[EM] Markus & Manipulability

[Forest Simmons]

On Mon, 4 Feb 2002, Blake Cretney wrote:

>   Joe Weinstein wrote:
> 
> >
> > As a citizen and voter, I don't want the election method to give 
> > gratuitous incentive to CAMPAIGN strategies which aim to confuse and 
> > entrap voters, e.g. thru introduction of incontestable fallacious poll 
> > data or of extra clones (pro or maybe con a given position).
> >
> > But I do have a VOTING strategy.  My strategy has me vote (or abstain) 
> > so as to maximize my expected overall satisfaction - call it 
> > 'utility'.  Ingredients which enter into this utility include:  how, 
> > apparently, my vote would most likely (or thereafter profitably) make 
> > a difference in picking the winner (instrumental effect);  how my vote 
> > would most likely (or profitably) bear weight as an expression of 
> > sentiment;  my costs in time and money to go and cast the vote; my 
> > regret and losses in self- or others' respect from not voting or from 
> > voting 'insincerely'; etc.
> >
> > I not only have a VOTING strategy:  I am in fact ENTITLED to conceive 
> > and have and use such a strategy.  Indeed I DEMAND that the election 
> > method give me scope for effective strategy.  (A decent respect for 
> > democracy requires me to concede the same rights to other voters too.)
> 
> The more a voting system allows effective strategy, the more it allows 
> those who understand the system to get more power than those who do not. 
>  This may be a necessary evil, but it surprises me that you seem to feel 
> this is actually desirable.
> 

The source of your misunderstanding is that Joe and I use the word
"strategy" inclusively, i.e. to include sincere strategy.  If the optimal
strategy for a method is to vote sincerely, then so much the better.

In Random Ballot the optimal strategy is to vote your sincere preference
order.  In Perverse Random Ballot the optimal strategy is to vote your
reverse preference order. The two methods are what I call "strategically
equivalent" like CR and Approval, or like Lone mark Plurality and
Cumulative Voting.  In each of these strategically equivalent pairs
transparent strategical considerations remove all essential expectations
in the outcome. 

The designers of Perverse Random Ballot, CR, and Cumulative Voting might
have reasons for wishing that they were not strategically equivalent to
the simpler halves of the respective pairs.  Well, they failed to meet
their design specifications, which is not the fault of the voters. 

> >
> > As Mike pointed out, most of us - including me - have been energized 
> > into election reform because we resent being often forced to trade off 
> > two quite reasonable strategic goals: defensive - to defeat the worst 
> > alternative; and affirmative - to support our favorite.  We know that 
> > this tragic trade-off is largely avoidable, through use of another 
> > method such as Approval in place of the prevalent Lone-Mark.
> >
> > It's bad enough being hobbled in the kind of strategy I can use.  
> > What's worse, though, is to be berated for - or prevented from - 
> > effectively strategizing at all.
> >
> > A totally non-'manipulable' (by me as voter) election - e.g., one 
> > where my choice has minimal effect (whether on account of someone 
> > else's prior choices, or gratuitous randomization, or a combination) - 
> > is of course of no civic interest or benefit to me.
> 
> Obviously being manipulable (as you define it) is a good thing, and 
> strategy is a form of manipulation.  But that isn't really an argument 
> for why strategy is a good thing.  In the same way, I could argue that 
> donuts are food, and that food is good for you, therefore donuts are 
> good for you.

Actually Nurmi defined the 'manipulable' that Joe put in quotes. That's
why he put it in quotes.

Again keeping in mind that "strategy" includes sincere strategy, if the
voter has no idea what his best strategy is, then he is at the mercy of
the winds of fate.


> 
> >
> > I must heartily second what Forest wrote so eloquently earlier today:
> >
> > '... Nurmi and Bartholdi are worried about voters "manipulating" the 
> > system to increase their expected utilities, i.e. to vote in their own 
> > best interest, as though voters' utilities had nothing to do with 
> > social utility.
> >
> > 'It is the prerogative of the voter to maximize their own utilities,
> > whether anybody else thinks they have social value or not. That's
> > democracy. We don't try to use voting methods to protect the public
> > against the public will. We use voting methods to ascertain the public
> > will.
> 
> We clearly use voting systems to protect the public against the will of 
> individuals.  As far as I can see, this is just a continuation of the 
> same error I mention above.

The public is made up of individuals.  The only way to ascertain the public
will is to somehow aggregate the wills of individuals.  If a method is so
opaque that it is unclear how to use it to express your will, then you
and others like you are effectively disenfranchised.

> 
> >
> > 'A more realistic worry is pollsters and pundits manipulating the 
> > voters by fooling them into voting against their own best interest. 
> > The more complicated the strategy and the more sensitive good strategy 
> > is to
> > information ...the easier the "experts" can manipulate the vote of the
> > gullible voter.'
> >
> Most voters seem to view IRV as impossible to manipulate (by strategy as 
> opposed to sincere votes).  So they don't try.

Although I have seen more evidence to the contrary than evidence in
support of this assertion, suppose for the moment that it were true. Then
what a pity that IRV strategy is so hard that people are stuck with the
lousy results of voting their sincere preference.

In a non-monotone system sincere voting doesn't effectively reveal the
public will.


> But it's obvious how 
> approval can be manipulated by expert opinion.  People will want to 
> approve of at least one option they think has a chance.  In run-off, 
> voters will try to avoid wasting their votes in the first round. 

How do they best avoid wasting their votes in the first round if not by
voting for an option that they think has a chance?

In other words, they use the same basic strategy that you suggest for
Approval voting. The difference is that they have to betray their
favorite, a greater 'manipulation' and violation of conscience than
elevating a lesser evil to the same level as favorite.

The constraint of strict ranking requires this more extreme 'manipulation'
in the case of IRV, and so makes IRV less stable than Approval.

Forest