Dilemma: a way to define the meta-problem
A meta-problem is about identifying the best problem to solve, knowing there is a tradeoff between the value of solving a particular problem and the effort it takes to solve that problem.
Dilemma is a word to describe problems where the path forward is less clear, and so we just begin with the high-level issue we want to address.
Beginning with a question can help you get clearer on what you want and how you can accomplish those things.
Understanding the problem neighborhood is just one part of the meta-problem cycle. You can also see illustrations of this idea in practice on the Examples page.
For a more technical view
Incompletely-Defined Problems
Solving a meta-problem starts by defining an incompletely defined problem.
An incompletely defined problem is a problem where we have defined some of the criteria, but especially may not know how to trade off one variable versus another.
To solve an incompletely defined problem, we can make assumptions about anything that is not known. For example, we could assume some tradeoff between criteria and see what the result is.
Completely-Defined Problems
Incompletely defined problems are best explained by showing what they are not.
A completely defined problem is a problem where there is a right answer.
To solve a completely defined problem you have some flexibility, but the value of the answer will be the same no matter how you get there. Incompletely defined problems in a classroom setting often become completely defined through the power of the teacher telling students what assumptions to make.
For example, if the teacher just taught a class how to do division with a "remainder," the students assume they should not use fractions to answer the question. If I ask what 3 divided by 2 is and 1 with a remainder of 1 is correct but 3/2 is not, then this is a completely defined problem.
Assumptions
Assumptions fill the gap between an incompletely defined problem and a specific completely defined problem
Assumptions determine some unknown in an incompletely defined problem.
With enough assumptions, we can get a version of a particular incompletely defined problem which is completely defined. Those assumptions could include simplifications of the problem, models of uncertainty, predictions of the future, estimates of preferences, and anything else.
Using the term assumption comes from engineering where we acknowledge if something is known for certain or not. Models require a lot of elements to be known, many of which we do not in fact know. By explicitly stating those unknowns engineers can discuss if the assumption is appropriate or not. If not, we often need to consider different methods.