Condorcet pairs on the ballot

[Steve Eppley]

Hugh T wrote:
-snip-
>(An exchange with Mike on this issue did not make it onto the list;
>I won't post it unless requested.)

For some reason, Hugh's messages to the EM list have a 
"Reply-to:Hugh" field in the message header.  I presume this is a bug 
in the mailserver, which ought to overwrite it with "Reply-to:EM".
Maybe Hugh could configure his email manager not to insert a Reply-to 
field?

>Even if one assumes that the hypothetical ballot above is likely in
>error, it does not follow that putting pairwise races on the ballot
>makes voting error-prone.  Does Steve assume, or have evidence,
>that any ballot with three two-person races for different offices
>produces a significant number of mistaken votes?  

It's an assumption on my part, derived from Murphy's Law.

>Would Steve at least agree that if circular ballots are prevented
>by the computer there is no reason to deny the voter the option to
>vote in pairwise races directly? 

I don't know if intransitive preferences must be forbidden, as long
as the computer can interact with the voter to verify it wasn't an
error.  But I haven't considered the consequences regarding new
tactical options this might allow, and it might take scenarios with
4 or more candidates to turn up problems.

With a computerized interactive ballot, the voter could be given a 
choice of interfaces.  A "power user" might want to use an interface 
based on rankings but which allows a way to enter the occasional 
intransitivities.  Here's an example ballot:
   D>B>C, F>D, A>B>C>D>E>F
In this example, the D>B>C order overrides the A>B>C>D>E order in 
the D vs B, D vs C, and B vs C pairings, and the F>D overrides 
in the D vs F pairing.

---Steve     (Steve Eppley    seppley at alumni.caltech.edu)