Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
The challenge of crafting intelligible intelligence
Weld D., Bansal G. Communications of the ACM62 (6):70-79,2019.Type:Article
Date Reviewed: Aug 5 2019

This thoughtful review of “intelligible intelligence” work is at the same time a manifesto on the desirability of intelligent systems that have the ability to explain their reasoning, at least to some extent.

The authors first explain why intelligibility matters and define the concept, and then describe a number of so-called inherently intelligible models. Finally, they touch on ways in which intelligibility may be important in artificial intelligence (AI) domains such as planning.

An important concept in the discussion of systems that make binary decisions is that of a foil. Thus, the user might ask the system: “Why P rather than Q?” In this case, Q is the foil. One strength of this view is that it allows for the modeling of context, as different users might invoke a different foil.

Two techniques for intelligible systems in particular are described. The first technique, which can be used for the construction of a system (GA2M), uses nonlinear shape functions and allows for the pairwise interaction of features. The second technique, which can be used to provide intelligibility for inscrutable models, is based on the idea of a local explanation in which a linear model is created for cases near the case of interest.

The paper covers a considerable amount of ground while managing to give enough detail for the reader to be able to grasp the state of intelligible systems construction. It is recommended for anyone interested in the construction of intelligent systems.

Reviewer:  J. P. E. Hodgson Review #: CR146641 (1910-0375)
Bookmark and Share
  Featured Reviewer  
 
General (I.2.0 )
 
 
Learning (I.2.6 )
 
Would you recommend this review?
yes
no
Other reviews under "General": Date
Artificial experts: social knowledge and intelligent machines
Collins H., MIT Press, Cambridge, MA, 1990. Type: Book (9780262031684)
Apr 1 1991
Catalogue of artificial intelligence techniques
Bundy A., Springer-Verlag New York, Inc., New York, NY, 1990. Type: Book (9780387529592)
Aug 1 1991
Knowledge and inference
Nagao M., Academic Press Prof., Inc., San Diego, CA, 1990. Type: Book (9780125136624)
Oct 1 1991
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy