See my op-ed on ORMS today:
To decide how we can best get our candidate elected, we need to define our goals as well as any constraints.
Some variables we could include as either goals or constraints:
Likelihood of our candidate winning
Cost
Resources like volunteers
Political capital / Favors (and what they might be worth in the future)
Goals like changing the political discourse on particular topics
Managing campaign promises and the ability to govern if elected.
The meta-problem of getting your candidate elected is largely about understanding how any action you take today / tomorrow / in the future helps your chances on election night. It also often depends on a whole portfolio of candidates and limited resources to support them all. In some cases, campaigns may even be more interested in publicizing issues rather than focusing only on getting elected.
Depending on how the landscape evolves over time, the choices available to you and opportunities may change dramatically. The uncertainty over time is typically the most significant challenge and campaigns spend many resources on polling to understand the odds.
Assuming we do not know exactly how our choices will impact outcomes, we have an incompletely defined problem to take actions over the planning horizon. If we knew exactly how our actions would impact both our opponent and our outcomes, the problem would be completely defined with a single right answer. But since it is an incompletely defined problem there could be multiple right answers depending on our priorities.
And so to solve the meta-problem we primarily need a model of uncertainty and how our decisions can impact that uncertainty. Depending on the result, we can identify our best actions today and in the future to achieve a better solution to the meta-problem.