Models of Bounded Rationality for Inference
Dissertation, The University of Chicago (
1997)
Copy
BIBTEX
Abstract
How do organisms make accurate inferences about their environments when the the information available to them and the faculties they reason with are bounded? In contrast to previous approaches which view the mind as either a computational demon or a careless user of shoddy heuristics, the satisficing view of cognition, put forth by Herbert Simon, suggests that simple minds can do well with simple algorithms if the minds and algorithms are adapted to natural environments. In this thesis, I detail a satisfying property called one-reason decision making. In the first part, I investigate whether simple satisfying mechanisms are capable of making good inferences. In the second part, I attempt to find out whether the recognition principle, a mechanism of one-reason decision making, is active in human thought. This research suggests that complex reasoning does not necessarily have to arise from complex mechanisms. If organisms satisfy, their minds may be smart and simple at once.