In many tasks, such as diagnosis, fault detection and reorienting yourself
within a familiar environment, an observer is faced with a certain amount of
uncertainty about what the true state of the world is. Furthermore, in each of
these tasks the observer is also given a specific goal (e.g., go to the
supermarket when disoriented or cure the patient in medical diagnosis) that they
are attempting to achieve. To achieve these goals the observer must collect
information about the true state of the environment (e.g., where am I) and
approach the goal state using a collection of actions (e.g., rotate left, move
forward in navigation or take a patient's temperature in medical diagnosis).
These actions can have vary in their cost in addition to how reliable that they
are. We have been using the Bayesian framework to study these types of
sequential decision making with uncertainty tasks using Partially Observable
Markov Decision Processes. Using this approach we are working to understand some
of the cognitive limitations associated with human sequential decision making
with uncertainty.
Copyright. All Rights Reserved.