Hopefully Anonymous, my point is that optimal is functional. If we find that our "optimal" policy is not functional, we need to expand the scope of our cost-benefit analysis.
If enough people are seriously disgusted by the possibly of compulsory trials (and I think they would be), the policy is unlikely to pass a cost-benefit test. When people balk that a particular policy will take their freedom, they are essentially saying "this policy would cause me harm, since I value my freedom." We need to look outside the most obvious costs and benefit when we evaluate policies.
A related example: by superficial utilitarian standards, compulsory medical trials for only the lowest-income members of society might seem a better policy than randomized trials in which anyone can be chosen regardless of economic status. After all, high-income people are far more likely to be meaningfully contributing to society. But the "poor only" law plainly violates our sense of equity and fairness, which is equivalent to saying it imposes large costs on us.
So I don't think, as savagehenry says, that a society built on minimizing harm would be much less free, provided we define "harm" sufficiently broadly. People value their freedom too highly, and loss of freedom is quite harmful given these values.
I am not convinced that a utilitarian legal system is much different than the systems of modern Western societies. Most laws in such societies are passed on grounds of ethics, efficiency, or a combination of both. Many people assume that laws passed on ethical grounds are inefficient by utilitarian standards, but I don't think that's necessarily true.
Consider murder laws. These are typically justified with a moral argument: life is sacred. But when a person is killed by another, the cost is not just some abstract violation of moral principle--since it is people that hold morals, and since those morals (usually) specify that loss of life is a bad thing, murders impose costs on members of society. These costs could conceivably be measured ("How much would you be willing to pay to revive person X?") and should be part of any cost-benefit calculation aimed at judging the value of a "moral" law. Of course, murder laws yield obvious gains in economic efficiency as well (disincentive to kill->more security->more commerce, peace of mind etc.), but I am considering only the moral side of things to make my point.
Why don't we have compulsory clinical trials? I wager they wouldn't pass a proper cost-benefit analysis.
Good idea. If it's successful, I suggest creating a new one every month or two.
Chapter 11 of the 9/11 commission's report, available here, shows the commission was very wary of hindsight bias. The failure to prevent the attacks is said to represent a "failure of imagination," meaning the intelligence community used the wrong model in evaluating terrorist threats.