Optimising and Satisficing under Partial Ignorance
Frans Voorbraak
Abstract:
In this paper, we study decision making in situations where the outcomes of
the options are (in general) uncertain, without making the assumption that
this uncertainty can be exactly quantified by means of a (single)
probability measure. In such situations of partial ignorance, the
traditional notion of optimising by maximising expected utility is in
general rather weak, and we will discuss some proposals to augment or
refine the criterion of maximal expected utility under partial ignorance.
We will argue that one interesting possible refinement, which is related to
the minimax regret rule originally proposed for decision under uncertainty,
is best viewed as a rule for satisficing rather than optimising. To make
this argument more precise, we first propose a formalisation of the notion
of satisficing under risk.