BEINGS: Knowledge as Interacting Experts

Knowledge may be organized as a community of in t f iac t ing modules Each module is granted a complex st iucture, to simulate a particular expert in some small domain An extended analogy is drawn to a group of cooperating human specialists Based on this, an internal constraint is imposed on the modules Then structure must be standard over the entire community Some advantages of a un i fo rm formalism are thereby preserved. An experimental community was implemented for the task domain of automatic programming It has managed to synthesize a few inductive inference LISP programs, nonformal ly. f rom specific restricted dialogues wi th a human user 1. Experts and Beings Consider an interdisciplinary enterprise, attempted by a community of human expeits who are specialists in and only in .. then own fields What modes of interactions wi l l be productive? The dominant paradigm might well settle into questioning and an sinning each other Instead of a cha i rman, suppose the group adopts rules for gaining the floor, what a speaker may do, and how to resolve disputes When a topic is being considered, one or two experts might recognize it and speak up In the course of their exposition they might need to call on other specialists Th is might be by name, by specialty, or simply by posing a new sub-question and hoping someone rould recognize Ins own relevance and volunteer a suggestion Such tiansfers would be more common at the beginning, when the task is (by assumption) too geneial for any one member to compiehend As the questions f o r m on more specific issues, single indiv iduals should be able to supply complete solutions If the task is to construct something, (hen the activities of the experts should not be strictly verbal. Ol ten. one wil l recognize his relevance to the current situation and ask to tto something, clarify or modi fy or (raiely) create Wha t would it mean to simulate the above activity? Imagine seveial little programs, each one modelling a different expert What should each program, called a Being, be capable of It must possess a coipus of specific facts and strategies for its designated speciality It must interact via questioning and answering other Beings. Each Being should be able to recognize when it is lelevant It must set up and alter structures, just as the human specialists do. Let us ie tu in to our meeting of human expeits. To be more concrete, suppose their task is to design and code a large computer p iog iam a concept formation system[2]. Experts who wi l l be useful include scientific programmers, non programming psychologists, system hackers, and management personnel What happens in the ensuing session? When an expert participates, he wi l l either be aiding a co l l egae in sume difficulty of else transferring a t iny, customized bit of his expertise (facts about his field) into a piogrammed function which can do something T h e final code reflects the member s' knowledge, m that sense One way the session might pioceed is for the specialists to actually do the concept formation task As they become fami l iar w i th what part of their own expertise is being called upon, and in what ways, they can begin to isolate it. When it is clear precisely what each is doing, they can take their extracted bits of knowledge, organize them, formalize them, and program them {A conscious erfoit along these lines was made in [8], where expeits gradually leplaced themselves by programs Instead of discussing how to write a speech p tog iam, they / speech recognition, unt i l each one could introspect sufficiently into his own activities to formalize them For our task, one expects the psychologists to dominate the eaily discussions, later yielding to programmers T h e project sponsor might be passive, submitt ing a single specification order for the program, or active, pai t ic ipat ing in the work as a (somewhat priveleged) member of the team. T h i s ind iv idua l is the one who wants the final product, hence wi l l be called the user How could Beings do this? Theie would be some litt le p r o y i i m containing information about CONCEPT FORMATION (much mote than would be used in wri t ing any single concept formation piogiam), anothei Being who knows how to manape a group to WRITE PROGRAMS, and many lower Ievel specialist's, for example INFO-OBTAINFR, TEST, MODIFYDATA STRUCTURE, UNTIl -LOOP, VISUAL -PF.RCI PTION, AVOIDCONTRADICTION, PROPOSE-PL AUSIHt.E-NAME Like the human specialists, the Beings would contain far too much in fo imat ion , far too inefficiently lepresented. to be able to say "we ourselves constitute the desired p iogtam'" They would have to discuss, and perhaps cany out. the concept format ion task They would write specialized versions of themselves, programs which could do exactly what the Beings d id to carry out the task, no more not less (although they would hopeful ly take much less time, be more customized). T h i s activity is relected to in the sequel as automatic programming Some Beings ( eg , TEST) may have several distinct, stieamlined fractions of themselves in the final program. Beings ( e g , PROPOSE-PLAUSIBLE-NAME) which only aided other Beings may not have any con elates in the final synthesized code An experimental system, PUP6, was designed and part ial ly implemented PUP6 synthesized a concept format ion p i og iam (similar to [7]). but the user. who is human, must come up wi th certain specific answers to some of the Beings' ct i t ical queues A grammatical infeience ptogram and a simple property list maintenance routine were also generated. A few new Beings had to be added to PUP6 to synthesize them T h e next section illustrates how the experts might have cooperated on the task of wri t ing the concept format ion p togram Section ? describes the program they produced. Next comes the Being hypothesis complex but standard anatomy Later sections explain this, both theoretically and by examin ing the behavior of the actual PUP6 pool of 100 Beings T h e appendix exhibits a typical Being.