Summary of PhD thesis, Computer Science Department, Stanford University, January, 1980,Stanford Rep. Nos. HPP-80-2, STAN-CS-80-784

Two premises, reflected in the title, underlie the perspective from which I will consider research in natural language processing in this article. First, progress on building computer systems that process natural languages in any meaningful sense (i.e., systems that interact reasonably with people in natural language) requires considering language as part of a larger communicative situation. Second, as the phrase “utterance and objective” suggests, regarding language as communication requires consideration of what is said literally, what is intended, and the relationship between the two.

The Heuristic Programming Project of the Stanford University Computer Science Department is a laboratory of about fifty people whose main goals are to model the nature of scientific reasoning processes in various types of scientific problems and various areas of science and medicine; and to construct expert systems — programs that achieve high levels of performance on tasks that normally require significant human expertise for their solution.

ACM Computing Survey 12:213-253

"In the fall of 1978 we decided to produce a special issue of the SIGART Newsletter devoted to a survey of current knowledge representation research. We felt that there were twe useful functions such an issue could serve. First, we hoped to elicit a clear picture of how people working in this subdiscipline understand knowledge representation research, to illuminate the issues on which current research is focused, and to catalogue what approaches and techniques are currently being developed. Second -- and this is why we envisaged the issue as a survey of many different groups and projects -- we wanted to provide a document that would enable the reader to acquire at least an approximate sense of how each of the many different research endeavours around the world fit into the field as a whole. It would of course be impossible to produce a final or definitive document accomplishing these goals: rather, we hoped that this survey could initiate a continuing dialogue on issues in representation, a project for which this newsletter seems the ideal forum. It has been many months since our original decision was made, but we are finally able to present the results of that survey. Perhaps more than anything else, it has emerged as a testament to an astounding range and variety of opinions held by many different people in many different places. The following few pages are intended as an introduction to the survey as a whole, and to this issue of the newsletter. We will briefly summarize the form that the survey took, discuss the strategies we followed in analyzing and tabulating responses, briefly review the overall sense we received from the answers that were submitted, and discuss various criticisms which were submitted along with the responses. The remainder of the volume has been designed to be roughly self-explanatory at each point, so that one may dip into it at different places at will. Certain conventions, however, particularly regarding indexing and tabulating, will also be explained in the remainder of this introduction." ACM SIGART Newsletter No. 70.

Earlier version in IJCAI 6 (1979). Artificial Intelligence, 14 (2), 165-203.

"The contract net protocol has been developed to specify problem-solving communication and control for nodes in a distributed problem solver. Task distribution is affected by a negotiation process, a discussion carried on between nodes with tasks to be executed and nodes that may be able to execute those tasks. We present the specification of the protocol and demonstrate its use in the solution of a problem in distributed sensing. The utility of negotiation as an interaction mechanism is discussed. It can be used to achieve different goals, such as distributing control and data to avoid bottlenecks and enabling a finer degree of control in making resource allocation and focus decisions than is possible with traditional mechanisms." IEEE Transactions on Computers C-29(12):1104-1113. PDF: http://www.reidgsmith.com/The_Contract_Net_Protocol_Dec-1980.pdf.

A program called "AM" is described which carries on simple mathematics research,defining and studying new concepts under the guidance of a large body ofheuristic rules. The 250 heuristics communicate via an agenda mechanism, aglobal priority queue of small tasks for the program to perform, and reasons whyeach task is plausible (for example, "Find generalizations of 'primes', because'primes' turned out to be so useful a concept"). Each concept is represented asan active, structured knowledge module. One hundred very incomplete modulesare initially supplied, each one corresponding to an elementary set-theoreticconcept (for example, union). This provides a definite but immense space whichAM begins to explore. In one hour, AM rediscovers hundreds of common concepts(including singleton sets, natural numbers, arithmetic) and theorems (for example,unique factorization).Summary of Ph.D. dissertation.Hayes, J.E., D. Michie, and L. I. Mikulich (Eds.), Machine Intelligence 9, Ellis Horwood.

Cognitive Science 3(3):177-212.

To choose their actions, reasoning programs must be able to make assumptions and subsequently revise their beliefs when discoveries contradict these assumptions. The Truth Maintenance System (TMS) is a problem solver subsystem for performing these functions by recording and maintaining the reasons for program beliefs. Such recorded reasons are useful in constructing explanations of program actions and in guiding the course of action of a problem solver. This paper describes (1) the representations and structure of the TMS, (2) the mechanisms used to revise the current set of beliefs, (3) how dependency-directed backtracking changes the current set of assumptions, (4) techniques for summarizing explanations of beliefs, (5) how to organize problem solvers into "dialectically arguing" modules, (6) how to revise models of the belief systems of others, and (7) methods for embedding control structures in patterns of assumptions. We stress the need of problem solvers to choose between alternative systems of beliefs, and outline a mechanism by which a problem solver can employ rules guiding choices of what to believe, what to want, and what to do.Artificial Intelligence 12(3):231-272