Open Stack

On Oct 3, 2016, at 1:39 PM, Clint Byrum <clint at fewbar.com> wrote:
> 
> Of course, I read the essays of those who I don't know more carefully, and
> I often go searching through my ML archives to see if we've interacted on
> threads in the past. Still, I'm very unlikely to rank somebody higher than
> a personal acquaintance unless I don't have many personal acquaintances
> that I agree with that are nominated.

I understand, and would be less than honest if I stated that I never do the same thing. I'm questioning, though, whether that's a good thing, or part of our tribal nature as humans.

> So I get where you're going, but I don't think that can be "the
> election". In addition to not allowing me to judge peoples' character,
> it also introduces a _massive_ fraud incentive. If you are a candidate,
> and you write a paper that is equal to all others, you can gain votes just
> by secretly telling your friends which one is yours, and their implicit
> bias will 1) make many think this is morally ok, because they know you're
> a good candidate, and 2) make them more likely to vote for you.

I suppose that's a possible downside, although I don't know that anyone who would do that would have enough people they could call "friends" to get them elected. I know that such a tactic would certainly backfire if someone tried to get me to vote for them.

> One way to counter the problems associated with the popularity contest
> is to have some appointed members. We can admonish the TC to appoint a
> nominee who did not win the most votes, but who is more likely to break
> a groupthink cycle. This would only work if people paid attention to TC
> voting records, which AFAIK, nobody really does.

Heh, let's change the bylaws so that the top 4 (or 5 in the next cycle) vote-getters win, and the other two seats are chosen by lottery from the bottom 5 vote getters! That's sure to liven things up! 

(kidding, of course!)

-- Ed Leafe