A hypothesis underpinning why I think the selection theorems research paradigm is very promising.
All intelligent systems in the real world are the products of constructive optimisation processes[5]. Many nontrivial properties of a systems can be inferred by reasoning in the abstract about what objective function the system was selected for performance on, and its selection environment[6].
We can get a lot of mileage simply by thinking about what reachable[7] features were highly performant/optimal for a given objective function in a particular selection environment.
For example, many observed values in humans and other mammals[8] (e.g. fear, play/boredom, friendship/altruism, love, etc.) seem to be values that were instrumental for increasing inclusive genetic fitness (promoting survival, exploration, cooperation and sexual reproduction/survival of progeny respectively). In principle, such convergent values are deductible apriori given an appropriate theoretical framework.
Furthermore, given empirically observed regularities/convergent features in different intelligent systems, we can infer that there was positive selection for said features in the selection environment and speculate about what objective functions/performance metric was responsible for said selection.
A hypothesis underpinning why I think the selection theorems research paradigm is very promising.