[EM] language/framing quibble

[Kristofer Munsterhjelm]

Fred Gohlke wrote:
> Good Afternoon, Kristofer
> 
> I'd like to start by thanking you for your analytical comments.  Your 
> forbearance on the points where my understanding lags is gratifying, for 
> I fear your level of education exceeds mine by a considerable margin. 
> Although I lack formal education, I don't lack the capacity for 
> independent thought.  If you wish to know a bit more about me, you'll 
> find a brief profile at:
> 
> http://whither-democracy.blogspot.com/

You're welcome. Actually, I don't have much of a formal education as 
regards election methods or societal dynamics, but I seem to naturally 
think in an analytical manner. That, the internet, and curiosity goes a 
long way.

Incidentally, why are your web log entries timestamped 2010?

> re: "There are two ways to regard the results of an election
>      method. The first is how accurate the representatives
>      display the properties or opinions of the people, and the
>      second is how good a job the representatives do."
> 
> In this, you describe a distinction I hadn't considered ...
> 
> I focus on how good a job our representatives do rather than how well 
> they reflect the opinions of the people, because, while I stand second 
> to no-one in my commitment to our right to govern ourselves, I'm not 
> blind to the fact that we can be manipulated.
> 
> We may, at some point, need to look at why we are susceptible to 
> manipulation, but, at the moment, I'll only say if we are to govern 
> ourselves, we must offset that susceptibility with leaders who have the 
> ability to guide us past our own shortcomings and the integrity to do so 
> to our common advantage.  We will only improve our society when we find 
> leaders who can transcend the vagaries of opinion without betraying the 
> public interest.

I can see the point you're making, but I think you should be careful not 
to go to the other extreme, too. Opinions may shift, but at the bottom 
of things, they're the people's priorities of in what direction to take 
society. The vagaries you speak of could be considered noise, and that 
noise is being artificially increased by the two main parties, since if 
they can convince their wing voters they represent their opinion (or 
change their opinion), then those voters are more likely to vote for 
them instead of not voting at all. That doesn't mean that there's no 
signal, though, and where that signal does exist, it should not be 
averaged out of existence or amplified in some areas and attenuated in 
others (as could happen if the majority of the majority is not equally 
much a majority of the whole).

> re: "... because politics is a process, if a method has biases,
>      then these may amplify in ways that can cause odd results
>      over time."
> 
> I had not understood that point and I'm glad you raise it.  Would it be 
> correct to imagine there is a difference between a systemic bias (as in 
> the plurality system) and the biases of ideology (as with Liberals and 
> Conservatives)?  I believe your reference is to the former.

Yes, it is. To be more specific, the bias I'm talking about is a 
distortion of the wishes of the people. A method that fails mutual 
majority might pick a candidate where a majority prefers one of a set 
that candidate isn't in, for instance; and more concretely, Plurality 
squeezes out the center and provides incentives for two-party rule.

> It is my belief (until I learn otherwise) that the process I've outlined 
> has a bias in favor of the most fundamental goals of society.  Humans 
> have a natural tendency to pursue their own interest, and the choices 
> they make will reflect this.  Time and repetition will refine the 
> choices from the extremes enunciated by the least thoughtful of us to 
> their essence as enunciated by the most thoughtful.

That means that your system acts less quickly to change. This is a 
method that can work, and it prevents the kind of spurious opinion 
oscillation you'd see in Plurality, but again, if it goes too far, it's 
too slow. I don't know if it is too quick, just right, or too slow, but 
I think it's more likely that it's one of the two latter than the 
former, especially given the observation that majority agreements turn 
into consensus at the higher levels.

> re: "Even if the country was nearly filled with paragons of
>      virtue, partisanship would happen nevertheless, simply
>      because of the long term feedback of the system."
> 
> Although I don't think you mean otherwise, partisanship is natural for 
> humans.  It exists independently of our political systems.  Politicians 
> exploit that tendency to our disadvantage.  They trade on it to attain 
> power.
> 
> An important consideration for a sound electoral method is the ability 
> to function independent of partisanship.  We all have partisan feelings. 
>  Indeed, they are a vital part of society, but they must always be a 
> voice and never a power.  The danger is not in partisanship, it is in 
> allowing partisans to control government.

What I meant was that even if the majority were (by some miracle) 
nonpartisan, parties would form out of necessity. Plurality simply can't 
support a horde of independents. As such, Plurality encourages the 
formation of parties, and of parties to coalesce until there are two 
main blocks -- even in the best case (where near-nobody is partisan), 
the nature of Plurality, the method itself, shapes the results, meaning 
that it definitely does so under less ideal conditions, such as in the 
world today.

> re: "... I try to find situations where your method loses
>      information, and see whether they could cause long-term
>      problems."
> 
> We know systems can be corrupted.  If we find flaws in this one now, we 
> have an opportunity to correct them.  I've spent many hours trying to 
> imagine weaknesses in the method, but that is a matter best investigated 
> by others.  I, like all humans, am constrained by my own blinders.

Yes, and if we find really bad flaws in it, then we can say "no, this 
can't be salvaged, let's try something else", without actually having 
society be hurt by our "experiment".

My fixes to the system would be to have a somewhat larger council size 
and use a PR method to pick more than one representative/delegate to the 
next level. This weakens your aim, which is to retain the experienced 
who can convince others, but hopefully not too much, and it also 
alleviates what I see as the problem, that legitimate shades of the 
people's idea of how society should be run would otherwise be excluded.

The exact size of the council would have to be found out by either 
trying, or by reasoning. I understand the reason for picking three, as 
you gave in your earlier post, so it's likely that inreasing the council 
size would make it less of a discussion and more formal, which we don't 
want. What we'd need would be to understand how quickly the council 
degrades as its size increases, in comparison to the gains elsewhere (in 
accuracy and in agreement).

> re: "... the intuitive idea is that if a group of three (or
>      some other number) agrees upon a compromise, then by
>      necessity, that compromise can't contain all possible
>      opinions of the group.  If this happens in many places,
>      the compromise of compromises will differ from the
>      compromise of all, had they been somehow able to elect
>      directly, simply because the minorities add up to a sizable
>      fraction that could have influenced matters if they had been
>      more lucky."
> 
> I understand your point.  I would agree with it ... if humans were stick 
> figures.  Since we are not, I think it misleading to attribute monomania 
> to all humans.  It is true there are among us people so obsessed with 
> the rectitude of a belief they will sacrifice themselves and others to 
> assert their point of view.  However, such individuals are but a tiny 
> part of the human race.  Most humans are amenable to reason, but they 
> must have an environment in which reason can thrive.
> 
> Taken as a group, members of a minority can appear monolithic, but when 
> you examine the individuals that make up the group, you find endless 
> variations in their belief in and commitment to the group.  You also 
> find they each embody parts of many other groups, some of which are 
> antithetical to each other.  There is infinite variety in their views. 
> Rare indeed, is the individual that does not favor some minority.  I 
> fear that excessive focus on minorities will distort one's view of 
> humanity.
> 
> In the case of three people selecting a representative, or, perhaps even 
> better, a spokesperson, the selection is not a compromise, it is a 
> conscious decision that one individual's advocacy of some dynamically 
> defined group of topics is superior to that of the others.  It is 
> correct to say the person selected can not contain (represent) all the 
> possible opinions of the group ... but that is an advantage rather than 
> a disadvantage.  Taken over a broad spectrum ... say the 3000 triads at 
> the initial level in the example ... the cumulative effect is a 
> conscious effort by all participants to make the best selection 
> possible, under the dominant biases of the time.

It's true that all methods must exhibit some distortion; the proof is 
simply the one I explained earlier, that if there's a council of size 
200 then the only way for someone who holds a single opinion to hold one 
  that less than 0.5% of the people holds if there's some other opinion 
than more than 0.5% holds that's not represented. Thus, to the extent 
that we want the representatives to have the full picture, the question 
is, compared to other methods, is the distortion/quantization 
particularly bad, or is it better than other methods?

That proof is, of course, a simplification which assumes that humans are 
stick figures. That isn't true in reality, as you point out. However, 
there is still a limit to how wide a span a single representative can 
hold - how many different solutions he can contemplate and argue in 
favor of - so the method (and any method) will still exhibit a 
quantization of the ideas of the people, and the same question returns; 
is it worse or better than other methods? (And also, is the quantization 
biased so that the method may give feedback like the two-party 
entrenchment of Plurality?)

When we model humans as stick figures, these questions can be, after 
some fashion, be answered by simulations. When we try to be more 
accurate, in that people are not, it becomes more difficult. There is 
both the objective neutrality ("judge manner", or the degree to which 
one is not corrupted, selfish, etc), and subjective positions (how many 
opinions? how wide a span can a single representative hold?).

To return to a more concrete response: if the councilmembers can hold 
many opinions, or a range of opinions, and deliberate among those, the 
effect of exclusion is significantly reduced, but it'll still be there, 
and it may or may not still exhibit the "shaving off significant, but 
thinly spread, areas of opinion" effect, only with ranges of opinion 
taking the place of stick-man type "either you're with us or without us" 
opinions. I don't know whether it would, since it'd depend not only on 
the system, but also on the integrity of the councilmembers.

> If I may presume to suggest an instance which (I think) bears out the 
> circumstance you describe, we can reasonably imagine most humans feel we 
> should avoid the mindless slaughter of war.  Yet, the chances of 
> selecting a candidate absolutely committed to opposing war is very small 
> because we also recognize the governments of our era are imperfect.
> 
> Even so, we can anticipate that when people are not obliged to choose 
> between 'hawks' and 'doves' they will choose candidates with a 
> preference for avoiding armed conflict; people who will work, within the 
> realities of their time, to eliminate war as an element of policy. Thus, 
> what is declared (by some) to be a minority public opinion (i.e., 
> anti-war sentiment) gains influence, not by attracting additional 
> adherents, but by the ascendancy of reason over partisanship.

Right, I see how the method would ideally work. The most hawkish would 
be pulled towards the center by those who are more dovish than them, and 
the same for the extreme doves. This would happen either by conviction 
(the representative/councilmember elected being convinced it's better 
not to be so extreme) or by replacement (if he doesn't); and would be 
augmented by the method picking objectively skilled members who'd be 
more likely to accept new ideas.

Let's see what ways it could fail.

The councilmember could lie his way to the top. This could be softened 
by recall; in some proposed council democracies, the councils are 
permanent and a majority at level (n-1) can recall a council at level n, 
but even with only the population to have recall at the end of the 
process, it would weaken the incentive to lie.
The closest example to councils of councils, that I know of, with the 
attendant objection that councilmembers are not accountable to the 
people because they can't be revoked, is the Commission of the European 
Union. The Commission is made up of members elected by the state 
governments, which are in turn elected by the people; or even with one 
additional link of indirection, through a parliament. Now, the analogy 
might be weak, since the EC is limited by the effects of the systems 
used to elect the intermediate steps, and possibly also by the 
partisanship incentives you've talked about earlier; and the "councils" 
are much larger than your triads as well; but that's the closest we 
have. If the EC is a good example, then accountability is going to be a 
problem.

The hawkish (say) could have fortuitious geographical distribution and 
also be unusually zealous. My simulations seem to show this is not as 
much of a problem as I thought, at least not with a one-dimensional 
issue of war vs peace. What happens if a triad is unable to agree upon a 
candidate? Would it then be done by vote, random selection within the 
council, or is that councilmember's position undetermined and filled 
randomly at the next level?

> re: "The second is that if there are multiple opinions (not just
>      yes and no), then the triad can't hold an opinion that's
>      held strongly by less than a third of the people."
> 
> As described in the response to the first point, neither humans nor 
> their opinions are so monochromatic.  People often favor opinions that 
> are not currently popular.  When they do, they work for their 
> implementation indirectly even though direct advocacy would be stymied. 
>  To assert that people are defined by the opinions attributed to them is 
> putting the cart before the horse.  Speaking of horses, I'll stop 
> beating this one ... it may already be dead.

I'll try to sum it up so we can see how we think here. My concern is 
that first, minorities can get shaved off in unpredictable ways, and 
second, that because of the small size, each triad may not have all 
information required to decide which is the more accurate with respect 
to what the people thinks. The first is particular to the triad system, 
the second would be the case for all systems with small bodies. You're 
saying that the triad councilmembers would make up for it by being 
unusually good at what they do; that they would have many opinions and 
not stick fanatically to any one of them, because the process would 
encourage those who were objectively skilled enough to do this.

If that's right, then the difference here is a focus on whether 
councilmembers accurately represent the people, versus how good each 
councilmember is at his profession-as-it-were.

To the degree I'm biased, I'm biased towards the first because it 
provides a worst case limit: even if you end up with bad councilmembers, 
you end up with a cross-section of the people, which is better than a 
gathering of bad majoritarian councilmembers, excluding all the others. 
Also, I've seen how single-member plurality works, and multiple-member 
methods (like STV) seem to produce much better results in practice.
The similarity would be that one could argue towards single-member 
plurality in the way you do, that the important thing is not to pick the 
partisan member, but the one who'd do his work the best. Then again, 
Plurality is flawed even as a single-winner method, so single-member 
districts under Plurality may not give a good idea of how far you can 
get with the single-member idea; Condorcet may be better.

There is also the amplification effect I've talked about earlier, on how 
the errors of the system could provide an unwanted dynamic (like 
two-party domination in Plurality), but yes... the horse is dead now, 
and I was just going to show how I reasoned.

> re: "... if the legislature can recall the executive, having a
>      majoritarian executive is of less problem than having an
>      unrepresentative legislature."
> 
> This topic is mentioned but not expanded in the second footnote to the 
> Sefton petition:
> 
>     "#2: The process is inherently bi-directional.  Questions on
>      specific issues can easily be transmitted directly to and
>      from the electors for the guidance or instruction of elected
>      officials."
> 
> The point was not expanded, in part, because it is not included in the 
> petition.  Another reason for not expanding it is that it is 
> implementation-dependent.  The natural bi-directionality of the process 
> allows many enhancements, not least of which is recall.  It also opens 
> the door to substantial digressions from the primary focus of the 
> process ... to harness our natures to the electoral process.  We can 
> anticipate that this topic can, and will, lead to considerable 
> dissension between those who feel representatives must be under the 
> direct control of the people and those who feel we elect people of 
> judgment and should (within limits) trust their judgment.

To briefly repeat what I've said, I think bidirectionality is going to 
be particularly important here, simply because the method contains 
multiple elections (one for each level), not just one, so the 
bidirectionality is not just from the final to the people, but to all 
the other levels as well. Thus, if there's a dilution of responsibility 
that must be compensated for by bidirectionality, that dilution happens 
multiple times, and so the compensation has to be all the stronger.

> re: "... say that there's only one political axis (left-wing to
>      right-wing). The two parties occupy positions some distance
>      away from center, say at 0.20 and 0.75. Then the compromise
>      of the left-wing party will be at 0.20, and that of the
>      right-wing party will be at 0.75; these are both more
>      extremely placed than the true compromise at 0.5."
> 
> I think I understand.  I took the .20 and .75 values to mean a position 
> within a wing rather than a position on the axis, but that line of 
> thought led me astray.  In an effort to enhance my understanding, I 
> tried to create a diagram but that didn't work for me, either, so I 
> tried creating a diagram of my own.
> 
> Is this representation, where each symbol (l,m,r) indicates 2% of the 
> electorate, close?  The idea I'm trying to convey is that the (l)eft and 
> (r)ight portions of the electorate are approximately equal at 40% each 
> and there's a middle segment, committed to neither side, of 
> approximately 20%.  The candidates for each side are represented as 
> being 20% (L) and 25% (R) from the least radical end of each wing:
> 
>                      L                  R
>      llllllllllllllllllllmmmmm mmmmmrrrrrrrrrrrrrrrrrrrr
>           -01%/-40%       -10%/+10%      +11%/+50%
> 
> If "L" wins:
>                      L (-18)
>      llllllllllllllllllllmmmmm mmmmmrrrrrrrrrrrrrrrrrrrr
>           -01%/-40%       -10%/+10%      +11%/+50%
> 
> If "R" wins:
>                                         R (+20)
>      llllllllllllllllllllmmmmm mmmmmrrrrrrrrrrrrrrrrrrrr
>           -01%/-40%       -10%/+10%      +11%/+50%
> 
> If that is a reasonable depiction of what you mean, even the most 
> cursory examination should show how unappealing it is, from the 
> perspective of the entire electorate.  Obviously, the more either 
> candidate moves toward its polar extreme (and we've often seen that in 
> the U.S.), the more offensive the concept is to reason. Nor is the 
> offensiveness ameliorated by the fact that a majority of the electorate 
> chose the final winner because their choices were limited to two actors 
> and were based on a multiplicity of emotional triggers inspired by 
> professional opinion-makers.

That is for the most part an accurate depiction. However, I considered 
the values to be absolute. Thus, for 2% per, you'd get

0.0  0.1  0.2  0.3  0.4  0.5  0.6  0.7  0.8  0.9  1.0
  #----#----#----#----#----#----#----#----#----#----#
            L              |            R
  llllllllllllllllllllmmmmm|mmmmmrrrrrrrrrrrrrrrrrrrr
                           |
          -01%/-40%   [-10%/+10%]    +11%/50%

This doesn't change much about what you said, though, it just makes the 
right-wing party candidate slightly more palatable than the left-wing 
candidate, since the right-wing party contains more of the 
close-to-middle voters than do the left-wing party. Presumably, the 
right-wing party candidate would win in a two-party Plurality election.

In any event, this is a sufficiently reasonable depiction of what 
happens in primaries. It's idealized and reality may be somewhat 
different, since parties may know this and strategize, picking a 
candidate closer to the middle. On the other hand, their internal 
primary election method might distort results, particularly for US 
parties, since they use Plurality.

The relevance of this problem as regards the council democracy / triad 
system is that the "l" and "r" voters are majorities, but neither L (the 
majority choice of the l-majority) nor R (the majority choice of the 
r-majority) is a good candidate. This shows how the true center may be 
eliminated for opinion ranges (not binary opinions) in a kind of 
real-valued variant of vote-splitting.

If the council deliberation works similar to Condorcet, the effect will 
be weakened, since a Condorcet election with a middle candidate inserted 
above (at 0.5) would elect the middle candidate:
   40: L > M > R    (left-wing group)
   10: M > R > L    (middle group)
   40: R > M > L    (right-wing group)
and M is the Condorcet Winner.

The effect might happen between councils, though, even if they don't 
happen (or happen only weakly) inside councils.

> re: "... one thing I observed from the simulations is that the
>      higher level councils often have not just a majority, but
>      consensus. Do you think this could lead the councilmembers
>      to consider their own positions to be held by more than is
>      actually the case?"
> 
> I suspect that is a virtual certainty, for projection is another natural 
> human trait (albeit one I did not consider and I'm deeply impressed that 
> your simulations revealed it.)  Since I had never considered this aspect 
> of the process and am an inordinately slow thinker, it will take me a 
> long time to fully appreciate the significance of this point.  At first 
> blush, I lean toward the opinion that it's favorable:  Those who would 
> advocate positions must have confidence in their rectitude.  Part of 
> that confidence will come from the belief that they speak for many 
> like-minded people.  If this leads to excess, it will affect subsequent 
> elections (or invoke some action via the bi-directional nature of the 
> selection process).

Confidence in their position is a good thing, as long as that projection 
does not lead from confidence to overconfidence and detachment from 
those groups of the people that do not share the councils' positions. 
Since a majority is transformed into near-consensus as the 
candidates/councilmembers bubble up through the levels, the effect would 
probably be more severe the further up the levels you got, and the more 
levels there are in general.

> re: "... you could either readjust the election method, or
>      compensate by some other means; I suppose it's the latter
>      that you're referring to when you say that we'd need to
>      change the way we maintain our laws."
> 
> Yes.  Although I don't want to digress unduly, I think it ludicrous that 
> a provision enacted by a tiny majority of a legislature should become 
> the law of the land in perpetuity. Laws should have a life defined by 
> the extent of their public support.  Oversimplified, and with a 
> percentage used for example only, a law enacted by a majority of less 
> than 52% should have to be re-enacted after one year.  Living under a 
> law is different than discussing it in preparation for enactment. 
> Therefore, laws should require annual re-enactment until they attract 
> enough support to raise them to a higher bracket with a longer life or 
> fail of adoption.

I don't want to digress too much, either, but let's just say that I've 
been considering similar ideas, myself; such as laws having a sunset 
that depends on how great a majority passed it, or on a president having 
a variable-time term depending on his victory margin.

Another, for presidential elections, would be one where it's possible to 
do a recall at any time. Then, at the start, you need a "yes" vote of at 
least 3/2 of the fraction that voted for the president, but where this 
margin gradually decreases.

For instance, you could have a decrease of 0.78 parts of a percent per 
month. So a president elected by a fraction of 51% would require a 
recall majority of 76.5% at day zero, 76.5 - 0.78 = 75.72% after one 
month.. and 76.5 - (0.78 * 3 * 12) = 48.42% after three years. The same 
kind of "gradual mounting opposition" could be applied to laws as well. 
There should probably be some sort of quorum (absolute limit) so that 
one can't do an unnoticed recall.

> re:  "I'm also saying that if 52.5% of the population holds a
>       given opinion, 4.1% of the time, a majority of the final
>       triad members hold the opposing opinion (a "majority
>       flip"). That's not so bad, and it actually surprised me"
> 
> I, too, am surprised ... but surprised it is possible to devise 
> simulations of such precision.  Although I never thought about the 
> possibility of a "majority flip", it doesn't surprise me that they 
> occur.  Indeed, part of my rationale is that, given time and an 
> environment in which to reflect, thoughtful people will learn (i.e., 
> change their opinions).  (In an attempt to interject a bit of humor, 
> I'll tell you one of my favorite sayings is that I change my opinions 
> more frequently than I change my socks.  All it takes is clearer 
> understanding.)

That was a stick-man model, therefore majority flips would be a bad 
thing. I expected the majority flips to be much more frequent, though.

> re: "The second representativeness issue would be similar to
>      there being four traffic solutions; one has to remain
>      unrepresented unless some can argue for more than one
>      solution."
> 
> There is little in life I find more stimulating than reminders that 
> there are valid views other than my own ... and they're the most fun 
> when they are described simply and clearly so I don't have to work for 
> the insight.  Thanks for this one.
> 
> 
> If I've failed to address anything of note, please let me know.
Same thing goes here, but I don't think you did.