Random Walks Across The Aisle: Introducing Certainty Through Randomness in Electoral Systems
The world sometimes feels chaotic & random. What if adding more randomness made it less so?
Today, electoral systems can be costly, corruptible, underrepresented, and prone to manipulation by moneyed interests.
Various proposals have been put forth that removes some of these issues by focusing on introducing randomness. To quote David Chaum & his proposals on random sample voting, it can result in:
“…(a) high effective voter turnout, (b) better informed voters rationally motivated to delve into issues, (c) increased effectiveness of results in shaping governance, (d) improved resistance to manipulation through advertising/campaigning, (e) increased indisputability and trustworthiness of results, (f) anonymity of voters with unsaleable votes, (g) reduced opportunity for selective denial of voter access, (h) voter fraud only through improper voter rolls, (i) equivalent but safer decisiveness, and all with (j) significantly reduced direct and overall cost.”
Let’s roll some dice and dig in.
Lottery Voting
Randomness in our electoral systems is not new. Athenians used a system called sortition, whereupon public officials were drawn by lot from a group of self-selected citizens (without consideration for specificity & skills required). For a thousand years until 1797, the Doge of Venice was elected through a complicated series of lots and elections. Historically, the arguments were that it provided a more democratic system.
The great Aristotle said:
“Democracy arose from the idea that those who are equal in any respect are equal absolutely. All are alike free, therefore they claim that all are free absolutely… The next is when the democrats, on the grounds that they are all equal, claim equal participation in everything.It is accepted as democratic when public offices are allocated by lot; and as oligarchic when they are filled by election.”
Such wow. The other Doge of Venice.
In modern electoral systems, however, Akhil Reed Amar, fed up with tyranny of the majority & political gerrymandering proposed a proportional representation system using random ballots. The goal, being, to increase the representation of minorities in legislatures.
In Amar’s proposal, when electing legislative bodies, each seat is randomly allocated from ALL individual votes. Instead of first-past-the-post elections in single-member districts, multi-member districts or party lists, you simply vote directly for any representative in the legislature. For each seat, a random voter’s ballot is chosen from all ballots and their choice occupies it. Thus, due to randomess: the more a person is voted for, the more likely they are to be represented. What’s particularly powerful about this is that it not only increases representation, but decreases the negative effects found in first-past-the-post voting: namely, reducing self-interest & increasing diversity.
From Amar:
“Lottery voting would thus gradually integrate growing outgroups” into government and would avoid the problems of political “tipping” that now occur when a former minority reaches fifty-one percent of the total district population. “Moreover, because a solid majority in a given locality could not guarantee itself a perpetual monopoly on the district seat in the state legislature,”’ it would have an incentive to treat the minority more kindly. The uncertainty of outcome that a lottery introduces is a check against injustice and self-interest, promoting more even-handed laws.”
David MacIver is a huge fan of random ballots. There’s a lot of questions around lottery voting and he gives quite great responses to various critiques: whether it is easy to game, whether people would understand it & how it affects elected officials. It’s a great summary & run-down on the benefits of lottery voting.
Once you have lottery voting, it could easily be extended to random sample voting. Due to statistical extrapolations, you don’t have to ask everyone to vote.
Random Sample Voting
To get a sense of scale from using random sample voting (proposed by David Chaum): For a population of 100,000 people, if you want a margin of error of 1% & 99% confidence, you only need to ask a sample of ~14,228 people to get, effectively, the same result. This grows to:
~16,317 for 1 million people.
~16,560 for 10 million people.
~16,585 for 100 million people.
~16,587 for 900 million people: the ballpark size of the voter roll for the largest democracy in the world, India.
The cost & scale is enormous: India votes for a whole month.
With random sample voting, a society can not only more regularly poll its citizens & decrease cost drastically, but also reduce the impact of electioneering & manipulation. With random sample voting it wouldn’t be known where the people would come from, and thus directed advertising & electioneering will be much more costly. In order to have the desired effect, it will be more costly to market, since they would have to effectively cover all their bases.
Amongst these benefits, the paper also claims that random sample voting adds a host of other benefits to society:
“…(a) high effective voter turnout, (b) better informed voters rationally motivated to delve into issues, (c) increased effectiveness of results in shaping governance, (d) improved resistance to manipulation through advertising/campaigning, (e) increased indisputability and trustworthiness of results, (f) anonymity of voters with unsaleable votes, (g) reduced opportunity for selective denial of voter access, (h) voter fraud only through improper voter rolls, (i) equivalent but safer decisiveness, and all with (j) significantly reduced direct and overall cost.”
The paper goes into more detail how it will technically work, how randomness is done, how to avoid voter coercion and so forth. Admittedly, although novel, it feels like the necessary technicality required makes it tougher to implement.
Through random sample voting it becomes much cheaper to know & understand what the populace wants at any point in time. This allows for an even further extrapolation: continuous elections. This idea was put forth by Steve Randy Waldman.
Continuous Lottery Voting
The premise is that on a regular schedule (say, every month), a random portion of the population is selected (as in random sample voting), and asked to elect new representatives. Of this smaller portion of people, a random ballot is drawn and the seat in the legislature is swapped for a new representative. In order to reduce volatile turnover, an additional 1/24 chance is added. eg: you flip 24 coins, and if only one of them is heads, you swap the seat. This means that the seat would only change every 2 years, on average. This gives elected candidates sufficient time to enact policies they were elected for.
I really love this proposal because it keeps representatives a lot more accountable. They are never really safe. In a world that changes much faster than when electoral systems were first developed, this allows the state to adapt more flexibly. Elected representatives can’t just rest on their laurels. Next month, their seat comes up for a shuffle and the randomly sampled population needs to vote to keep them in power. Politicians can’t just wait until election year to suddenly get busy and fix potholes.
Electioneering is also reduced. It becomes a lot more costly to manipulate the populace or attempt to buy votes. It doesn’t just happen around a singular event every few years. It’s also not known before-hand what portion will be randomly sampled, making it even harder to buy off a people.
Issues
All this sounds quite nice. The ability to reduce many of the issues plaguing modern electoral systems by simply introducing random ballots in the right places.
It’s cheaper, less corrupt and more representative. What’s the downsides?
I see potential issues around legitimacy (consent of the governed) & the unexpected consequences of complexity.
Legitimacy
Randomness has formed part of historical electoral systems. The Athenians used a Rube Golderg kleroterion for sortition, and the Doge of Venice were selected through a process that included choosing the first boy they see at a specific time & place to draw lots. Although these choices seem a bit bizarre, my hunch is, is that in the era of fake news, a modern electorate might push back on the use of randomness in legislative elections.
A Kleroterion
This is especially the case if a sample is chosen, because people might not understand how a sample comes to be representative. This is particularly in the case when you look at extremes like India. Less than 20,000 people have to be polled to be representative with a reasonable margin of error for a country of over 900 million registered voters. People might question the legitimacy if they are not a part of the sample: “Why am I governed by someone I didn’t even have a chance vote for?”
Many people have a hard time understanding randomness and statistics. A good example is how people play public lotteries. There’s biases towards gambler’s fallacies & hot hand fallacies. It can easily be dismissed that the randomness was tampered with, because the complexity involved in understanding very well-designed random systems is hard/tough. Would modern electorates trust a kleroterion for example? It’s really just balls falling down a pipe in random order. Maybe, but would a modern electorate trust something like a blockchain RANDAO, which is similar to David Chaum’s proposal on where randomness for random sample voting will come from?
That being said, today, randomness *is* used in areas where legitimacy is necessary. Notably, even though the outcomes are different, juries are elected randomly. It’s a part of *some* parcel of some societies. Some studies have shown that random selection is perceived to be legitimate. Pek, Kennedy & Cronkright have studied this in student governments in Bolivia.
In order to improve legitimacy, I see that there are 2 different ways to do so. One way is to add some ritual theatre & the other is to introduce the ability for a reasonable portion of society to participate in random number generation.
Ritual
Ritual & theatre, even when the practical outcomes are the same is important to generate a sense of legitimacy. In a paper comparing leader selection in consensus protocols & elections of Doges in Venice, Mowbray & Gollman, suggest a simpler variation that reduces the complexity of the election process.
As described before: electing a Doge of Venice consists of multiple rounds of lots & elections. One reason, I suspect, legitimacy was achieved, was the fact that it was simply a complex process. The reason is that the cost to understand & participate requires social interaction & consensus which adds eustress to the engagement, and thus a sense of legitimacy. Thus, even if the random outcome, might be the same, adding ritual & theatre to it might improve legitimacy because it feels like the populace participates in the process. It’s part of the same parcel of how humanity uses myths to coordinate. How add this ritual & theatre could be quite varied, but I’d like to put forth a simple proposal.
Davic MacIver argues that once all the ballots are counted in a full, electoral, lottery vote, there’s a virtual simulation that shows all the ballots next to each other. Like, traditional elections, it’s important to maintain secrecy of ballots to avoid electoral coercion & fraud, so you won’t know which ballot is a specific person’s ballot. From here, this deterministic simulation shows a specific ballot being chosen. Anyone can run a program that reproduces the outcome. What this simulation shows, however is important: it needs to show that all votes mattered in the production of a random outcome. Merely just being able to *see* all the ballots can add legitimacy.
However, I think the theatre & ritual could be more interesting.
A Giant Random Ballot Knockout Tournament
What if the ballots compete in rounds against each other in a knockout tournament, until there is one winner (for each seat)? It’s a giant random ballet, March Madness bracket. The winner of a coin flip (50%) per ballots competing against each other, advances to the next round. It will just be fun to watch, even though the statistical outcome would be the same. What’s interesting about this variation of ritual theatre is that one can see & understand how a small percentage of votes actually DOES matter, even though the chance of getting out at the top is small. If a candidate has 1% of the votes [say 500,000 ballots in a 50 million voter roll], you would see that the minority ballots would at least advance a few rounds, before losing to majority ballots. It would make minority voters see & understand that their vote does matter. On some circumstances, their candidate would actually proceed to win a seat.
With random sample voting, however, this is less likely to happen, since you won’t have a spread of votes as with a full lottery vote. Although you can query the populace more often, to reach everyone, even at a more rapid pace, it will take quite a while.
In order to attain more legitimacy this way would be to require all candidates (at least) to participate in the random number generation. This is a proposal also put forth by David MacIver. Using commitment schemes, such as what David Chaum, also proposes, everyone can contribute their own “random number” into the scheme, such that it becomes impossible to game. Joel Wietelmann has a great succinct summary (after he discovered a technical vulnerability in these schemes):
* Each participant picks a number and keeps it secret.
* Each participant publishes a hash of their secret number.
* Once all participants have committed, each participant reveals their secret number.
* The numbers are verified against their respective hashes.
* We combine all participants’ numbers to produce the pseudo-random output.
What’s interesting is that one can extend this to include various civic society institutions or even citizens themselves to participate in this.
So, in general, I think although there’s additional complexity involved, I don’t necessarily think it’s insurmountable.
Managing Complexity
If you use random ballots to achieve a continuous voting system, you might encounter unexpected emergence. The benefits seems great, but the act of having a system that responds more readily isn’t necessarily able to deal with the volatility or shocks as well. There’s potentially, less shock absorbers.
Sometimes systems need space to breathe. Nature has, for example, evolved pulsatile systems, for a reason. You can see this in mammalian hearts or the secretion hormones. In a continuous electoral system, you might have too much volatility to sustain longer-term change.
Partisan electoral shifts after every 4-5 years already has an effect on this, never-mind the likelihood of having elected officials change in shorter time frames due to randomness. The gain you get from having elected officials always be on the potential chopping block, might not be worth the trade-off of having them change too often. Certainty matters.
Even in the scenario where you have a turnover of roughly 2 years per seat, short-term volatility in randomness might be detrimental when it matters. Continuous voting proponents argue that this randomness could lead to elected officials opting for bi-partisan polices or bi-partisan staff in their elected offices, because if they get unseated, they could be back the next month (in the shortest time).
Regardless, there’s uncertainty around continuous preferences and randomness and how it might effect governance. Sketching extremes, we can see how it matters.
Take the following example: a continuous electoral without intervals (in other words, the populace votes ALL the time), and if the percentages for a candidate is 51/49 and shifts continuously, you might have an elected office continuously swap in/out of a candidate, resulting in nothing getting done. So, somewhere between current electoral systems and continuous, random elections is a great middle ground.
A middle ground that results in cheaper, more effective governance, less corruption, better representation & higher costs to manipulate/buy a populace.
Conclusion
As the pace of change increases, we do feel that the world is chaotic & random. If we add more randomness to our electoral systems: a cornerstone of modern, civic society, we can actually achieve more certainty. It allows us to poll the populace more often & more cheaply. It could lead to substantially better representation, favouring minorities more so than any current proportional representative system & reduce the power of money to manipulate what society wants.
The biggest hurdle is ensuring that because randomness & statistics are often hard to grasp, we introduce measures to improve legitimacy. These include exploring ways for people to participate in random number generation, alongside adding ritual theatre to the process. Adding lottery voting to current legislatures might be met with trepidation. We can however create citizen assemblies using these methods as civic society institutions. We poll the people using these methods and present the voice of the people to current legislatures to help inform them. These methods can be used for civic society institutions above and beyond nation states. It could also be used to poll governance in supra-national systems such as blockchain ecosystems.
There’s a of research in the latter (blockchain space) that also readily informs how we improve decision-making, and allocate resources. The rabbit hole goes deep and I hope to supplement this with research and writing from this domain (or should I say rabbit hole). We have tools like quadratic funding for allocating resources to public goods, holographic consensus, budget boxes, curation markets, & social sensor fusion to draw from.
By adding a dash of random walks across the aisle, we can be more certain of our future.
—
And now, for the outro: A Live Rendition of Radiohead’s “Meeting In The Aisle”.
