[EM] Quick and Clean Burial Resistant Smith, compromise

[Colin Champion]

Daniel wrote:

 > Given a choice between two methods that are equally good in other 
respects, I would pick the one where the best strategy is to vote 
honestly. It's hard to trust that an otherwise good voting method will 
produce good results if people are motivated to lie on their ballots.

This is a valid point. My own metric implicitly assumes that an innocent 
error is neither more nor less serious than an error induced by tactical 
voting. This is a reasonable starting position but not ultimately 
tenable - ideally we'd want to weight the two sorts of error 
differently, and recognise that providing people with an incentive to 
vote dishonestly is an additional fault besides getting the wrong 
result. But a weight of 0 for innocent errors is even less tenable, and 
if we're to quote a single figure we can't avoid adopting some implicit 
weight.

In fact even innocent errors aren't really innocent. If a method 
produces errors which aren't due to tactical voting, they're probably 
due to the configuration of candidates, and therefore provide perverse 
incentives to candidacy which in turn distort political discourse. In 
the end I've been inclined to stick with equal weighting while admitting 
that it needs a prominent health warning, but this is partly because 
I've never seen the position argued that tactical errors are more 
critical than innocent ones.

On the magnitude of errors, my own figures suggest that Condorcet/Hare 
makes fewer but larger errors than minimax when confronted with false 
cycle tactics, and therefore looks marginally better in terms of 
accuracy (and no doubt much better in terms of strategy resistance) but 
significantly worse in terms of Euclidean loss. See
http://www.masterlyinactivity.com/condorcet/condorcet.html#results
(where Condorcet/Hare=Condorcet+AV).

And on IRV... I don't favour it at all, but having condemned it for its 
faults, I don't want to additionally condemn it for its few merits, one 
of which is allowing a strategic escape from the centre squeeze.

CJC

On 17/01/2022 09:31, Daniel Carrera wrote:
> On Mon, Jan 17, 2022 at 2:48 AM Colin Champion 
> <colin.champion at routemaster.app> wrote:
>
>     A bit of a side question, but is strategy resistance a good metric
>     to quote?
>         FIrstly, you'd never rank methods on it - you'd end up choosing a
>     coin toss.
>
>
> It's not my #1 priority, but it seems pretty important to me. Given a 
> choice between two methods that are equally good in other respects, I 
> would pick the one where the best strategy is to vote honestly. It's 
> hard to trust that an otherwise good voting method will produce good 
> results if people are motivated to lie on their ballots.
>
>
>         Secondly, voting methods recommended for their resistance to
>     tactical voting sometimes have the property that when they go wrong,
>     they make larger mistakes than those made by methods targeting
>     sincere
>     voting. If you don't take this into account, you can draw misleading
>     conclusions.
>
>
> Hmmm... I did not know that. Can you give me an example? Yeah, that 
> sounds like something we'd really want to avoid.
>
>
>         Thirdly, consider a 3-candidate election using IRV. The central
>     candidate is the Condorcet winner, but supporters of one of the
>     non-central candidates realise that the opposed non-central
>     candidate is
>     likely to win owing to the operation of a centre squeeze, and
>     therefore
>     compromise on the central candidate. This satifies JGA's
>     definition of
>     strategic manipulation but is certainly not an additional fault in
>     IRV;
>     on the contrary, it's a mitigation of its weakness under sincere
>     voting.
>
>
> That does sound like a fault in IRV. If the goal is to get the 
> Condorcet winner (and I agree!) why not use a Condorcet method? The 
> example you cite is one of my biggest complaints with IRV. If people 
> are driven to vote dishonestly to compensate for a failure of the 
> system, that tells me that the method is flawed.
>
>     My own practice has been to measure the performance of methods in the
>     presence of tactical voting exactly as I would in its absence, simply
>     skipping over any attempted manipulation which doesn't lead to a
>     worse
>     result.
>
>
> If two election methods elect the same good candidate, but one of them 
> achieves that by letting people vote honestly and the other requires 
> people to learn that they need to manipulate their votes... to me that 
> seems like a relevant difference.
>
> -- 
> Dr. Daniel Carrera
> Postdoctoral Research Associate
> Iowa State University

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20220117/d02eaade/attachment.html>