Models of Bounded Rationality

Thomas Icard

  • Area: LoCo
  • Level: A
  • Week: 1
  • Time: 09:00 – 10:30
  • Room: C2.01


The dominant normative theories of inference, reasoning, and decision making typically work at a level of abstraction that eclipses the resource limitations of actual, physically instantiated agents. Happily, this level of abstraction often results in the sorts of elegant theorizing familiar from logic, semantics, decision theory, game theory, and related areas. Meanwhile, practical business of computational implementation, concrete cognitive and linguistic modeling, and applications usually requires shortcuts, heuristics, and approximations to those normative theories, which often diverge quite radically from what is prescribed. Theorists across the disciplines have been well aware of this mismatch between theory and practice, at least since Herbert Simon’s seminal work. While there has been much ink spilt on the topic, as well as a number of promising research programs exploring bounded rationality from different perspectives, the theory of bounded rationality has not reached a consensus view.

The aim of this course is to cover some of the most successful and fruitful approaches to bounded rationality—from logic, artificial intelligence, cognitive science, and game theory—with an eye toward the possibility of unification. Topics will include:

  • Classical perspectives by H. Simon and others.
  • Bounded rationality in game theory (A. Rubinstein, etc.).
  • Ecological approaches (G. Gigerenzer et al.).
  • Bounded Optimality and Metareasoning in AI.
  • Approaches inspired by ideas from thermodynamics and information theory.
  • Applications of these ideas to methodological problems in cognitive science.


Additional References