On 1/22/07, Jan Kok <[EMAIL PROTECTED]> wrote: > On 1/22/07, Abd ul-Rahman Lomax <[EMAIL PROTECTED]> wrote: > > > Converting Range to Approval, though, requires the specification of > > an Approval cutoff. Various cutoffs have been proposed. 50% is the > > obvious simplest one. This gets a lot clearer if the Range is > > expressed as -N to +N. 50% is zero, i.e., is easily interpreted as neutral. > > Based on my my thoughts about strategic voting and the need for > intermediate values in Range Voting, posted to the RV list a few weeks > ago, I think the Approval cutoff should be set just below (or maybe > just above) 50%. I don't see a need for setting it anywhere else.
On further thought... If there are three frontrunners, Good, Middling and Bad, then one would vote strategically by voting max (say 1) for Good and min (say 0) for Bad. What about Middling? Calculate a score for Middling by scaling your utilities for Bad, Middling and Good to the 0-1 interval. If you want to vote sincerely for Middling, just vote his score. Otherwise, vote max or min for Middling depending on whether he is above or below your approval cutoff. But, where should you set your approval cutoff? It's your expected utility of the election not considering Middling, on the 0 to 1 scale. In the case of just the 3 candidates, it's the probability that Good will win. So, if you think Good has a .66 chance of winning, then your approval cutoff should be .66 . > > > > > Our votes here are only advisory. The advice will be more useful if > > the collection of opinion that it is based upon are expressed in more > > detail. There is no risk of strategic voting in Range causing some > > harm, because if ballot analysis shows the possibility of strategic > > voting, the analyst can compensate for that. > > Uhh, how would you know that a ballot was voted strategically? Even if > someone votes all max and min values, maybe the voter just has strong > opinions about the choices. And how would you compensate for it? > > Cheers, > - Jan > ---- election-methods mailing list - see http://electorama.com/em for list info
