How Electability Frustrates Democracy

This site might be interested in an article I’ve recently published at OpEd News. It specifically addresses the virtues of evaluative voting, an important property of approval voting and score voting. The article is just one of a series of more than forty articles on voting methods.

The best argument against democracy is a five-minute conversation with the average voter

This problem is undeniable, and is exploited with demagoguery.
To solve it, rules are needed to determine who to exclude from the vote (through a “voting license”), but these rules must satisfy some essential and non-trivial conditions: useful (applicable to all political areas), wisdom (exclude ignorance / stupidity) , not exploitable, simple (to understand and accept).
Rather than having rules that do not fully meet these conditions, it is better to let everyone vote, because it also applies to:

The best argument against dictatorship is a five-minute conversation with the dictator (or just watch what he does)

Democracy is the worst form of government, except for all the others.

Similar to what I always say: “the less worse, is the best”.

a democracy would have to, on occasion, take the trouble of determining the public will

In fact, the best democracy is not representative democracy but liquid democracy (if you want to satisfy this point).

We argued instead for evaluative systems of voting that ask voters simply to evaluate each of the candidates separately … regardless of electability

Voters will never evaluate the candidates separately, that is, if they have a range [0,10] then they will put the preferred candidate at 10, and the worst candidate at 0, and the others in the middle by comparing them with 10 and 0 (also applies to AV).
However, I’m extremely in favor of independence from electability, in fact DV is much more independent from electability than SV or AV.


If BAV is AV with negative ratings, then look at this example:
Candidate order (honest interests): A,B,C

  • If the most likely winner are A and B, then the vote will be: A[1] B[-1] C[-1].
  • If the most likely winner are B and C, then the vote will be: A[1] B[1] C[-1].

The votes (candidate B) depend very much on the eligibility of the candidates (i.e. it depends a lot on how likely it is that certain candidates win).

Negative votes are bad in dealing with unknown (or little known) alternatives.
Eg candidates A,B,C with little known C (mainly by those who support it):
45: A [1], B [-1], C [0]
45: A [-1], B [1], C [0]
3: A [1], B [-1], C [-1]
7: A [-1], B [-1], C [1]
C wins even if it is appreciated only by 7 people.
Now, hypothesize that in the same example, the 45 people who support A (or B) think "I don’t know C but I know that B sucks a lot, so in doubt I put 1 also to C); this way of thinking could happen sometimes, making C even more advantaged (despite being very unknown overall).

If the unknown candidates were favored by the voting method, the stronger political factions would create on purpose “many” little-known puppet candidates, with the hope of winning in this way (tactic exploitable even by Nazi factions, wanting to extremes).

Why do you use so many non-standard terms like “ration voting” and “evaluative voting”?.

“Balanced-Approval Voting” does not even have an electowiki or wikipedia page i can find.

Anyway, neither of these balanced systems are good systems.


Sometimes I have to invent a term for some concept that I invent. Perhaps this is an artifact of my mathematics background, but I think it is pretty common practice in many other fields as well. It just makes it easier to talk about a new concept if it has a name.

When I invent a new name I do try to make clear its meaning and in later articles I try to review the meaning of the term or at least provide a link to where I did define it. I suggest you take a look at the links in the article or perhaps review the list of my articles where you may find the term in the article’s title.

Specifically, you can find balanced ration voting and evaluative voting described at their respective links. The second of these terms is not my own invention, but a term I have seen used in popular literature.

Sorry, which of these systems do you claim invention of? It is pretty hard to invent a new single winner system.

Yes, “evaluative voting” is used but mostly in articles written in France. Pretty much everybody calls it Score voting or Range Voting if the choice is a continuous spectrum.

I think what you call ration voting is much more commonly known as cumulative voting.

1 Like

The assumption you are making is that voters always do have a strict linear ordering of preferences. This may occasionally be the case, but with several candidates it seems unlikely to happen often. Granted, this assumption has a long history, with voters being modeled as simply such a strict linear order of preferences, but that is little justification for continuing to make the assumption.

More likely, in an election with say 10 candidates, a typical voter may find two or three as good choices, perhaps another two or three, maybe even four as unacceptable and the remainder as neither particularly good or bad, perhaps even including one or two who are unfamiliar to the voter. If forced to formulate a strict linear order, the voter will be forced into a number of arbitrary decisions, perhaps based on strategic considerations or perhaps based on a mental flip of the coin.

His arguments seem to work even if you allow for equal-ranking (i.e. the voter might have multiple 1st choices, a couple 2nd choices, etc.)

I don’t take a linear order (I support ranges not rankings). In the example I gave, I indicated the order that the voter would use if he was “forced” to choose between couples, which is the best candidate.
Furthermore, I assume that if a candidate knows the likely winners A and B, then at least between A and B he will know how to find one better than the other (thus generating the problem I described in the previous comment). Sorting doesn’t have to be done across all preferences.
If you say to a voter “by assigning 1 to both A and B, your vote becomes completely useless (since A,B are the likely winners)” you can be sure that the voter will find an order at least between A and B.

You say that I have “little justification for continuing to make the assumption” but you also make an assumption that doesn’t seem very justified to me.
The fact remains that, if I vote using range [-5,5], the voters will be able to decide whether to use your assumption (voting only with values ​​-5,0,+5) or whether to use my assumption with sorting (also using the other values ​​in the range).
SV can be used as a BAV (if the voters really want to, as you assume), while BAV cannot be used as a SV, so SV is better (offers more freedom to the voter). This is a fact, not an assumption.
In SV, using range [-5,+5] or [0,10] is the same thing, at the procedural level.
And I support ranges [0,10] as a form of representing interests, but I don’t support SV as a count (because SV is also very subject to the problem described in my previous comment).

As Keith_Edmonds notes, range voting and score voting are two terms used for the same thing. It is not uncommon for a concept to go by more than one name.

When your only example assumes that a voter’s preference is A > B > C, that seemed to me to be assuming a strict linear order.

One thing that have come to believe is that human psychology has a lot to do with voting and that should be a consideration when it comes to voting methods. Unfortunately, this does force us into the realm of speculating about how people will act and what will guide their thinking. So I will grant you that I am speculating, but let me just do some speculating about how people think.

  1. People tend to view support and opposition as two separate emotions. They do not generally consider “support at the lowest level” to be the same as "opposition at the highest level.

  2. People at large generally are not well-versed in mathematical arguments, so voters tend to take instructions at face value. Some for example are intimidated by being asked which of two candidates they like best when they do not like either of them at all. But when voters are asked simple questions that seem to them to have easy answers, most will not go out of their way to look for reasons to avoid giving those simple answers.

  3. Different people make different judgements about assigning numerical values. Even when asked the simple question of whether or not to support a candidate (as with AV), a voter may have difficulty deciding on a given candidate; where to draw the line will be the difficulty. Having an even wider range of options only magnifies that difficulty by introducing even more uncertain boundaries.

Let me now turn to the tutorial on SV and say a bit about that system as it is described. Of course I would prefer just three levels instead of six (based on 3. above), but I don’t consider that a particularly important issue. Based on 1. above, I would certainly change the label “Zero Support” to read “Maximum Opposition”. It would help promote understanding (in view of 2.) on the part of the voter to change the numbering to -3, -2, -1, 1, 2, 3.

But this raises the question of what to do about the voter who fails to provide any score for some candidate, say Chris? Would it be appropriate to treat that as -5, maximum opposition? I doubt anyone would agree with that. Perhaps we would count it as the middle value 0 - thus giving the voter one additional value for scoring. The result would be a balanced system.

I’ll have to give some more thought to the instant-runoff aspect of SV. When I first heard of the notion I was reminded of the old adage that a camel is a horse designed by a committee, but the idea does seem appealing in some respects. It may encourage some strategic voting, however, as it places some value in manipulating who comes in second.

But we can’t have people constantly renaming stuff that has existed for a long time. Cumulative voting is used in california under that name. If you reinvent it and give it a new name you are not helping anybody.

Range and score are technically different, too.

Wikipedia seems to be saying that Score Voting and Range Voting are exactly the same. You should probably submit a correction to their article.

I came up with the idea of Balanced Ration Voting in 2016 and I did not learn of the term “cumulative voting” until this week. In any event, in 2016 I was more interested in exploring the concept than in researching whether there was another term in use for the similar concept. I just did a search for the term and found that it seems to be used primarily for elections with multiple winners.

It would be wonderful if there were a single word that accurately describes each thing we want to name. I don’t expect that to ever happen, but good luck with that project.

I was using “evaluative voting” for a broader notion than just range voting or score voting. As I used the term it includes any voting system in which the voter is not faced with choosing one, or in fact any fixed number of candidates; rather the voter is asked simply to evaluate those candidates that the voter feels comfortable evaluating.

I agree with practically everything but I would add that, although it’s true that with regards to “negative” candidates it’s difficult to establish who is the worst, among the “positive” candidates the voter generally has clear ideas.
If you ask me “which food do you prefer”, I don’t waste time evaluating foods that I don’t like (I just discard them), but the ones I like instead have a very specific order.
In the DV I solved this problem, that is: you can put a “dislike” (0 points) or a number of “like” between 1 and 10 to a candidate. This I think is the best way.

But this raises the question of what to do about the voter who fails to provide any score for some candidate, say Chris?

This is exactly the criticism I have often made to the SV. In DV the unknown candidate (just like the rejected ones) receives 0 points (doesn’t receive any likes), so he cannot win.

Wikipedia is full of inaccuracies on many topics. It tends to be pretty good for facts but not for technical stuff. I am not really interested in fixing wikipedia. We built to get around the issue.

When people first get into this stuff they often reinvent systems. I thought I invented approval voting when I came up with it but it turned out to be 1000s of years old.

Given that there is 100 years of research you should really read some of it. Exploring a concept on your own is not really a common path to scientific discovery.

So Cardinal voting. People do use the word “score” for this but it implies an aggregation method. Score is explicitly a single member sum of discrete cardinal vote system. You are talking about classifying based on the ballot so you want the term Cardinal. You see how difficult it is to follow your arguments when you use your own semantics? As with using score for cardinal, nobody is super strict about semantics but if you come in with totally new semantics it is going to do nothing but cause confusion.

Its not really a project. I have been doing this for many years and you are the first person to come along and claim to have invented a whole set of standard systems by simply taking old systems and renaming them. I do not see how this is a good strategy. Nobody in the community will understand what you are talking about.

What would be more productive would be to use the semantics of the people who existed prior to you getting involved and talk about what system you endorse. The systems you propose have been shown to be too flawed for consideration ages ago. It is hard to understand what you are advocating exactly but I think you are proposing a score [-1,0,1] system. If this is the case it was pointed out above why this is not a good idea. There was a recent paper with actual voting data proving why this is bad.