It seems to me that every heuristic presupposes some form of inductive bias in the form of simplifying assumptions that are arrived at via induction on past experience. A good heuristic makes assumptions that simplify (or optimize) the task it is designed for, with at most minor inaccuracies for the vast majority of common instances of the task (until a black swan arrives, at least). Determining which assumptions to use, in turn, involves relying on some more general form(s) of inductive bias, like Occam’s Razor or Minimum Description Length.
It seems to me that every heuristic presupposes some form of inductive bias in the form of simplifying assumptions that are arrived at via induction on past experience. A good heuristic makes assumptions that simplify (or optimize) the task it is designed for, with at most minor inaccuracies for the vast majority of common instances of the task (until a black swan arrives, at least). Determining which assumptions to use, in turn, involves relying on some more general form(s) of inductive bias, like Occam’s Razor or Minimum Description Length.