Andrew Harless says that the rest of us want there to be gamblers in financial markets who trade in public:
A Note: Prolegomenon to Any Useful Discussion of Modern American Finance: I'm not sure whether there are a whole lot of trades I don't like. A typical trade (perhaps) is between two intelligent, rational, well-informed investors, each with their own set of private information. Necessarily, one set of private information will turn out ex post to have been better than the other, but both pieces may be good in an absolute sense, and neither side would be willing to reveal its information to the public except as implied motivation for the trade.
Even if it is mostly zero-sum (and if we ignore time and risk preferences, etc.), trading is useful as a way to get market participants to disclose (admittedly in extremely vague form) their private information. If we see movement in a price, we have new information about the value of the asset, which we wouldn't have had if there weren't a bunch of agents out there trying to take advantage of one another.
When a trade is motivated by moving money across time--savers and builders--or by getting decision-makers skin in the game--principals and agents--or by hedging risk--principals and agents--in general both sides of the trade will be happy both ex ante and ex post. As Andrew says, when a trade is motivated by disagreement, one side of the trade must be unhappy ex post. That means that the side that has the lower-quality information ought to be happy ex ante. So in order for the trade to go, not only must one side be wrong in its information, but both sides must think they have the better information and at least one of them must be wrong in that also.
Andrew is making a somewhat subtle argument that even though a disagreement-driven trade cannot be Pareto-optimal for the two parties ex post and hence cannot be Pareto-optimal ex ante, it is still good from a social-welfare point of view because of the informational externalities arising as gamblers are motivated to discover useful information and then partially reveal it through their trade patterns.
This is correct. The question is: how big are these informational externalities, and how much of a gambling propensity would we then want to see in the market?