The easiest one would simply be removing the down vote mechanism. If with an existing site then can compare with the years' prior stats - assuming they've been recorded; I don't know if HN does. There are a number of things you could monitor - would be interesting to see if any sets of people with the same behaviour change their behaviour more than just down voting, like if they stay on the site more or less, if they comment more or less; perhaps doing word use analysis on these different sub sets before and after the down vote change, what language are people who are prone to down voting using vs. those who only comment or only up vote but don't down vote, etc.