One argument that is sometimes made in papers about heuristic search algorithms is that an algorithm is good because it required very few parameters to be adjusted by the user. The argument is that such algorithms are better because they are easier to use—the user doesn’t require expertise in the use of the algorithm to make effective use of it by choosing parameters well.
This is all very well—but, it could be argued that all such algorithms are doing is substituting explicit parameters for tacit parameters that are part of the structure of the algorithm design. As a result, the user is forced to just work with what is there and has no scope to change the algorithm at all; as such, it isn’t really any better than a parameterised algorithm with the parameters fixed.
Perhaps the solution to this is the well-known approach of parameterised algorithms with adaptive parameter search as part of the optimization process. With modern hardware, this is not to hard. I wonder, though, if there is scope for extending this by taking a more sophisticated approach to the parameter optimization by taking ideas from control theory.