How can we move past the debate between representative and participatory democracy? The study of the social organization of animals helped researchers to design algorithms that provide an original solution to this unsolvable issue.
In an ideal world, citizens should be able to express their preferences on any subject that affects them. Hence the rise of referendums, as a response to the flaws of representative democracy. And yet, even on a single issue, a referendum is far from an ideal solution. The fundamental question, once raised by Kenneth Arrow, is whether we can find at least one democratic method for aggregating individual preferences which a priori respects minimum democratic and rational conditions. The study of the social organization of animals may help us find new ways to solve this unsolvable issue. It has inspired new algorithms, especially for robotics. Even if the political or economic issues are highly multidimensional, it seems promising to be inspired by these innovations.
How can we move past the debate between representative and participatory democracy? In the search for a better form of democracy, in which citizens would be more involved and their preferences better taken into account, participative democracy seems to offer a viable solution: the annual number of referendums in Europe has tripled since the 70s.
Indeed, in an ideal world, citizens should be able to express their preferences on any subject that affects them. To better understand the differences between a referendum and a representative system, let us borrow the following example to Laslier: suppose that we vote on three societal themes, and that each theme divides public opinion into two distinct positions (e.g. for or against the construction of a new highway). Suppose also that the vote is shared between just two political parties, which we will call R and L. Of course, R and L have preferences regarding each theme. Now imagine that public opinion is divided as follows:
In this table, the first group of 20% of voters, for example, prefers the choices of party L on themes 1 and 2, and the choice of party R on theme 3.
We can see that L is clearly in a minority position on each of these themes (in the three cases, it represents only 40% of the voters), so it would lose all three referendums on these themes. On the other hand, in a representative system where the vote concerns a program (which can be seen here as an aggregation of choices concerning each theme), L is in the majority against R, since for each of the segments of voters, its positions are preferred in two out of three cases! This is called the Ostrogorski paradox.
The possibility of electing a minority political party on each theme of public policy seems to logically work in favor of the single-question referendum. This type of referendum is frequently practiced in the United States at the local level, in both states and municipalities. Conversely, the representative system presents certain pathologies: at the US federal level, we observe the proliferation of “catch-all” laws drawn up precisely when the underlying provisions do not have majority support. The 2017 presidential election in France offers an example of Ostrogorski’s paradox, in that the winning candidate did not have an implicit majority regarding many of the major themes of the campaign, based on the results of the first round.
And yet a referendum, even on a single issue, is far from an ideal solution. Can we guarantee that popular consultations will not lead to an inconsistency in legal standards over time? The results of referendums in California may have led to new spending and, at the same time, to a lowering of taxes. Are referendum initiatives outside the scope of the government likely to paralyze government action? On the theoretical level, at least, McKelvey and Wendell have shown that the repeated application of the majority rule to a succession of simple questions could, after sufficient iterations, produce a decision that contradicts a previous decision! The majority rule is then described as “chaotic.”
The problem is no longer just about examining the superiority of one rule of political decision-making over another. The fundamental question is whether we can find at least one democratic method for aggregating individual preferences which a priori respects minimum democratic and rational conditions. This is the problem that the famous Arrow impossibility theorem tries to solve, with the following conditions:
1. Condition of effectiveness: if all individuals prefer A to B, the same must be true of the community
2. Condition of independence: the choice between A and B depends only on the individual opinions about A and B
3. Condition of universality: a collective choice must be possible in all configurations of individual choice.
Surprisingly and disappointingly, the only rule of collective choice that meets these three conditions is “dictatorial” rule, whereby a single individual imposes their choices on the community.
The aggregation of collective preferences is therefore fundamentally problematic. There is an additional dimension to the problem posed by Arrow’s theorem: its inter-temporal dimension and the necessary coercive power of what Simon Nora called the “priests of the long term” which allows decisions to take effect. Indeed, how effective would a system be if it allowed permanent debate and voting about public policies? The control of time and the political agenda must therefore be orchestrated over time. But by whom?
Rather than seeking the opinion of citizens on all occasions, is the real concern not to enrich the government’s work via new expertise, not for the benefit of lobbies but in the public interest? Our institutions are the legacy of a time when it was enough to recruit brilliant senior officials to make decisions in centralized institutions. Thanks to modern technologies, including the opening of public data and crowdsourcing, collective decision-making can be reinvented in a co-construction of public policies by the government and the citizen, that is to say in political decision-making informed by the community.
But would such a system not call into question one of the foundations of modern democracy, namely the identical weighting of individual choices (one citizen, one vote)? If one wishes to improve the quality of political decision-making by incorporating the work of collective intelligence, it seems crucial to carry out a critical review of the diverse expertise or of the simple opinions expressed, to filter them, to attribute to them a differentiated weight, and in particular to exclude those which would be in contradiction with the facts or violate the most basic logic.
This procedure is only the exacerbation of an old phenomenon: the opposition between the majority voting process and the administrative process. By nature, the latter conducts a selection or, at best, a synthesis, but does not take into account the opinions of all voters.
If this rule is of a discretionary nature, we are brought back to enlightened despotism. To avoid any risk of human arbitrariness, it may be tempting to rely on machines. However, advances in artificial intelligence will not necessarily lead to greater transparency. In the extreme case, replacing the government by a “machine learning” type process could lead to the least bad solution, without being able to explain it to the citizen, whose adherence is nevertheless crucial.
If the procedure of discrimination of collective intelligence is of an economic nature, that is, following the pattern of Census suffrage such as was the rule in most of the 19th Century, there is a real danger of losing social cohesion. Muro and Liu, in an article published by the Brookings Institution, have shown that if the vote in the US presidential election had been weighted by the Gross Domestic Product of the US constituencies, Hillary Clinton would have won the election with 64% of the vote (and Al Gore in 2000 with 54%). But then, the latent frustration that led to the election of Donald Trump would certainly have expressed itself outside the regular democratic framework, i.e. in the street.
More generally, one could imagine a new social contract whereby citizens accept that their wishes are aggregated in a process of optimizing public decisions, where, unlike in traditional majority rule, constraints apply on the choices available to individuals and disrupt the weighting of their voice.
How can this idea be implemented? One of the most fruitful and misunderstood political principles is that of subsidiarity, i.e. the decentralization of political decision to bring it as close as possible to the individuals affected by it. In this respect, it seems appropriate to look at local, informal, simple and fast decision-making systems. Here, as elsewhere, multidisciplinarity can be particularly fruitful: the study of the social organization of animals has inspired new algorithms, especially for robotics. Even if the political or economic issues are highly multidimensional, it seems promising to be inspired by these innovations.
Coordination in a flock of birds is done through a small number of individuals without particular privileges that guide the entire group towards a collective consensus on topics such as the direction of a migration.
In a human society that can be compared to a highly decentralized set of units with multiple and different interests, political, hierarchical and vertical coordination is often slow and erratic. In contrast, coordination in a group of animals (a flock of birds for example) is done through a small number of individuals without particular privileges that guide the entire group towards a collective consensus on topics such as the direction of a migration.
These animal behaviors inspired Nagpal and other researchers from Harvard University to propose an algorithm called “implicit leadership” that allows all the agents of the system to agree on a decision motivated by some better informed individuals, based only on local interactions and not on global optimization. The model even allows the objectives of the “informants” to be divergent, which would of course be more than likely in a political framework. For the authors, this is an effective decentralized decision-making approach with information asymmetry, without the explicit emergence of a “leader” and without elaborate communication: each individual modifies its behavior or position according to those of its neighbors.
Could such a system ever replace the simple citizen vote for a candidate or on a referendum subject? We do not claim, at least in the current state of science, that there is a clear convergence of models of political and economic governance towards animal protocols or robotic algorithms. We simply believe that there are many new territories to explore, in a multidisciplinary spirit.