I think a lot of arguments which support the idea of HFT being too much liquidity seem to take, as a premise, that the extra liquidity costs consumers -- namely, non-HFT participants in the market trading against the HFTs, the "buyers" of liquidity. Does it? If you have too much liquidity, doesn't that just mean that the buyers don't buy any of it? They saw as much liquidity as they needed, completed their trades, and went home. It seems like the actual cost of too much liquidity is HFT firms which send orders that don't get traded against, so they have no benefit (to the HFT), but still have a cost (the firm has to run, after all). Yes, there's some cost to society here, in the same way that there's a cost any time someone takes on a speculatively profitable business that turns out not to be.
I'm sure there are some strong arguments you could make that HFTs are unnecessary, but the argument of "too much" liquidity always felt shaky to me (in part, because it's one of those arguments that relies on playing fast and loose with nebulous terms).
Said a different way, with your "trading costs" substitution: If x amount of human effort is required to reduce trading costs from, say, 5% of a transaction to 0.001%, perhaps (let's assume) that's justified. But is it then wise for a society to encourage expenditure of 10x more effort just to reduce costs to 0.0001%? (This analogy doesn't really hold, because cost of trading could be 0 and there could still be 0 liquidity, just no other parties interested in buying or selling, but it still seems to point at a real issue when restricted to talking about liquidity.)