The System

At both the local and global levels The Matching System shall be used to enable The People to run their government. All individuals will be guaranteed access to it as a fundamental human right.

Our current political systems are generally classified as “representative democracies”, where The People are required to delegate responsibility for the vast majority of decisions about the laws and policies that affect us to “representatives”. The general theory is that these “representatives” are somehow “representative” of us, and would therefore make decisions that the same decisions we would make if we were in their position. But what if none of the candidates for a given office are like you at all? For example, which of the two major 2016 US presidential candidates are you like: Donald Trump or Hillary Clinton? For most people the answer is “neither”. And yet these were the only two realistic choices for a leader who would be placed in charge of making a vast number of decisions for us. The situation is even worse when deciding which candidate for senate/congress/county commissioner/city council/etc.: Not only is it unlikely that they will vote like you, in most cases you very little information about what they’re like and so it is impossible for you to predict how they will vote at all!

The System, an electronic direct democracy and information management system, will work from the opposite direction: Rather than requiring you to research and then pick “representatives” from a very limited set of choices and then trust them to make good decisions on your behalf, you will instead provide information that will be used to match you to proxies that have already voted on a proposed law or policy, specifically those proxies that most likely voted like you would if you had the time, experience, and motivation to properly research that proposal. You will then have the opportunity to verify this automatically-calculated vote by examining the justifications for those votes provided by your proxies. If you agree with their reasoning, or if you have learned that you can trust the system’s proxy-matching ability and so don’t even check up on them, this proxy vote will be recorded on your behalf. If you don’t agree, you can cast a direct vote to override your proxies.

General Voting will happen over a weekend for maximum convenience of the voters. The proxy batch calculation will run every Friday at midnight UTC (4pm US Pacific time). After the proxy votes are calculated each voter get an email or text message telling them what vote has been calculated for them. They will then have the weekend to look at how the proxy candidates voted and why, and to override or selectively eliminate each proxy if they disagree with the justifications supplied. Voting closes Sunday at midnight UTC and an email/text will sent to each voter with the result and an announcement of the next weekend’s proposal(s).

If a voter has the time and inclination to do the research and consideration necessary to cast a direct vote and wants to make themselves available as a proxy to other voters they’ll have at least 10 days to do so (the proposals will be put up at the latest the Monday the week prior to the General Vote). These early voters will have access to the information in the proposal and to comments provided by any proxies who have voted before them, but will be reminded to vote carefully: Other voters will be relying on these decisions and votes cannot be changed once they have been cast.

The distinguishing characteristic of The System is the way voters are matched to their proxies/delegates/representatives. Unlike other systems, this system is specifically designed to maximize the chances that the vote recorded for an individual is most likely to be the same vote that voter would cast if they had the time, experience, and motivation to properly research an issue prior to casting a direct vote (i.e., casting a “correct vote”, as defined in Ha & Lau 2015 and Milic 2012). It will do this by matching the voter based on relatively invariant personality characteristics such as the Five Factor Model (aka the Big 5 personality traits) along with other measures of personality, political orientation, and possibly even their past votes if a voter has cast a sufficient number of direct votes. Although there isn’t any direct research to support the viability of this design yet, there is considerable research that these traits have high predictive value for things like political orientation and party affiliation (for a review see Gerber, Huber, Doherty, & Dowling 2011), which in turn is at least somewhat predictive of positions on specific issues (indeed, by bypassing parties and the commonly used yet relatively crude measures of political orientation as intermediates we may find that the predictive power of the statistical match is far greater).

The System also differs significantly from other systems in that these calculations are done automatically and by statistical analysis rather than relying on human’s rather horrible ability to pick their own proxies. For example, Kling, Kunegis, Hartmann, Strohmaier, & Staab 2015 looked at the operation of the LiquidFeedback system as it was used by the German Pirate Party and found that when votes cast prior to delegation were compared, the votes of voter-selected delegates only correlated 0.6 with the votes cast by the voters themselves. Yet there was a 0.5 correlation among voters chosen at random, which means that manually selecting delegates is not significantly better than using random sampling (albeit without the flaws in lottocracy/demarchy/sortition, all of which have at least the same fatal flaw exhibited by the judicial jury selection process: Anybody who wants to get out of serving can almost certainly do so, leaving only the least capable (or most authoritarian) to render the decision, which is exactly the opposite of what you want when making public policy decisions). This inability to choose appropriate proxies is undoubtedly due to the fact that people use inappropriate characteristics to select their representatives (name familiarity, friendliness/extraversion (who you’d most like to have a beer with!), recommendations of other non-representative individuals, etc.) just as they do in our existing misrepresentative democracies.

Kling et al. also found several other serious flaws in the operation of the LiquidFeedback delegation system, including inadequate utilization (less than 15% of participants ever used this feature), dead ends (voters delegating to a person who didn’t cast a vote), and a significant effect of what they called “supervoters”, individuals who somehow managed to amass hundreds of proxies and were therefore granted the ability to act as oligarchs, subjecting them to the same sorts of corrupting influences that currently plague our existing misrepresentative democracies. By automatically recalculating proxy matches for each vote, The System will solve all of these problems, and achieve vastly better representation to boot. Further improvement will result from providing the voter with multiple proxies (5 in the current version) whose votes are averaged, reducing the effects of outliers, some of which could be individuals who are attempting to game the system by providing fake personality profiles.

Automatic proxy matching also solves what is undoubtedly the biggest problem of both LiquidFeedback and our existing systems: Low voter turnout. In the Pirate Party LiquidFeedback runs most decisions were made with only 20% of the members casting a vote. In the US 2016 general election only about 40% of the population voted. Even the Swiss system, currently the most widely used direct democracy system in the world, generally has turnouts below 50% of registered voters. But once established, every vote in The System will have near 100% turnout, ensuring that the opinions of the all of the population, most of whom are not represented at all in our currently elections, will be taken into account when making public policy decisions. The vast majority of these newly-represented voters will be Neurotypical and so far less likely to be aggressive, discriminatory, or authoritarian (i.e., SDAP) in their voting preferences.

After voting, each voter will be given the ability to write a comment (or even a whole dissertation) on why they voted the way they did, or to choose posts that they found most influential in making their decision. To prevent these voters from attempting to game the system by suppressing the opinions of or explicitly attacking the other side, they will only be allowed to contribute to their side of the argument (i.e., if they voted “no” they’ll only be able to post in the column of reasons to vote against a proposal, or to choose the posts from that column that they found influential). Other voters will be able to use this information to make their own decision, or to evaluate the quality of their matched proxies, an instinctive ability every human has from a young age even when they have little personal expertise in a domain (e.g., see Keil 2010). This maximizes the information each direct voter will have access to, while preventing the same type of trolling behavior that has become endemic in all other social media: Unlike them The System is designed to allow The People to make decisions, not to collect eyeballs to sell to advertisers. The best social engineering techniques can therefore be used to ensure a civil and efficient debate rather than just using the system to stir up emotions to attract marketable attention.

Other key features of The System will be:

  • A resilient design that is distributed across servers in multiple locations.
  • Robust cross-platform support, from browsers and apps on all platforms to support for voting by email or SMS in very low bandwidth environments.
  • Open source, such that anyone can inspect and improve it.
  • Security sufficient to protect against or at least detect vote rigging or other tampering, although nowhere near enough to guaranty 100% accuracy of results immediately (and of course individuals who insist on this are probably authoritarian and so not interested in efficiency or fairness, but primarily in control).
  • A “party prevention” design: Not only will The System not include any support for political parties, it must be specifically designed to prevent them from arising spontaneously. Political parties arise because they are a means of aggregating political power from individuals of a particular personality type and then wielding it to achieve that group’s goals. Unfortunately, parties don’t even properly represent their members because individual variations are far too large to be captured by any party system unless it supports at least dozens of parties: Any party system by definition therefore prevents people from casting a “correct vote”. Because Matchism is based on the fundamental assumption that the power of all personality types must be weighted in proportion to their prevalence in the general population, political parties can only be considered to be a means of “gaming the system” because they facilitate some personality types wielding political power out of proportion to their numbers. This is obviously incompatible with Matchist philosophy and social engineering principles. It probably should even be considered explicitly immoral in all philosophies because some personality types simply don’t aggregate the way others do. For example low Big-5 Openness scores (authoritarians and other conservatives) naturally aggregate, whereas organizing high Big-5 Openness individuals (some Greens, but most of which are not well represented by any current political party) is akin to herding cats.
  • A testing system to regulate Credentials, information that will be available to voters to provide them with an additional tool that can be used to assess the credibility of the proxies they have been matched with.
  • The ability to block certain proxies from being matched with them if they find that a proxy takes positions or makes comments that the voter disagrees with on principle. This will also address any proxy’s attempt to game the system by faking their personality profile: The research has shown most people will be able to detect this because they can determine most relevant dimensions of the Big 5 personality profile using only the potential proxy’s on-line posts (Darbyshire, Kirk, Wall, & Kaye 2016).
  • The ability to record a changed vote after the voting has closed. If a large number of these reversals are recorded, the proposal will automatically be rescheduled for a revote.
  • Although it expected that the vast majority of proposals to be voted on will come from qualified Social Engineers in the Executive branch of government, add-on systems will aid in creation and refining of proposals submitted by individuals. Note, however, that The System itself is not a deliberation or consensus-building system, all of which have the fatal “last man standing” flaw where even a single user who insists on getting their way holds up all progress until a compromise acceptable to them is made. This kind of behavior was simply not tolerated in the EEA, is fundamentally incompatible with human social Clinations, and so does not need to be included in any modern decisionmaking system.

The proxy system will address most of the problems raised in Hibbing’s 2002 Stealth Democracy which showed that although people greatly dislike our current systems and yet don’t want to be involved in all of the low-level details of running a government, they do want to be able to ensure that the people they chose to represent them are not capable of exploiting those positions for their own personal gain. Most people also do want to retain the ability to vote directly on those few issues that are important to them.

The results of all votes would be available at closing so that The People can check that their vote had been recorded and that there was no ballot-box stuffing, just as is the practice for representatives in representative-based government. After each vote, they will also get an text/email showing what vote was recorded for them along with the vote totals. This virtually eliminates the possibility of fraud because any tampering would be instantly detectable by anyone who even glanced at their phone. It also vastly simplifies the infrastructure: No need for fingerprints or other exotic authorization methods, no need for The System to be able to verify individuals using their financial information (i.e., recording credit card or bank account or other information useful for identity thieves), and no need for armies of system administrators to monitor and protect the system from attack. Indeed, it may not even be necessary to validate any sort of voter ID at all so long as citizenship rolls are properly maintained: If someone else votes using your name/ID, it’s your own fault, and this alone should provide a proper incentive to make sure that every person at least signs up with the system and does at list a minimal monitoring of the votes cast on their behalf. If the system does at some point get hacked, large numbers of users will find out about it and the only negative consequence will be that a vote or two may have to be redone.

Although this is different from the secret ballot system used in most representative democracies, the choice between support for proxies and making fraud straightforward to detect versus the possibility of making vote buying and voter intimidation easier is a clear one. Indeed, by making the latter easier to detect (e.g., via statistical analysis of voting patterns and comparisons with the personality inventories) public voting would seem to be the best of both words in a technologically advanced civilization.

This type of open voting is not unfamiliar to most people: Caucuses used in primaries in many states work this way, as do most votes held in meetings in most organizations. Most people do not even have any reluctance to share their voting preferences with others (e.g., Gerber, Huber, Doherty, & Dowling 2012). Kling et al. (private communication) received virtually no reports of reluctance to cast a direct vote in LiquidFeedback as a result of it having open votes. It has also been shown that the individuals who are most insistent that a secret balloting system be used are the very people who are the least informed on the issues and so would most benefit from having their votes cast by a representative proxy instead of directly (Gerber, Huber, Doherty, Dowling & Hill 2013).

Screen names will be used primarily in The System, but it should be possible for people to match screen names to real-world names with some effort. Not only does this allow independent inspectors to check up on the voting process, but prevents the system from being held hostage by a hacker who manages to break into the main database. Social engineering tools should be sufficient to keep the screen-name/real-name information from being used inappropriately (e.g., someone including this information in one of their posts in an attempt to influence or intimidate another user would receive some sort of censure).

It is a common criticism of direct democracy systems that the large number of decisions that (at least supposedly) need to be made by government effectively prohibit individual participation in making them. But if you actually look at the number of substantive decisions made at the local, state, and national levels you’ll find the number is already quite manageable: A weekly vote on 1 or 2 proposals would be sufficient (although a lot of bridges and post offices will probably end up with boring and utilitarian names, nor is there any chance there will be over 60 votes to repeal Obamacare like the US congress has). For each proposal an individual would at most have to read a few concise and well-written statements in favor of the proposal, and equally concise and well-written statements opposed. Of course, if they needed more information prior to voting they could read the proposal itself along with every comment (in ratings order) about it that anyone else had written (and there will be a lot of that information available because instead of whining about the way things are on Facebook or some other forum, people will instead express their opinion in a place where it actually matters!). Making a full evaluation and casting a direct vote should only take an hour or so a week for most people, and once people learn to trust the proxy-matching system the vast majority of them will only spend a few seconds a week monitoring the system.

It is also frequently observed (including by the Founding Fathers of the United States) that direct democracy can’t work because The People are not up to the job (too stupid, too apathetic, too prone to mob effects, etc.). Many evolutionary psychologists would extend this assessment to human performance in large-scale political systems in general: Humans simply don’t have the evolution-tuned hardware to perform well in that environment (Geher, Carmen, Guitar, Gangemi, Aydin, & Shimkus 2016, Petersen & Aarøe 2012). But these criticisms assume The People are unaugmented by technology. The Systemwill be designed and engineered to improve human decisionmaking ability by providing exactly the information needed and in a directly usable form. It will facilitate voters making decisions rationally using only this information instead of relying on lower-level processes, biases, or prejudice. It will provide an environment that behaves and feels more like a small-scale society even though it is designed to scale to accommodate the entire global population. It will amplify collective power while minimizing mob effects and therefore will allow us to function far more efficiently and competently than we are able to now in any current political system.

Only votes from individuals with a Level 1 Credential in Matchism and with their primary residence in that Locality will be counted for proposals affecting a Locality. Globality elections will count the votes of all individuals with a Level 1 Credential, regardless of their location.

Yes, this means that issues affecting the Globality will at first be decided primarily by people living outside of Localities that have adopted the current Matchlist. There are two reasons why this is not only necessary, but optimal. The first is that statutes and policies in the Matchlist will eventually apply to all human beings and so all human beings should have a say in how they are defined. Allowing the first few adopters of Matchism to define these laws and policies is a recipe for disaster since they may implement culture, language, or location-specific policies that might indirectly (or even directly) harm individuals in pre-Matchist nations. This could in turn cause unacceptable delays in the adoption of Matchism by those nations.

Secondly, the first few adopters will likely lack the financial and organizational resources to implement Matchism on their own. All The People will need to pool their efforts to ensure that these early adopters are successful so that we will all eventually be able to live in Matchist Localities.

Also, while Matchism is generally compatible with any true democracy, it is fundamentally incompatible with the existence of any authoritarian regime. The research on SDO and RWA clearly shows that as long as any authoritarian regime exists in the world, no one is safe. While many westerners seem fascinated and sometimes even obsessed with the Chinese citizens’ failure to fight their government’s efforts to restrict dissent and access to information (“Don’t those people value freedom?”) this is the equivalent of worrying about the fleas on a rabid dog. The real problem is that any government, even a highly authoritarian government, works adequately during an interval of plentiful resources. It’s when an era of resource scarcity returns that authoritarian governments turn to aggression, either against outsiders or even against their own people, to maintain their grip on power.

Only by eliminating all sources of external threat is it possible to control the activation level of Authoritarians in one’s own country. Replacing all SDAP-led governments (and this includes most representative democracies) with Matchism must therefore be a high priority. The most efficient way to do that is by having these citizens living under SDAP rule participate in the Globality to the greatest extent possible until the Matchish grow to critical mass in that country.

This also means that Matchism is not suitable as a cult or government for a commune-like small-scale society that must rely on external support or a self-selected population: Matchism either works for all of us or it works for none of us.

The decision criterion for elections shall be a simple majority of votes cast. To change the Matchlist itself, however, the criterion shall be a majority vote in an election where a majority of eligible voters cast a direct ballot.

Note that there is no provision for Locality-count input, i.e. as is the requirement in the US Senate or for Swiss cantons in national referendums. Allowing this encourages splitting up Localities into smaller units, reducing their efficiency. Localities should be competing for citizens to grow to the largest size possible, not provide individuals with an incentive to do the opposite. Locality-count voting also encourages efficiency-killing hostage-taking behavior (i.e., tribalism, an authoritarian mindset and tactic). If there is a need to protect Localities from tyranny-of-the-majority issues, it should be sufficient to move responsibility for some decisions (e.g., on resource extraction policies) further into the local domain.

Next: On Representative Democracy