Sortition is interesting, but certainly not a panacea.
Generally, it works best when all or at least most candidates are competent, and the task they’re being selected for is well defined. The position of party secretary or party treasurer, for example. As long as all the people who put their hand up are capable then might as well randomly select who does it. Could probably select all party positions this way, and there are arguments both for and against doing so.
When forming a representative government body by sortition it’s a bit different. The task isn’t really well defined at all. Procedure aside, policy can be taken in all sorts of directions, and if you ask a bunch of random people you’ll get all sorts of different answers. If the candidates have to rely on consulting the public then sortition is strictly inferior to election, as sortitioned candidates won’t have information on how the public reacted to policy campaigning. So the only sensible way to get a result that’s more or less consistent with what the public wants is to have enough people selected to form a statistically representative sample.
I’m not an expert on statistics, so I had some fun playing around with a survey sample size calculator. Turns out that for a given confidence level and interval, the number of people you need grows very quickly up to some asymptote that’s independent of the total population size.
At a confidence level of 95% with an error interval of +/-5%, for example:
total pop 100 -> need 80 reps
total pop 500 -> need 217 reps
total pop 1000 -> need 278 reps
total pop 2500 -> need 333 reps
total pop 5000 -> need 357 reps
total pop 10000 -> need 370 reps
total pop 16 million (approx current electoral roll size) -> need 384 reps
On the other hand, if there’s only 76 reps in the Senate like there is now representing the same 16 million, and still assuming a confidence level of 95%, then the interval is +/-11.2% for issues the public is evenly split on. That’s wildly inaccurate.
Based on this, a sortition based upper house does not appear to be practical in Australia. Unless you’re all fine with tripling the size of parliament or thereabouts.
But what’s really interesting is the same reasoning suggests sortition to be a really good idea for USA congress and UK parliament. In the USA they’ve long since reached the point where each representative has too many people they represent to actually meaningfully consult with them, and more reps won’t help since the size of Congress is already nearly 600. They could get quite accurate sortitioned results with the same number of reps. And the UK already has the House of Lords with 794 seats. Again, plenty of room for representative sortitioned results, possibly even including a reduction in size.
(Another result of these stats is that most citizen’s juries are apparently very undersized if the goal is to figure out the opinion of the general public.)