MikkW

Karma: 940

MikkW 23 Sep 2021 23:10 UTC
1 point
on: MikkW’s Shortform
When a republic is considering reforming the way it elects holders of power, there are certain desirable criteria the chosen method should have: officeholders should be agreeable to all citizens, not just to a fraction of them, since maximizing agreeableness will select for competence and the desire to do right by the people; in bodies with many members (e.g. a legislature, but not a singular executive such as a president or prime minister), the various view points of the voters should be proportionately represented by the various members of the body; and the system should be relatively simple, to allow the members of the republic to verify that things have been done properly, ensuring the system is trusted.

There are systems I have considered that suceed better at the first two criteria, but at the cost of being sufficiently complex to make me worry that practical issues could arise if implemented, and that it could be hard to convince enough people that they are better, which will be needed in order to get them made into law. The following is my current best hypothesis for an ideal system that meets these three criteria, particularly aiming to be simple both to actually use and to explain to people:

Voters are provisioned with a ballot with three fields: first a field to elect the representative of their single-member district, where they can either approve or disapprove of as many candidates as they wish (approval voting); second, a field where they may indicate a single republic-wide list of candidates that they feel best represents them; and third, a field where they may name as many candidates from their chosen list as they wish, to indicate their approval (party-list proportional representation with within-party approval voting).

The single-member districts will select roughly 2/3rds of the body, with the most approved candidate winning in each district. A “similarity score” will be calculated for each winner comparing them to each list: if half of the voters who approved of Charlie voted for the Aleph List, Charlie’s similarity score for Aleph will be 0.5; if instead only a tenth of his approvers voted for the list, the similarity score will be 0.1

For each list, the sum of that list’s similarity score with each candidate will be considered to be the initial number of effective representatives that list has in the body (note that this means the district candidates’ party affiliation does not factor in to this system, since this is based on voter behavior patterns). For each list, consider the ratio of number of votes for the list divided by the number of representatives for the list: whichever list has the most votes per representative is the most underrepresented list, so the most approved candidate from the list will be appointed, increasing the number of effective representatives by one. Keep adding seats to the least represented lists, until the remaining 1/3rd of the body has been appointed, thereby giving the entire composition of the body.

Note that the use of approval voting ensures district candidates will be incentivized to appeal to a broad base, encouraging them to find solutions that take care of everybody’s needs, instead of rewarding them for creating division within their community; while the leveling seats ensure that all voices, not just the most centeral perspectives, contribute to the government.

MikkW 22 Sep 2021 23:52 UTC
2 points
on: MikkW’s Shortform
A commonly given reason for why Nordic countries tend to rank highly as desirable places to live, is because the people there are supported by a robust welfare system. In America, I’ve often heard it said that similar systems shouldn’t be implemented, because they are government programs, and (the argument goes) government shouldn’t be trusted.

This suggests the government as a potentially important point of comparison between the Nordic countries and the US. Are there features that differ between the American and Nordic governments (keep in mind that there’s noticeable ethnic diversity even in the Nordic countries, and many of them are part of the EU, which provides a roughly similar large-scale government as the American federal government). Do the politicians face different incentive structures that allow more honest and more competent politicians to do better in the Nordic countries than in the US?

MikkW 22 Sep 2021 2:35 UTC
1 point
in reply to: gwern’s comment on: MikkW’s Shortform
Widdershins is from a different root according to Wiktionary. I was not aware of those before, but I do still prefer my made-up terms.

MikkW 22 Sep 2021 0:36 UTC
1 point
on: MikkW’s Shortform
For a long time, I found the words “clockwise” and “counterclockwise” confusing, because they are so similar to each other, and “counterclockwise” is a relatively long word at 4 syllables, much longer than similarly common words.

At some point in time, I took to calling them “dexter” and “winstar”, from the Latin »dexter« and Middle English »winstre«, meaning “right” and “left”, respectively. I like these words more than the usual “clockwise”, but of course, new words aren’t worth much of others don’t know them, so this is a PSA that these are words that I use in my vocabulary.

(As an addendum, I have also taken to calling the Danish political party «Venstre» as “Winstar”, because despite their name meaning “Left”—which was fitting at the time the party was founded—they are actually part of the right-of-centre coalition)

MikkW 20 Sep 2021 7:18 UTC
3 points
on: MikkW’s Shortform
There is probably overlap between the matter of aligning AI and the matter of aligning governments

MikkW 20 Sep 2021 1:36 UTC
1 point
in reply to: Pattern’s comment on: MikkW’s Shortform
Gotcha. My main explanation is just that the American political framework is old, having been around since the start of the modern democracy movement, and voting theory wasn’t a thing people thought about back then; that, plus the particular historical reasons many countries adopted proportional representation didn’t play out to the same degree in the US.

MikkW 19 Sep 2021 16:04 UTC
1 point
in reply to: MikkW’s comment on: MikkW’s Shortform
It occurs to me that this is basically Babble & Prune adapted to be a writing method. I like Babble & Prune.

MikkW 19 Sep 2021 4:15 UTC
1 point
in reply to: Pattern’s comment on: MikkW’s Shortform
Do not both the resources needed to run a government and the resources a government can receive in taxes grow linearly with the size of a country? Or do you have different size dynamics in mind?

MikkW 18 Sep 2021 7:48 UTC
1 point
in reply to: MikkW’s comment on: MikkW’s Shortform
This post was written in 5 blocks, and I wrote 4 (= 2^2) branches for each block, for 5*2 = 10 bits of curation, or 14.5 words per bit of curation.

As it happens, I always used the final branch for each block, so it was more effects of revision and consolidation than selection effects that contribute to the end result of this excercise.

MikkW 18 Sep 2021 7:43 UTC
5 points
on: MikkW’s Shortform
This Generative Ink post talks about curating GPT-3, creating a much better output than it normally would give, turning it from quite often terrible to usually pround and good. I’m testing out doing the same with this post, choosing one of many branches every few dozens of words.

For a 4x reduction in speed, I’m getting very nice returns on coherence and brevity. I can actually pretend like I’m not a terrible writer! Selection is a powerful force, but more importantly, continuing a thought in multiple ways forces you to actually make sure you’re saying things in a way that makes sense.

Editing my writing can be slow and tedious, but this excercise gives me a way to naturally write in a compact, higher-quality way, which is why I hope I do this many times in the future, and recommend you try doing this yourself.

MikkW 17 Sep 2021 7:44 UTC
3 points
on: MikkW’s Shortform
A personal anecdote which illustrates the difference between living in a place that uses choose-one voting (i.e. FPTP) to elect its representatives, and one that uses a form of proportional representation:

I was born as a citizen of both the United States and the Kingdom of Denmark, with one parent born in the US, and one born in Denmark. Since I was born in the States with Danish blood, my Danish citizenship was provisional until age 22, with a particular process being required to maintain my citizenship after that age to demonstrate sufficient connection to the country. This process is currently in its final stages.

I have had to (still am) deal with both the American Department of State and the Danish Ministry of Immigration and Integration in this process, and there’s a world of difference between my experience with the two governments. While the Danish Ministry always responds relatively promptly, with a helpful attitude, the American DoS has been slow and frustrating to work with.

I contacted the American Dept. of State in order to acquire a copy of certain documents that the Danish government needed to verify that I indeed qualify to maintain my citizenship. This request was first sent in October 2019, or 23 months (nearly 2 years!) ago, and I still have been unable to acquire a copy of this needed documentation, with no timeframe provided for when it might be available. The primary reason for this delay is precautionary measures taken in order to slow the spread of COVID-19 preventing anybody from entering the building where the records are kept.

But the Federal government didn’t even start to take precautions against Covid until March 2020, four months after I first sent the request! While such a delay isn’t surprising when dealing with the American government, there really is no excuse for such a long delay to start with. And even once the virus hit, I’m still left scratching my head.

Yeah, we want to take precautions to keep it from spreading. Yeah, it makes sense that in the first many weeks, hasty measures were taken that wouldn’t always make perfect sense. But you don’t close down a building containing vital records for a year and a half! Let a selected group of employees enter, and have them follow some set of protocols that ensures people are safe. The building is literally called the Office of Vital Records! You don’t close a building that’s capital-v Vital for such a long time just to contain the spread of Covid.

Meanwhile, everything that I needed from the Danish government, I received a reasonable and helpful response within a very quick timeframe. I don’t think I ever waited more than 10 days to receive help, and usually it was a good bit quicker than that, with my requests being responded to within the day after I sent any request.

So why is there such a big difference? Imagine that the slow, unhelpful government processes showed up overnight in Denmark, which uses proportional representation. It wouldn’t take long for some citizens to get frustrated, and for them to rally their friends to their cause. So far that’s not so different from what could happen in the US. But while the two major US parties would be so focused on a narrow set of polarized topics, with neither having any incentive to address this growing discontent among the populace (and both being able to get away with ignoring it, since the other party ignores it too), in Denmark, even if the two biggest parties ignored this discontent, the smaller parties would be able to use this discontent to win big in the next election, by promising to do something about the problem, winning them large votes away from the parties that ignore it.

This dynamic is what causes Danish government, not just in this aspect, but in almost every aspect I have seen during my time living there, to be so much more competent and pleasant to interact with than the American government: smaller parties in Denmark always make the larger parties work hard to maintain their lead, while in America, the two parties can compete on a few high-profile banner issues, and sit on their laurels and ignore everything else.

Another framing, is as a two-dimensional political spectrum. While the two-dimensional spectrum I’ve seen most often pairs “Right vs. Left” with “Authortarian vs. Liberal”, I think a more important grid would pair “Right vs. Left” with “Competent and Aligned” vs. “Incompetent and Unaligned”. For a political party, being competent and aligned takes time, energy, and money away from being able to campaign to win elections, so in the absence of sufficient pressure from voters to have aligned parties, the parties will drift towards being very incompetent and very unaligned.

Because Competence is generally orthogonal to Right vs. Left, in a two-party system the main forces from voters will be on the Right-Left axis, allowing the parties to drift towards incompetence more or less unchecked (if you doubt this, pick up an American newspaper from the past 5 years). However, in a multi-party system (I believe this also applies to Instant-runoff, despite my disdain for IRV), since there are multiple parties on both the left and the right, voters can reward competence without having to abandon their political tribe, pushing strongly against the drift towards incompetence, and instead ensuring a highly competent outcome.

(One last note: yes, I have considered whether the anecdote I give is a result of the culture or the more diverse ethnic makeup of the US compared to Denmark, and I am unconvinced by that hypothesis. Those effects, while real, are nearly trivial compared to the effects from the voting system. I have written too much here already, so I will not comment further on why this is the case)

MikkW 14 Sep 2021 23:43 UTC
1 point
on: MikkW’s Shortform
The following is a mantra I have decided to install, and will likely be the basis of my yearly theme for 2022:

Say no to everything, decuple down on what is most important

MikkW 14 Sep 2021 7:07 UTC
1 point
in reply to: Measure’s comment on: MikkW’s Shortform
Correct, yeah

MikkW 12 Sep 2021 6:44 UTC
1 point
on: MikkW’s Shortform
I was considering methods that are well-suited for electing groups of 3 or 5 candidates (for example, choosing regional representatives in a larger body, or appointing a small city coucil). I know of Single Transferrable Vote, which uses a ranked ballot; but I know ranked ballots in single-winner elections are inferior to score-based ballots, due to Arrow’s Theorem. This made me consider cardinal (i.e. score-based) multi-winner elections.

Wikipedia names “proportional approval voting” and “sequential proportional approval voting” as such methods, but both come up short- the first (PAV) is not computationally tractable for large enough races (it defines a score, and says to consider every single possible combination of candidates to find which one maximizes this score), while the second (SPAV) will always appoint the approval winner, even when a more proportional result can be acheived by appointing a different candidate instead.

I devised the following algorithm that runs in O(n), but will often produce a more broadly acceptable result than SPAV:

An algorithm for choosing a good winner in a multi-member form of approval voting
1. Start with an empty stack of candidates
2. Tally the approvals of each candidate, and push the most-approved candidate to the stack
3. Reweight the votes, such that if a voter supports n candidates currently in the stack, their vote is worth 1 / (n + 1) of their original weight
4. Push the most-approved candidate to the stack
5. Pop the earliest added candidate off of the stack (they are no longer in it)
6. Reweight and push a candidate onto the stack
7. Reweight and push again
8. Repeat 5 − 7 until the number of candidates on the stack equals the total number of seats
9. Calculate the satisfaction score of this stack as the sum of [1 + ¹⁄₂ + … 1/n where n is the number of candidates in the stack that the voter supports] over all voters. Record this stack as well as its satisfaction score.
10. Pop and push, then record the satisfaction score of the new stack
11. Repeat step 10 until the number of stacks considered is equal to the number of candidates
12. Return whichever of the considered stacks has the highest satisfaction score as the output of the election

MikkW 12 Sep 2021 5:39 UTC
1 point
on: MikkW’s Shortform
Recently here I have been mentioning the idea of using California as a nucleation point for encouraging electoral reform in the USA. Beyond state-level change, a potentially easier target than amending the US constitution is to change the ways that one or both of the major parties chooses its candidates, particularly in the presidential race. This can help address some of the scarier problems we’ve been seeing in national-level politics, without requiring all the effort and activation energy needed for constitutional change; but it will likely be harder than and downstream of state-level change.

MikkW 12 Sep 2021 5:25 UTC
1 point
in reply to: Measure’s comment on: MikkW’s Shortform
I think that used to also be the case in Denmark, that a vote for candidate within the party was also a vote for the party, but that was changed for the reasons I mentioned above to the current system where one can vote for a different party than the chosen candidate’s party.

MikkW 11 Sep 2021 22:06 UTC
3 points
in reply to: MikkW’s comment on: MikkW’s Shortform
One aspect that drives my curiousity in this matter, is to see how this information can be used to implement a better system in my home state, California (I know I mentioned I want to leave California and the States, but even if I do leave, laying the groundwork for a better system here will be a good thing for California itself, America as a whole, and even the entire course of the history of humanity, and I care deeply about that).
One difference that stands out to me is that Finland’s Parliament is much larger (at 200 members) than any of California’s representative bodies—The State Senate has only 40 members, the State Assembly is larger at 80 members, and our delegation to the US House is 52 members large. While the state-level bodies could be made bigger (and maybe even the State Senate could be abolished? It’s not clear to me that it has any real purpose), the delegation to the US House is fixed, at least on the scale of effort I’m focusing on; and getting real change to happen will require the support of the people, and the fewer things that have to be changed, the more easy that will be to get, so I’d rather not try to change the size of the state legislatures unless it’s really needed.
Adopting a system similar to Finland, while holding the size of the bodies constant, will require either much smaller regions than in Finland, which would introduce substantial distortions, or will require a big reduction in the number of electoral districts, which I worry will not be popular in California (while I personally suspect regional representation is overrated, particularly in the context of proportional representation, people are used to electing regional representatives, and reducing regional representation is a criticism / concern I have heard mentioned seriously by people who don’t support proportional representation).
That causes me to suspect that while Finland’s system works well over there, it would be better to focus on systems that work well with districts with 3 or 5 members, with 15% − 20% of the seats used as leveling seats (to level out the inevitable distortions introduced with such small districts), in the context of electoral reform in California.

MikkW 11 Sep 2021 21:46 UTC
1 point
in reply to: MikkW’s comment on: MikkW’s Shortform
One common point of feedback I received from my recent posts is that perhaps I’m a little too grounded in theoretical aspects, and not focused enough on what’s actually going on. As part of my plan to address this, I am digging in deeper into what the actual systems are; another path that will be worth taking to address this is to look deeper into the reality of the situations in the countries I am looking at, and try to illustrate why their systems are leading to better or worse outcomes (without denying cultural factors, of course; but 1) I have a better grasp on how to change constitutions than how to change cultures—in California, the former is actually quite straightforward as long as there’s public support, and 2) I suspect that culture is largely downstream of constitutions, with constitutions shaping incentives, and the incentives then shaping people’s beliefs and values; more aligned constitutions will ultimately lead to more aligned culture).

MikkW 11 Sep 2021 21:37 UTC
2 points
on: MikkW’s Shortform
How Parliamentary Elections Work in Finland
(These are my notes after skimming the Finnish Election Law)
For purposes of electing the Finnish Parliament, the country is divided up into 13 regions. Åland elects one representative, and the rest elect multiple (between 6 and 35) representatives. All representatives are elected through regions; there are no supra-regional representatives.
Candidates are generally grouped together into parties, joint lists, or electoral alliances. The distinction is not relevant to my notes here; in each, multiple candidates are named, without further distinction. Voters identify one candidate who they support; this vote will then be counted as a vote for the candidate’s list, in addition to influencing the ranking of the candidate within their list. (That is, a list’s vote count is the sum of the votes received by each candidate on the list)
I originally found the method of using the votes to determine the precise representatives a little bit confusing, so I will describe it in two different ways here: first, an intuitive way that I believe is equivalent to the prescribed manner, then I will detail the manner as it is actually described in the law.
The Intuitive Explanation
The number of representatives a list receives within a region will be proportional to the number of votes that list receives; so if Party A wins twice as many votes as Party B, Party A will receive twice the number of representatives. Within each list, the n candidates which received the most votes in the list will be appointed, where n is the number of representatives from that list.
The Prescribed Manner
Candidates will be ranked within their list according to the number of votes they receive. They will then receive a score, which is (Total List Votes / Rank within list); So the 3rd place candidate in a list that received 60,000 votes will get a score of 60,000 / 3 = 20,000.
All candidates in the region will then be listed in a single roll, with spot #1 belonging to the candidate with the highest score, and continuing accordingly. The first n candidates will be appointed as representatives, where n is the number of representatives given to that region.
It is left as an exercise to the reader that the two descriptions I gave match.
Commentary
I will contrast this to Denmark, which is also a Nordic country, has a roughly similar system, and similarly ranks highly on various rankings of countries, including the World Happiness Report. While I haven’t read the Danish election law yet, I am roughly familiar with that system, and have strong ties to the country, and lived there for a year and a half.
One noticeable difference is that there are only regional representatives in Finland, with no representatives that balance things out on a national level. I don’t think this poses a big problem, maybe resulting in a distortion resulting in a couple percent one way or the other. I actually find myself confused as to why leveling seats are needed in Denmark (which also uses regional party lists), which I may resolve by actually reading through the Danish election law.
Another difference (in theory, not in practice) is that Denmark affords parties the choice to be closed-list, meaning that a party could choose the ranking of candidates instead of letting voters decide. In practice, I don’t think any parties actually do that, so it’s not a real difference (though, this can be contrasted with Israel, which uses closed lists exclusively. I cannot currently demonstrate this, but I suspect this explains part—not all—of Israel’s current political troubles).
I believe in Denmark, voters may indicate a different party to support than the candidate they name. There are gaps in my knowledge as to exactly how this is operationalized, but the idea is to allow voters to support their preferred candidate, even when they think weird things are going on in that candidate’s party at large.
A final difference is that there is a threshold of 2.5% of the vote required in Denmark for a party to receive seats (I’m not sure if this is on both the regional and national levels, or only the national level; but I think I’ve heard that it’s specific to the national level, in which case it’s not actually that different to the Finnish system), whereas in Finland, there is no such threshold.

MikkW 8 Sep 2021 19:20 UTC
1 point
on: Gradient descent is not just more efficient genetic algorithms
I suspect the thesis is true and can be valuable to appreciate, but I’m left feeling that this doesn’t explore the thesis nearly as much as I want. You give one example where there might be a divergence, but I would be more satisfied if you provided more examples where they diverge. I also found myself wanting to read more thoughts on cases where the difference can impact things in a way that I would care about. Where would descent be more useful, and when would there be benefits from using evolution? I think there are examples of both, but this post doesn’t touch on those.