[EM] SODA strategy

[Jameson Quinn]

2011/8/9 Juho Laatu <juho4880 at yahoo.co.uk>

> I checked the definition of SODA at the wiki page. Since the method
> consists of multiple phases and has many rules, it was difficult to find a
> simple mapping from that to one simple claim that could be proved or
> falsified. I also had some problems with terms semi-honest, non-semi-honest,
> self-reinforcing and defensive strategy below.
>
> I had multiple thoughts on where SODA might be vulnerable and where not,
> but on the other hand I didn't know which phases were supposed to be
> "strategy free" and which way (the phases whose role I wondered were
> nomination, preference order declaration, voting and vote transfer). Maybe
> one could do this in smaller pieces, like handling separately the chicken
> problem for one of the phases etc. Another approach would be simply to list
> all identified possible vulnerabilities and then prove that all those cases
> are harmless. Is there one major claim that could sum it all (at least the
> claims) in one sentence or should we start from smaller pieces?
>

I think starting small is probably the best way to go.

I believe that there are worthwhile strategy-proofness claims to be made
about each of the phases except preference order declaration. For that
phase, the principal "defense" against strategy is the assumption that
dishonesty in this phase would be detected and punished by the voters,
either by voters withdrawing support, or in a weaker way by their turning
delegated votes into non-delegated votes.

The system I want to use for strategy considerations is SODA-DAC with
candidate preferences completed to strict rankings by approval order. This
differs in two ways from the simplest basic SODA version, but I believe that
the latter is close enough to share most advantage.

I'll try to develop a list of clear sub-claims in a later message.

JQ


>
> Juho
>
>
> On 9.8.2011, at 16.14, Jameson Quinn wrote:
>
> > SODA is not strategy free. Even if you make the assumption that candidate
> preferences are honest because dishonesty will be detected and punished by
> voters -- an assumption which puts the system beyond the reach of the
> Gibbard-Satterthwaite proof -- the fact remains that you can construct
> strategic scenarios.
> >
> > However, it seems to me that SODA is not just a less-strategic system
> than most others, but radically so. Unlike Approval, semi-honest approval
> strategy is not something voters must deal with at least implicitly. But
> like approval, non-semi-honest strategy is relegated to a tiny minority of
> voters in a tiny minority of cases. The system can deal with all the
> commonly-discussed strategic problems, including chicken, center squeeze,
> and honest cycle. I honestly suspect that strategy under SODA would be
> favored less than half as often as any other good deterministic system I
> know of, including Approval, Asset, Condorcet (various), IRV, Median, and
> Range.
> >
> > So, how would you set out to make this idea demonstrable or falsifiable?
> What rigorous statement about strategy and SODA could I make that would be
> testable, preferably using simulated elections or mathematical
> demonstration/counterexamples? What voter model could capture enough of the
> sophisticated strategic thinking of which humans are capable?
> >
> > How about "SODA requires no self-reinforcing or defensive strategy"?
> >
> > These are honest, not rhetorical questions. I appreciate good responses,
> good research questions, from anyone, whatever you expect that the results
> of that research would be.
> >
> > Thanks,
> > JQ
> > ----
> > Election-Methods mailing list - see http://electorama.com/em for list
> info
>
> ----
> Election-Methods mailing list - see http://electorama.com/em for list info
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20110809/8b629b6a/attachment-0004.htm>