Further to Graeme's piece, I note that his point about the foolishness of weighting last and first choices equally has two components. The first might be called the "assumption of equal caring" and is more complicated and more contentious that the second component, which I shall call the "assumption of useful informedness".
Under the assumption of equal caring, we take it that someone's preference for their first choice over all other alternatives should carry equal weight as a preference to their next-to-worst preference versus what they like least. Now, I think that's a pretty tricky equal weighting to defend, but it's also a bit difficult robustly to dismiss it. After all, under first past the post we don't really know how much someone voting for, say, the Conservative really prefers the Conservative over the Labour candidate. For all we know, it could be that in some case where the Conservative gets 12,100 votes and Labour 12,000 votes, in fact 12,050 of those Conservative voters were virtually indifferent between Conservative and Labour whilst all of the 12,000 Labour voters would prefer slow disembowelment to having a Conservative MP. So strength of preference may be poorly reflected even under first past the post - it's unclear that we lose that much under AV.
What is more problematic for the AV advocate is the assumption of useful informedness. Under first past the post, with universal suffrage, we assume that everyone is sufficiently informed to contribute usefully to the decision over who is his/her representative. That's pretty fundamental to democracy, and first past the post facilitates it since for most of us the decision boils down to whether we prefer the incumbent (Labour or Conservative) or the alternate (Conservative or Labour). If we are particularly interested we can research matters a little further and decide we actually prefer a Lib Dem or a Green or a Communist or whatever.
But under AV we assume that people have usefully informed opinions regarding the relative merits not merely of their preferred candidate versus the others but across the whole spectrum. Well, it's difficult enough to gather information about the parties and their beliefs (and even more challenging in respect of how the candidates in our area differ from the national parties in subtle ways) so as to be able to decide who is our favourite. How many people are really in a position to offer any vaguely informed preference between their fifth and sixth options? Most of us reading this site are sad political nerds - how many of you really feel informed enough to express a useful preference between your fifth and sixth best options?
Now of course under the form of AV under consideration voters are not obliged to express preferences across the whole range. But the reality is that many of them will do so, without having the foggiest basis for why. They will do things such as express their first two or three preferences, then rank the candidates alphabetically (or, if it differs, rank them on the order they appear on the ballot paper). But we know that under AV races can and do come down to preference allocations well down the list. So election results will be decided by the entirely frivolous and uninformed random rankings of those that have not been able to gather enough information to offer a sensible view.
People have all of their time to garner sufficient information to be able to express one preference under first past the post. We can grant that few of them will be perfectly informed, but it's reasonable to take account of what they say anyway - the tradeoff between informedness and inclusion favours inclusion pretty clearly. But do you really think, oh AV advocate, that that trade-off is attractive across the whole span of voter preferences?
