October 27, 2007
Decision Meta-Strategy
On CNE, I blog about the new european health strategy. The good: it understands that health is important, that individualized health is coming and that we need to benchmark it. The bad: it doesn't get the strategic benefits of life extension, it focuses on politically correct issues and wants to force inclusion of health concerns into *every* policy everywhere.
I wonder about that last part. In an ideal world every decision would take into account every kind of consequence - the environmental, social, spiritual and economical effects of every little playground or software utility ought to be considered. It is rational to try to get the best outcome, and at the very least we ought to state what outcomes we are aiming at. But clearly it is impossible to assess all possible aspects: there is not only merely 24 hours in a day, we also lack the knowledge of how 1) the policy will actually be implemented, 2) how this implementation actually affects the world and 3) how we in the future will judge these effects. So we have to use a bounded rationality approach, where we look at the most likely effects that have the biggest or most certain consequences. It is essentially a search tree situation with a probabilistic adversary ("nature").
Adding new areas that ought to be assessed is essentially demanding that the search tree branch more strongly. That means that we will have less resources to do depth searches and should use robust heuristics, or simply go for satisficing. But the current administrative vision is that we should all be optimizing and use as much information as possible. How can that be avoided?
One approach would be to have the administrators bear the costs of the information demands they make. If the Commission wants to add another dimension to every policy, they need to give every policymaker affected an increased budget. While this might be nice if you are a bureaucrat, it sounds like a recipe for balooning bureaucracy.
Another approach, which would also promote satisficing and finding better heuristics, is to make policymakers affected by the utility of their decisions. If it goes well, they get more money/resources, if they make stupid decisions they lose. Ideally, decisionmakers should be able to make a fortune (real money or political capital) from supporting a decision that turns out to be very good - and lose it if they also support bad or wasteful decisions. This would encourage rationality and accountability. To get it to work we need better follow-up, which is also good for transparency and as a basis for future decisionmaking. It may also be tricky to determine what constitutes a good result, and how to factor in external forces, as well as avoiding perverse incentives. But these issues can likely be solved; what we need from the start is the recognition that policymakers should get incentives to make better decisions, not just more decisions.