitally important.

While it is beyond the scope of this book, there is another issue of participation that is vitally important. The core principle underlying democracy is that the people should control decisions that affect them (rather than allowing some elite to make those decisions). But what about decisions that will dramatically affect future generations, rather than those currently alive – or even other species? While it is impossible to directly include actual members of these groups in decision-making, researchers in the Japanese Future Design movement, have argued that such a mindset can be encouraged within participants, or that such interests should have designated representatives within the deliberative system who will specifically look out for them. Experiments with asking deliberators to take the role of an imaginary future generation advising the current generation have shown promise. Researchers have coined the word “futurability” and suggested that

“one must design social systems that activate a human trait called futurability, where people experience an increase in happiness because of deciding and acting toward foregoing current benefits to enrich future generations.”

The climate change crisis, and rapid advances in AI have brought this issue front and center. It is important that ways of expanding virtual participation beyond the current generation of people be put on the agenda, so that a good democratic deliberative process can decide how these unique interests can be incorporated into future decision-making.

1. It is important to remember that self-reported “reasons,” may not always reflect reality. Those interviewed could feel some need to give more respectable reasons than were actually underlying their refusal. It is also often true that people don’t even know why they make some particular decision, and their brain generates a plausible story that they tell themselves and others.

2. Some people reference Dunbar’s Number, suggesting that around 150 participants is the upper limit for good deliberation, while others propose much smaller or larger bodies.