Note: This reflects my thinking on the subject as of late 2017. My thinking has of course updated, but this text has not.

OPHELIMO is a new type of democracy.  The name is based on ωφελιμίσμος, the Greek word for utilitarianism, because ophelimist democracy borrows from classical utilitarian thinking about achieving the greatest happiness for the greatest number.

  1. Motivation: What’s wrong with our current form of democracy?
  2. Goals: What characteristics should a better form of democracy have?
  3. Explanation: How does ophelimist democracy work?

I. Motivation: What’s wrong with our current form of democracy?

It’s badly designed. A badly designed form of government makes it easier for us to end up with bad consequences and harder for us to achieve good consequences.  All existing forms of democracy are badly designed, and the design of the system in the United States, where I live, is exceptionally bad.  Briefly, here are a few of the worst flaws:

  1. Many people who are affected by the results of an election cannot vote.
  2. Many people who can vote choose not to because they do not like their options.
  3. Many people who can vote choose not to because their votes are unlikely to change the outcome.
  4. We use a winner-takes-all form of democracy, where the losing side gets nothing. And the losing side is often nearly half of those who did vote.
  5. We use First Past the Post (FPtP), a voting method that uses the least possible information about voters’ preferences.
  6. FPtP causes vote-splitting and the spoiler effect when similar candidates run.
  7. Instead of ditching FPtP, we tried to fix it by having partisan primary elections, which locks in political parties as part of the system, giving party leaders immense control over the system from behind the scenes.
  8. FPtP plus a locked-in system of political parties forces us toward long-term two-party dominance, as each major party evolves to appeal to about half of the voters. This narrows the range of policy options to just two.
  9. Two-party dominance rewards negative campaigning, where each candidate focuses on stoking fear and hatred of the other candidate rather than making a rational case for their own ideas, since fear and hatred are more powerful motivators.
  10. Long-term two-party dominance and negative campaigning cause factionalism, where people form “tribal” hatred of the other side, and are unable to cooperate with, sympathize with, or understand each other.
  11. Factionalism leads both parties to criticize each other’s actions without regard for whether the actions are good or the criticisms are correct, so that political debate is dominated by dishonesty and hypocrisy.
  12. We elect representatives from geographic districts though it has been generations since people felt their geography was a defining part of their identity.
  13. Parties and politicians often gerrymander districts to produce electoral results as favorable as possible to themselves, regardless of how the people vote.
  14. In districts where one party is sure to win the general election, the partisan primary becomes more important. But there is no primary-to-the-primary, so the vote splitting and the spoiler effect problems are back, leading to victories by highly unrepresentative candidates.
  15. Furthermore, partisan primaries are dominated by the voters who are most committed to the party and most pleased with its politicians, so the candidates they choose grow more extreme over time compared to the constituents as a whole.
  16. In fair, competitive districts, candidates trying to win both primary and general election will strategically want to appeal to extremist primary voters and then pivot to appeal instead to moderate general-election voters. When this happens, the primary supporters and the moderates being pandered to realize that politicians will say anything to get elected, and so even if they do vote they will grow to loathe and distrust politicians as a group.
  17. Candidates need to raise large amounts of money to persuade uncommitted voters to show up and vote, and so their focus is on fundraising and constant campaigning rather than on the quality of their ideas.
  18. Candidates often seek office out of a desire for power and status, rather than a desire for the common good.
  19. After winning office, an official still must optimize for electability next cycle, focusing on getting media attention that will earn them activists and donors rather than focusing on producing good laws, or else they will be vulnerable to another candidate who is optimizing for electability in those ways.
  20. Our representatives then vote by majority-rules, adding a second layer of the winner-takes-all problem.
  21. Furthermore, among the representatives the majority party elects a Speaker, and the Speaker alone decides what bills will even be voted on.
  22. At the national level, we also have the Senate which is not democratic in the first place, and the electoral college, which is winner-take-all in most states and warped in favor of low-population states.
  23. Because it’s so hard to pass a law and incredibly hard to amend the Constitution, the legislature continually resorts to delegating more power to the executive branch.
  24. When our legislators finally do pass a law, they often do so while blatantly lying about the effects of the law and their reasons for voting.
  25. Any politicians that manage to gain power in the above system will have incentives to make it even less democratic, since that way they gain power, whereas if they tried to fix it, they would lose power.
  26. Legislators don’t have incentives to pass good laws. They get many expensive perks regardless of whether they pass good laws or bad laws. It’s often not clear to the voters until much later whether a law was good or bad, by which time it’s too late to hold the legislators accountable.  And if they solve problems, their voters stop showing up to support them.
  27. Even when people can vote and their vote matters, it turns out that the vast majority of voters are extremely ignorant, irrational, biased, short-sighted, and selfish.

Given all of that, to what extent is the U.S. actually democratic?  Just barely, if at all. The American system does not have sufficient “people power” for the people to ensure that we get laws serving the common good — not even when there is supermajority support among the population.  And in situations where the U.S. is democratic, the process and results are ugly enough often enough that people grow skeptical whether they want better democracy.

Some countries do have slightly better designs for democracy: proportional representation, unicameralism, parliamentarianism. Some scholars and activists promote even better alternatives like score voting, sortition, liquid democracy, storable votes, futarchy. Each of these ideas fixes some of the problems above.

But ophelimist democracy fixes them all.

II. Goals: What characteristics should a better form of democracy have?

A better democracy should be directly democratic but non-majoritarian.  Ophelimist democracy has no arbitrary thresholds at all.  It’s not winner-take-all; instead, it treats proportional influence as a critical part of the design.  “Proportional influence” means that every voter achieves roughly the same amount of effect on the real-world outcomes. Every vote meaningfully affects the results; not a single vote or fraction of a vote is ever wasted.

A better democracy should be anti-factional. In ophelimist democracy, there is zero advantage to be gained from forming a political party or indulging in tribalism. But there is also no disadvantage to banding together and working to raise consciousness about issues, so ophelimist democracy doesn’t lead to social atomism either.

A better democracy should be “ethocratic“, meaning that its design makes moral legislation easier to pass and immoral legislation harder to pass.  Ophelimist democracy achieves this via its polling system, which measures total happiness in a utilitarian sense. In the real world people find it impossible to usefully think about pure “utility”, so the polling system’s measurements are broken down into manageable issues that people understand and care about.  Every voter gets equal consideration for their happiness and also for their ideas about what happiness really is.

A better democracy should be “epistocratic“, meaning that its design makes the honest use of knowledge essential to the legislative process.  Ignorance and knowledge should never be treated the same. In ophelimist democracy, every voter’s knowledge contributes by making predictions about the effects of laws — and the predictions are scored later. The basic notion of prediction-driven democracy is borrowed from futarchy, but ophelimist democracy eliminates futarchy’s unfortunate dependence on money and markets. Instead, each voter’s prediction is treated proportionally to how accurate that voter’s predictions have been in the past.

A better democracy should be capable of considering many options simultaneously without encouraging strategic dishonesty. Ophelimist democracy does this using score-voting. It inherits from score-voting immunity to most impossibility theorems about voting (e.g. Arrow’s theorem, the Gibbard-Satterthwaite theorem, the Duggan-Schwartz theorem, etc.). Ophelimist democracy even triumphs over Gibbard’s theorem over the long run because of its use of proportional influence: you can vote strategically to give yourself more influence now, but only at the cost of having less influence later.

A better democracy should have built-in feedback loops that cause it to do the smart and good thing whether or not the members are smart and good, to preserve and strengthen itself whether or not the members have a long-term vision, and to inclusively spread the democratic franchise to previously disenfranchised people whether or not the members have inclusive beliefs and attitudes.  In ophelimist democracy, all this is achieved via the predictions and the polling.

A better democracy should be easily built within the shell of the old. No revolution or miracle of social coordination is required to implement it.  No politician has to vote against their own career interests.  In our current democracies, most of the best reforms have no possibility of being achieved given that opponents have the incentive and ability to block the reforms.  But for ophelimist democracy, there really is a way to get from here to there.  Just download the app and start using it for your organization.  The “laws” the group passes start out as merely the organization’s bylaws, of course.  The feedback loops within ophelimist democracy automatically encourage growth.  As ophelimist democratic organizations grow large enough, they will have three routes forward: either the organizations gain enough power to boss around the politicians, or the government adopts a similar ophelimist democratic structure, or weak governments get replaced by an ophelimist democratic organization that the people view as more legitimate and more effective.

III. Explanation: How does ophelimist democracy work?

Ophelimist democracy is relatively complicated, so I imagine that it begins with an app that anyone can use via personal mobile phone, PC in a public library, or another device.  Anyone can create a group on the app and edit its membership roster.  The app handles all the technical details of this system.  Voters simply see and respond to poll topics unless they choose to look into the workings of the system.  (Note: The app does not yet exist!  My work hours are long and exhausting, so I may never build it.  You are welcome to build and release such an app if you wish.)  The app is not strictly necessary; all this could be handled manually by a bureaucracy, though if it was handled manually it should be simplified.

Ophelimo has several elements.  The key elements that create the positive feedback loop are polls and predictions.  The polls ask voters how satisfied they are on several topics, and people try to predict what the answers will be.  The predictions then are used to pass the bills that will make people the most satisfied.  Then the consequences of the passed bills affect how satisfied people are, and the new polls are used to evaluate the accuracy of the predictions.

In addition to the polls and predictions, there are preferences and officers.  Preference votes serve two functions.  First, they guarantee proportional influence.  Second, they help keep the predictions honest.  Officers are elected using a special modified sortition to ensure that they are qualified and of good character.

Finally — here come the nitty-gritty details!

  1. Choosing Polling Topics
  2. Polling
  3. Choosing Bills to Consider
  4. Making Predictions
  5. Passing a Bill
  6. Electing Officers

(1) Choosing Polling Topics

What the user sees:

Anyone can propose a new polling topic at any time. The purpose of each topic is to help the voters evaluate how satisfied they are about the part of life mentioned in the topic.  The topic can be a question or a short phrase or statement, and each voter is supposed to consider the topic and then give their own personal judgment of how well or badly things are going.  For example, topics might include:

  • How happy are you with your job?
  • Road safety in your neighborhood
  • Scientists report that ocean animal species are rapidly going extinct.

To choose polling topics, every member gets 1 vote which is divisible down to many decimal places. Members can search through the topics that people have proposed.  Any member can at any time vote for a topic using the score voting method.  That is, voters can give any topic any score in the range from 0 to 1.  The higher score they give a topic, the more likely it will be shown to voters, and if it does get shown, the bigger effect it will have on the final outcomes.  The lower score they give a topic, the less likely it will be shown to voters, and if it does get shown, the smaller effect it will have on the final outcomes.

Behind-the-scenes details

  • Non-members of the group are included in “anyone”. They might have important ideas for topics that members didn’t think of.
  • The form of the topic is not policed, but each member of the group uses their own judgment no matter how it is phrased. For example, if a fan of the Pittsburgh Steelers sees, “The Steelers sux this season! Say it’s the best the more they sux!”, he can and should ignore what the topic literally says and instead honestly report that he’s happier when the Steelers are doing well.
  • Every voter will have the chance to respond regarding all the same topics. Therefore, it makes most sense to vote for topics where nearly everyone agrees with you about what kind of outcomes are good and what kind are bad.  This means that the optimal strategy for individual voters is to vote for topics that are about a genuine common good, universally relevant, secular, and non-partisan.
  • Reweighted Range Voting is used to select a proportionally representative, manageably small set of topics. The number of topics can be set to any convenient value, such as 20.  Using a small set is important for four main reasons: to foster a sense of cohesive purpose or “political mandate” for the group, to encourage members to actually consider topics other people submitted before submitting their own, to keep out prank topics and crazy topics, and to make the prediction task easier.
  • Once that set is determined, the total score for each topic determines the weight given to that topic when the poll results are aggregated.

(2) Polling

What the user sees:

Every member is polled periodically by the app regarding the topics that the group chose. (Responding to the poll is voluntary, of course.)  When responding to a poll topic, the voter is presented with a scale and can select any position on that scale that describes their view of the situation, all the way from “worst possible” to “neutral” to “best possible”, with many subtle gradations in between. Each topic is presented 3 different ways:

  • once regarding the present and recent past,
  • once regarding the near term future, and
  • once regarding the long-term future.

This is done because it’s essential to take long-term thinking in account, and the easiest way to encourage that is to explicitly ask about it in a separate way.

Behind-the-scenes details:

  • Each answer on the scale is treated as a number in the range [-1,+1]. The textual explanations of the scale positions are merely the human-readable versions.
  • The duration of the polling period can be set to any convenient value, such as 1 week.
  • At the end of each polling period, the answers from all members are totaled. Both the totals for individuals topics and for the weighted aggregate of all topics are reported. The totals from each period are used to evaluate the predictions for the most recent winning bill.

(3) Choosing Bills to Consider

What the user sees:

Anyone at any time can propose a bill.  Anyone can make a modified version of someone else’s bill.

To choose a set of bills to consider, every member initially gets 1 vote which is divisible down to many decimal places. That number can change later. Members can search through the bills that people have proposed.  Any member can at any time vote for a bill using the score voting method.  That is, voters can give any topic any score in the range from 0 to X, where X is the number of votes they currently have.  The higher score they give a bill, the more likely it will be passed.  The lower score they give a bill, the less likely it will be passed.

Behind-the-scenes details:

  • Non-members are included in “anyone” again. They might have important ideas for bills that members didn’t think of.
  • The form of bills is not policed, but badly-written bills are unlikely to voted up. Similarly, in the prediction phase, badly-written bills are unlikely to be predicted to be high value.
  • Choosing bills works almost exactly the same way as choosing polling topics. We end up with a proportionally representative, manageably small set of bills, and a weight for each bill according to how strongly people preferred it.
  • “every member initially gets 1 vote” – See step 5 below.
  • The votes for bills are tallied separately from the votes for topics.

(4) Making Predictions

What the user sees:

Anyone can try to predict what the polling totals will be. For one bill and one topic from the small sets, the voter uses the app to predict a probability distribution over what the polling totals will be each polling period for that specific topic, conditional on that bill becoming law. The app gives a convenient and intuitive interface for entering the distribution, so that a voter doesn’t need any advanced statistical knowledge — though having statistical knowledge will surely help.  Each voter may give predictions for however many topic-bill combinations they wish.

Predicting higher polling totals for one bill than for another is the same as predicting that people will be more satisfied with the effects of the first bill than the effects of the second bill. Also, predicting higher polling totals for a bill makes it more likely that the bill will become law. Importantly, the more accurate your predictions are, the more weight your predictions will have in the future.

Behind-the-scenes details:

  • Non-members are included in “anyone” again. They might have valuable knowledge to include in the predictions. Their scores are merely advisory, however. They don’t change the results.
  • To avoid putting less active users at a disadvantage, every user actually gets a default prediction for every bill-topic combination. The default is the historical distribution of polling totals for that topic, or, if it’s a new topic, the historical distribution across all topics. That means users only have reason to manually make a prediction when they believe the polling totals for a topic will be different than the status quo.
  • The system keeps a record of the past accuracy of each user’s predictions using a slight variation on relative information scoring. Each user starts with a score of 0 until, by making correct predictions, they gain a higher score. The slight variation has two differences from ordinary relative information scoring.  First, wrong predictions do subtract from a user’s score, but the user’s score is never treated as being below 0. Second, in each polling period, the total magnitude of changes to members’ scores is normalized to 1, so that there is no double-counting of information.
  • The predictions are aggregated like a weighted average, where the weight given to each user’s predictions is their current score.
  • Using those aggregated predictions and the weights on the polling topics, the “value” of each bill is its median predicted polling total.

(5) Passing a Bill

What the user sees:

Bills are passed periodically.  If the bill is passed by a group, it becomes one of the group’s laws.  Sometimes a voter has more or less than 1 vote for choosing bills. It’s usually a fractional number of votes. When you got your way last time, your number of votes dropped a little. When you didn’t get your way last time, your number of votes increased a little.

Behind-the-scenes details:

  • The winning bill is the one with the highest weighted value.  The “weighted value” of a bill is the product of the value from step 4 multiplied by the weight from step 3.
  • Only one bill becomes law each cycle, but that bill can be arbitrarily complex. It may address many diverse subjects if its authors wish. Or if people don’t want any changes, they can pass a “status quo” bill that doesn’t change anything.
  • The length of a legislative cycle can be set to any convenient value, such as each quarter of the year.
  • The votes cast for a winning bill are considered “spent”. They are redistributed equally to everyone who voted.
  • The redistribution together with the polling and predictions guarantee, over the long run, that everyone’s preferences are taken into account to the degree that they are humane, and that everyone’s humane preferences are realized to an equal extent.
  • Strategically, the redistribution strongly incentivizes the majority to produce compromise bills that are similarly satisfying to all groups.  If they do not, then the majority will lose power to those other groups, who can repeal the bills they disliked and replace them with their own attempt at a fair compromise.
  • At the time a bill is passed, the predictions for it are fixed–they can no longer be changed.  Similarly, the set of topics for that bill also becomes fixed, and those topics will be shown in polling until the next bill is passed.
  • The predictions for the most recent winning bill are evaluated every polling period. The polling periods should be much shorter than the legislative cycle so that there is quick feedback.  This means the predictions get evaluated many times across changing conditions; voters who want to make predictions will need to take that into account.
  • These preference votes also serve a technical requirement of the prediction phase: they introduce a bit of uncertainty regarding the winning bill, which is the only way to ensure a risk of punishment for people who give false lowball predictions in an attempt to kill bills they dislike.

(6) Electing Officers

What the user sees:

Sometimes there are bills about electing officers. Members can vote for the bills in the usual way and make predictions about the bills’ (candidates’) effects in the usual way.

Behind-the-scenes details:

  • This method of electing officers is not strictly part of the basic design. Parts 1–5 are what establish a feedback loop optimizing for humane values. But insofar as effective governments and organizations must always put executive power in the hands of specific individuals, I believe it’s essential to start with a system already in place to choose executives in a manner consistent with the optimization.
  • Every member is rated on the accuracy of their predictions, as discussed.  But in the background, the system also keeps track of the correlation between their votes and their predictions. The product of their accuracy score and that correlation is called “candidate quality”.
  • When the law requires officers to be elected, first use sortition to select a random slate of candidates with probability proportional to their candidate quality.  (The number of candidates can be set to any convenient value, such as 10.)  Sortition ensures that the type of person who craves power over others has no direct route to acquiring it. Instead, this sortition relies on candidate quality, which is a measure of the degree to which a person understands what policies serve the common good and of the degree that they value the common good rather than their own personal interests.
  • A special set of electoral bills are automatically generated: one for each candidate in favor of electing that candidate, one in favor of selecting a new slate of candidates, and one more in favor of leaving the office unfilled. Predictions are made regarding the electoral bills in the same way as for ordinary bills, to help pick the candidate who will best satisfy the needs of the people.
  • When it is time to elect one of the candidates, a special vote is held in which only the electoral bills are used. The one that passes determines the winner, whether new candidates are needed, or if the office should remain empty this cycle.
  • The predictions on electoral bills are on a different cycle from the legislative bills. They continue to be evaluated until the next election to that office.
  • The same votes are used for electoral bills as for legislative bills, to ensure fully proportional influence.
  • If an elected officer proves to be a bad choice, it’s always possible to pass an ordinary bill calling for a new election.


If ancient Athens was democracy 1.0, and modern rich nations have democracy 2.0, then Ophelimo is at least democracy 4.0. It’s a giant leap forward.

It isn’t final. Not all the necessary details are specified in the explanation.  There are elegant ways to refactor the design; I’ve kept this modular version because it is simpler to explain. To be more fully utilitarian, I would want to include two additional major features: a “Voice of Nature” and a “Voice of the Future”.  The Voice of Nature is a fancifully named institution that represents the interests of animals, plants, and other nonhuman living things.  The Voice of the Future is a fancifully named institution that represents the interests of people many generations hence whose lives will be shaped by decisions taken today.  But neither is part of the core Ophelimo design.

Ophelimo isn’t perfect. Predictions can fail. Unforeseeable circumstances may dash even the best plans. Voters with low knowledge or low tolerance for complexity may dismiss it as “technocratic”. Voters with racial bigotries or class-war strategies may fight against it due to its proportionality.  Voters with dogmatic politics may reject it because it does not try to force the future into any preconceived mold; rather it is wide open to whatever kind of political answers actually work at satisfying people’s needs, whether they be liberal, conservative, socialist, libertarian, nationalist, something new, or some unprincipled but practical mixture.

Any political thinker who is confident that his or her answers are the right answers should welcome ophelimist democracy as the chance to prove it. Any political thinker who is not confident about the best path forward should welcome ophelimist democracy as the fairest way of deciding based on the best information.

Ophelimo is a tool to change the system, a weapon to fight “Moloch“, a funnel to guide the future toward happier ends.