■Although knowledge representation is one of thecentral and, in some ways, most familiar concepts in AI, the most fundamental question aboutit—What is it?—has rarely been answered directly. Numerous papers have lobbied for one oranother variety of representation, other papershave argued for various properties a representation should have, and still others have focusedon properties that are important to the notion ofrepresentation in general.In this article, we go back to basics to addressthe question directly. We believe that the answercan best be understood in terms of five importantand distinctly different roles that a representationplays, each of which places different and, attimes, conflicting demands on the properties arepresentation should have. We argue that keeping in mind all five of these roles provides a usefully broad perspective that sheds light on somelong-standing disputes and can invigorate bothresearch and practice in the field.What is a knowledge representation?We argue that the notion can bestbe understood in terms of five distinct roles that it plays, each crucial to thetask at hand:First, a knowledge representation is mostfundamentally a surrogate, a substitute for thething itself, that is used to enable an entity todetermine consequences by thinking ratherthan acting, that is, by reasoning about theworld rather than taking action in it. Second, it is a set of ontological commitments, that is, an answer to the question, Inwhat terms should I think about the world?Third, it is a fragmentary theory of intelligent reasoning expressed in terms of threecomponents: (1) the representation’s fundamental conception of intelligent reasoning,(2) the set of inferences that the representation sanctions, and (3) the set of inferencesthat it recommends.Fourth, it is a medium for pragmaticallyefficient computation, that is, the computational environment in which thinking isaccomplished. One contribution to this pragmatic efficiency is supplied by the guidancethat a representation provides for organizinginformation to facilitate making the recommended inferences.Fifth, it is a medium of human expression,that is, a language in which we say thingsabout the world.Understanding the roles and acknowledging their diversity has several useful consequences. First, each role requires somethingslightly different from a representation; eachaccordingly leads to an interesting and different set of properties that we want a representation to have. Second, we believe the roles provide aframework that is useful for characterizing awide variety of representations. We suggestthat the fundamental mind set of a representation can be captured by understanding howit views each of the roles and that doing soreveals essential similarities and differences.Third, we believe that some previous disagreements about representation are usefullydisentangled when all five roles are givenappropriate consideration. We demonstratethe clarification by revisiting and dissectingthe early arguments concerning frames andlogic.Finally, we believe that viewing representations in this way has consequences for bothresearch and practice. For research, this viewprovides one direct answer to a question offundamental significance in the field. It alsosuggests adopting a broad perspective onArticlesSPRING 1993 17What Is a KnowledgeRepresentation?Randall Davis, Howard Shrobe, and Peter SzolovitsCopyright © 1993, AAAI. All rights reserved. 0738-4602-1993 / $2.00This paper is copyrighted by the American Association for Artificial Intelligence. All rights reserved. Role 1: A Knowledge Representation Is a SurrogateAny intelligent entity that wants to reasonabout its world encounters an important,inescapable fact: Reasoning is a process thatgoes on internally, but most things it wantsto reason about exist only externally. A program (or person) engaged in planning theassembly of a bicycle, for example, mighthave to reason about entities such as wheels,chains, sprockets, and handle bars, but suchthings exist only in the external world.This unavoidable dichotomy is a fundamental rationale and role for a representation: It functions as a surrogate inside thereasoner, a stand-in for the things that existin the world. Operations on and with representations substitute for operations on thereal thing, that is, substitute for direct interaction with the world. In this view, reasoningitself is, in part, a surrogate for action in theworld when we cannot or do not (yet) wantto take that action.1Viewing representations as surrogates leadsnaturally to two important questions. Thefirst question about any surrogate is itsintended identity: What is it a surrogate for?There must be some form of correspondencespecified between the surrogate and itsintended referent in the world; the correspondence is the semantics for the representation.The second question is fidelity: How closeis the surrogate to the real thing? Whatattributes of the original does it capture andmake explicit, and which does it omit? Perfect fidelity is, in general, impossible, both inpractice and in principle. It is impossible inprinciple because any thing other than thething itself is necessarily different from thething itself (in location if nothing else). Putthe other way around, the only completelyaccurate representation of an object is theobject itself. All other representations areinaccurate; they inevitably contain simplifying assumptions and, possibly, artifacts.Two minor elaborations extend this viewof representations as surrogates. First, itappears to serve equally well for intangibleobjects as well as tangible objects such as gearwheels: Representations function as surrogates for abstract notions such as actions,processes, beliefs, causality, and categories,allowing them to be described inside anentity so it can reason about them. Second,formal objects can of course exist inside themachine with perfect fidelity: Mathematicalentities, for example, can be captured exactly,precisely because they are formal objects.Because almost any reasoning task willwhat’s important about a representation, andit makes the case that one significant part ofthe representation endeavor—capturing andrepresenting the richness of the naturalworld—is receiving insufficient attention. Webelieve that this view can also improve practice by reminding practitioners about theinspirations that are the important sources ofpower for a variety of representations.Terminology and PerspectiveTwo points of terminology assist our presentation. First, we use the term inference in ageneric sense to mean any way to get newexpressions from old. We rarely talk aboutsound logical inference and, when doing so,refer to it explicitly.Second, to give them a single collectivename, we refer to the familiar set of basic representation tools, such as logic, rules, frames,and semantic nets, as knowledge representation technologies.It also proves useful to take explicit note ofthe common practice of building knowledgerepresentations in multiple levels of languages, typically, with one of the knowledgerepresentation technologies at the bottomlevel. Hayes’s (1978) ontology of liquids, forexample, is at one level a representation composed of concepts like pieces of space, withportals, faces, sides, and so on. The languageat the next, more primitive (and, as it turnsout, bottom) level is first-order logic, where,for example, In(s1,s2) is a relation expressingthat space s1is contained in s2.This view is useful in part because it allowsour analysis and discussion to concentratelargely on the knowledge representation technologies. As the primitive representationallevel at the foundation of knowledge representation languages, those technologiesencounter all the issues central to knowledgerepresentation of any variety. They are alsouseful exemplars because they are widelyfamiliar to the field, and there is a substantialbody of experience with them to draw on.What Is a Knowledge Representation?Perhaps the most fundamental questionabout the concept of knowledge representation is, What is it? We believe that the answeris best understood in terms of the five fundamental roles that it plays.a representation… functions asa surrogateinside the reasoner…Articles18 AI MAGAZINEThis paper is copyrighted by the American Association for Artificial Intelligence. All rights reserved. encounter the need to deal with naturalobjects(that is, those encountered in the realworld) as well as formal objects, imperfectsurrogates are pragmatically inevitable.Two important consequences follow from
the inevitability of imperfect surrogates. One
consequence is that in describing the natural
world, we must inevitably lie, by omission at
least. At a minimum, we must omit some of
the effectively limitless complexity of the natural world; in addition, our descriptions can
introduce artifacts not present in the world.
The second and more important consequence is that all sufficiently broad-based reasoning about the natural world must
eventually reach conclusions that are incorrect, independent of the reasoning process
used and independent of the representation
employed. Sound reasoning cannot save us: If
the world model is somehow wrong (and it
must be), some conclusions will be incorrect,
no matter how carefully drawn. A better representation cannot save us: All representations are imperfect, and any imperfection can
be a source of error.
The significance of the error can, of course,
vary; indeed, much of the art of selecting a
good representation is in finding one that
minimizes (or perhaps even eliminates) error
for the specific task at hand. But the unavoidable imperfection of surrogates means that we
can supply at least one guarantee for any
entity reasoning in any fashion about the
natural world: If it
đang được dịch, vui lòng đợi..
