[ CogSci Summaries home | UP | email ]

Russell, S. (1995). Rationality and Intelligence. Artificial Intelligence, 94, 57–77.

  author =	"Stuart J. Russell",
  title =	"Rationality and Intelligence",
  journal =	"Artificial Intelligence",
  year =  "1997",
  volume =	"94",
  number =	"",
  pages =	"57--77",

Author of the summary: Neil Conway, 2006, neilc NO-SPAM samurai.com.

Cite this paper for:

This paper is available online: http://www.cs.berkeley.edu/~russell/papers/aij-cnt.pdf.

Defining intelligence

Perfect rationality

Calculative rationality

Metalevel rationality

Bounded optimality

Future work

The remainder of the paper sketches some areas for future research related to bounded optimality and ABO agents.

Calculative rationality improvements

ABO proofs

Optimization techniques

Offline vs. online

Variable agent architecture

Agents that learn can adapt their behavior, but prior research on adaption has usually applied learning techniques only to select components of the agent. Given finite memory, there are problems where "almost complete revision of the agent's mental structure is needed to achieve good performance" [p20] -- that is, learning techniques would be applied to the entirety of the agent's architecture.

Summary author's notes:

Page numbers are from downloaded version.
Back to the Cognitive Science Summaries homepage
Cognitive Science Summaries Webmaster:
JimDavies (jim@jimdavies.org)