> You may argue that it's in everyone's best interest for us to use our free downvotes wisely, but if we could cooperate like that, we could make linear and 25% curation work as it's to everyone's detriment that we all engage in content indifferent behavior. Yet here we are.

I don't find the situations analogous at all. Altruistic or socially-optimal voting under the current system has an obvious and huge immediate cost, so it is a pretty easy to expect it to be heavily disfavored under the current system. To cooperate and sustain it would be a huge effort with large coordination cost.

By contrast, free downvotes have very little direct cost. There _might_ be some retaliation, but that also _might_ be implausible (if someone is downvoted by 5 or 10 different voters, are they going to retaliate against all of them; vote power limits alone might make this impossible). It is far more likely to expect that some altruistic or non-myopic self-interest to kick in there, when the cost is much lower, the way it does when people make small (positive) edits to wikipedia and such.

I would agree it may not be sufficient, but I don't think it is clear it is insufficient just because it isn't directly rewarded and therefore perfect game theory might suggest ignoring the option altogether.  Though to be perfectly mathematically precise, if you have any active content eligible for a payout, downvoting does benefit you, however slightly, and likewise, long term good-of-the-platform considerations also benefit you, if also slightly.  Again, it is more likely to expect these considerations to matter when they aren't offset by a huge direct cost. After all, it doesn't cost much to physically click downvote if you don't like something, as people do millions of times per day without any incentive on reddit, etc.

BTW, I do think downvoting (assuming it happens a reasonable amount, which is uncertain; see above) brings price discovery, not directly, but via its effect on upvoters. That's precisely non-linearity in action. Upvotes have more 'oomph' if they don't get downvoted, just as they would have more 'oomph' if combined with other upvotes (or, at least, more stake) in a superlinear upvoting system. So people upvoting who want their votes to have maximum value (either for curation or reward purposes), a natural desire, need to consider what is more or less likely to be downvoted. That brings price discovery. In the extreme case, if you upvote for N rshares and get downvoted for N rshares your vote then has no value at all. That's clearly inefficient and unprofitable voting you would prefer to avoid.