Two things that I would really like to see:

  • Estimation of policy intervention effects: Methods for running realistic simlulations for evluating the effects of real world policy interventions, so that any planned intervention (say, a tax hike on a certain subpopulation) could be tested to determine whether an intervention will have the desired effect, whether there will be unintended consequences, etc. Most current day simulations are too simplistic and need to make simplifying assumptions that ultimately make it hard to justify taking their results seriously. But if we could create a framework for running realistic simulations and trust their results, imagine the huge implications this would have: a government (or indeed a corporation), could run a sim to figure out whether a change they are planning will be successfull, before actually implementing it.

  • The inverse problem: A computational framework for designing the details of a policy intervention that would meet the desired goals. Apart from a few wins in very specialized settings such as auctions, mechanism design as field, has under-delivered. The grand vision of the field is to allow one to design the rules of a game that has a desired equillibrium. I would venture a guess that practitioners in the field (is there such a thing as mechanism design practitioners? or are they all acemic researchers?) are too bogged down by the need to provide formal guarantees - after all if you cannot provide formal gaurantees, then what good is your mechanism? This is especially true if your goal is to get your work published rather than create mechanisms for th real world. I for one would really like to see this field priduce more real world practiotioners!