| Sumario: | John Pollock aims to construct a theory of rational decision making for real agents--not ideal agents. Real agents have limited cognitive powers, but traditional theories of rationality have applied only to idealized agents that lack such constraints. Pollock argues that theories of ideal rationality are largely irrelevant to the decision making of real agents. We need a theory of "real rationality" in its place. Thinking about Acting aims to do just that. The most contentious conclusions of the book are: (1) For purely computational reasons, the primitive evaluative database that real agents employ in evaluating outcomes cannot be preference ranking. An agent's evaluative database must instead assign real numbers to outcomes. It is argued that, contrary to initial appearances, this is psychologically plausible. (2) Subjective probability makes no sense when we are talking about real (resource bounded) agents. Rational decision making must instead be based on a species of objective probability, and it is discussed how this can be done. (3) Classical decision theory is based on the optimality principle, according to which rationality dictates choosing actions that constitute optimal solutions to practical problems. It is argued that the optimality prescription is wrong, for a number of reasons: (a) actions cannot be chosen in isolation--they must be chosen as parts of plans; (b) we cannot expect real agents to find optimal plans, because there are infinitely many alternatives to survey; (c) plans cannot be evaluated in terms of their expected values anyway, because different plans can be of different scopes. Pollock constructs an alternative theory that accommodates these difficulties. This theory is intended not only to solve the philosophical problems, but also to provide the basis for an implemented system of decision-theoretic planning in artificial intelligence.
|
|---|
