To t~hoose their action s, rea soning programs must be able to make assumpt ions and subsequently revi se their beliefs when discoveries contradict these assumptions. The Trut h M~ u'rn '~~e Sysi 'o'i (TMS) is a problem solver subsystem for performing these functions by recording arid maintaining the rea sons for program beliefs. Such recorded reason s are useful In constructin g exp lanauons of program actions and In guiding the course of action of a problem solver This paper describes (I) the representat.otis and structure of the TMS . (2) the me-hanisms used to revise the current set of belIefs , (3) how dependency-directed backtrackin g changes the current set of assumpil ons . (4) techniques for summarizin g e~1'lanauon s of beliefs, (yr) how to organize problem solvers Into dia lectlca fly ar guing modules. ~F) how to revise models of the belief systems of others, arid (' 1) met hods for embedding control stru cTu res in patterns of assumpt ions We stress the need of problem solvers to choose bet ween alternative systems of belief s. and outline a mechanism by which a problem solver can employ rules guiding choices of what to believe, wha t to want. and what to do This research w a s conducted at the Artificial Intelligence Laboratory of the Ma ssachus etts Inst itute of Technology Support for the Laboratory's art ificial intelligence research is provided in part by the A dvanced Research Proj ects Agency of the Department of Defense under Office of Na val Research contract number N000I4-75-C-0643, and In part by NSF grant MCS77 O4*28 D D C I DI!TRIBUTION STATEKINT A fl~~~E flfl 1Ef1 App,o~.d tot pub~%c teleossI ~ aic 17 1919 '~-'.~~~~n

No abstract

The economic theory of rationality promises to equal mathematical logic in its importance for the mechanization of reasoning. We survey the growing literature on how the basic notions of probability, utility, and rational choice, coupled with practical limitations on information and resources, influence the design and analysis of reasoning and representation systems.

No abstract

Ceteris paribus preference statements concisely represent preferences over outcomes or goals in a way natural to human thinking. Many decision making methods require an efficient method for comparing the desirability of two arbitrary goals. We address this need by presenting an algorithm for converting a set of qualitative ceteris paribus preferences into a quantitative utility function. Our algorithm is complete for a finite universe of binary features. Constructing the utility function can, in the worst case, take time exponential in the number of features. Common forms of independence conditions reduce the computational burden. We present heuristics using utility independence and constraint based search to achieve efficient utility functions.