[EM] Re: simplcity of range v condorcet

[Rob Lanphier]

On Sat, 2005-08-13 at 19:42 -0700, Rob LeGrand wrote:
> Rob Lanphier wrote:
> > In fairness, the specification for counting votes is something
> > that voters will probably care about, and it is one of the
> > biggest liabilities of Condorcet.  Part of the uphill battle for
> > Condorcet advocates is to convince people that even if they don't
> > understand exactly how it works, it's still a better system (the
> > tactic I've usually advocated is endorsement from trusted smart
> > people).
> 
> This tactic seems possibly dangerous to me.  There's a fine line
> between asking the public to trust a decentralized network of
> experts (open-source software, "mainstream" science: good) and
> asking them to place their trust in a centralized "expert"
> authority (governments, cult leaders: bad).  I think many people
> can't tell the difference between the two and either trust both or
> neither.  I'd prefer that a public election system be simple for
> everyone to understand in the first place.

It's a tradeoff, though.  Either the mechanics are simple, or the
strategy is simple.  There doesn't seem to be an example of both.

In spite of Range and Approval advocates claim that the optimal strategy
for Range and Approval is simple, I have a hard time getting my head
around it myself.  Quite frankly, I don't truly understand why it works
the way it does, but I'm only taking the word of people on this list.
I've also got some doubts about how well the strategy works in all
cases.

The trust in Condorcet that I'm talking about is more of a
trust-by-inspectability, very analogous to the open source example you
bring up.  It's not as though you can expect everyone to understand the
system, it's rather that you expect everyone can choose whether they
want to study up and learn it themselves, or find a smart person they
trust and have them give it the thumbs-up, thumbs-down.

Regardless, the basics of Condorcet methods are actually pretty simple.
It's only in dealing with edge cases (i.e. when there's more than one
candidate in the Smith Set) that things get complicated.  Up until that
point, it's basically a round-robin tournament.  Anyone with experience
as a sports spectator should understand the basics well enough to feel
confident that the system is fair, and that sometimes, tie-breakers get
messy.  It can't possibly be any worse than NCAA football ;-)

> I think Warren Smith makes a good point when he says that many
> voters would be tempted to use Borda-like strategy under a
> Condorcet system, however effective it can be shown to be.  If my
> sincere vote were A>B>C and C obviously had almost no chance to
> win, I'd be very tempted to vote A>C>B to hurt B's chances.  And
> under a winning-votes system I'd strategize by voting A=B>C even if
> I knew nothing or expected it to be a close three-way race.
> Following this intuitive Borda-style strategy under Approval or
> Range Voting never requires expression of an insincere pairwise
> preference.

If voters still try to game the system and shoot themselves in the foot,
then its their own damn fault.  We shouldn't be /too/ sympathetic to
people trying to game the system.  We should only be sympathetic to the
point where it's actually rational to use strategic voting.

Why?  The goal is to make the best decisions in the fairest way
possible.  If people manage to disadvantage their voting power by doing
something they shouldn't, while smarter and more honest people get more
of a say simply by voting sincerely, isn't that better?  Won't better
decisions get made over the long haul by listening to the smarter and
more honest people?

When a system forces one to choose between "smart" and "honest", that's
when we've got a genuine dilemma.  We can't assume that people will
subvert self-interest in the name of honesty, hence we assume that
they'll do the self-interested thing first.  But I think it's folly to
worry too much about voters that employ dumb strategies.

Rob