<div class="gmail_quote"><div>Reductio ad absurdem. One voter, two candidates. Preference-based criterion: "If the voter votes for A but actually prefers B, then B should win". But, people say, what if they meant to vote for A? Nope, you say, doesn't fit the criterion. So fine, I can find whether system X meets this criterion; but we can't have a reasonable conversation about it, because half of the relevant examples are somehow arbitrarily outlawed because they don't "fit".</div>
<div><br></div><div>Preference-mentioning criterion: "Imagine the voter prefers B, but due to an epileptic seizure, votes for A. The correct winner in this case would be B. Therefore, whenever we see a vote for A, we should elect B." Now, it is easy to say why this is a silly criterion: "But if they meant to vote for A, then A is right."</div>
<div><br></div><div>I'm not implying your criteria are this silly, or even silly at all. I'm just saying that a criterion can be justified on the basis of preferences, but, like the system itself, must ultimately speak in terms of ballots and results.</div>
<div><br></div><div>Jameson</div></div>