K. Myers and D. Smith. The persistence of derived information. In Proc. of the 7th National Conference on Artificial Intelligence, pages 496--500, 1988.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....we mean actions whose effects can be described by a set of literals. Analogously for the constraints. As a consequence at the formal level, as far as we have to formalize applications of this kind, we do not have to deal with all the complications caused by disjunctive information (see for example [36, 34]) e.g. ramifications are not possible. As a consequence at the application level, the frame problem can be and indeed it is solved by simply updating the knowledge base of the scenario. Such a theory can be used to verify system properties and plans. We propose in Section 4.3 an example ....

K. Myers and D. Smith. The persistence of derived information. In Proc. of the 7th National Conference on Artificial Intelligence, pages 496--500, 1988.

.... a criticism leveled on Ginsberg s approach involving its essentially syntactic foundations [ Winslett, 1988 ] Consider, for example, the case of ramifications: if ff makes OE true and is a ramification of OE then if some action is performed after ff that makes OE false, will no longer hold [ Myers and Smith, 1988 ] To deal with such problems, model based revision schemes were proposed based on cardinality of differences between models of the counterfactual antecedent and models of the counterfactual consequent [ Dalal, 1988 ] 2.4.1 Belief updating and Gardenfors s triviality theorem Winslett [ 1988 ] ....

....anything lumped together with it. Example (4c) is, then, a contingent generalization. Kratzer [ 1981 ] presents a theory of counterfactuals that addresses these points. Lewis [ 1981 ] examines his approach given those requirements identified by Kratzer. 6 See also the green cheese problem of [Myers and Smith, 1988]. 13 2.5 Which change operation The above discussion suggests that some syntactic form of updating should ground the semantics of counterfactuals, as applied to problems in causal reasoning. Approaches that have defined update in terms of changes to a belief base which has been converted to ....

[Article contains additional citation context not shown here]

Myers, Karen L. and Smith, David E. 1988. The persistence of derived information. In AAAI 88. 496--500.

....may be crucial in properly reasoning with such constraints. This needs to be investigated further. 141 9.2.3 Dealing with Unwanted Persistence of Fluents Though we usually want the values of fluents to persist, there are situations where we do not want this to happen. The following example [Myers and Smith, 1988] illustrates this. Example 2. If an object is on the table, it is baby proof. Initially, Block A is on the table. Picking up an object causes it to be not on the table. Question: Is Block A baby proof after picking it up Intuitively, the answer to this question is Unknown . But, if we ....

Karen Myers and David Smith. The persistence of derived information. In Proc. AAAI-88, pages 496--500, 1988.

....safe or not. Note also that closing and opening the closet door will only affect the status of objects in the closet and not those at other locations. Example C. 33 (Extended Baby Protection Scenario, Giunchiglia et al. 7] This example is a modification of one originally due to Myers and Smith [20]. Action Symbols: Close(door) Open(door) P ut(thing; location) Remove(thing; location) Feature Symbols: closed(door) safe(thing) loc : thing location Object Symbols: door1 : door, gun : thing, toy : thing, table : location, closet : location, f loor : location 46 obs1 [0] loc(toy) f ....

K. L. Myers and D. E. Smith. The persistence of derived information. In Proc. AAAI-93, pages 496--500, 1988.

....p. 15] Taking the transitive closure in this way makes his analysis vastly more plausible, but it also makes it much less useful as a direct justification for minimisation based approaches to the frame problem. The following pair of examples is due (in slightly different form) to Myers and Smith [24] (see also [7] 28, pp. 289f. Example 3 (Wet baby) The scenario is this: There is one agent (a baby) which has two actions, crawl in and crawl out, which, respectively, take the baby into and out of a lake. The baby is initially dry; crawling into the lake causes it to be wet, but crawling ....

K.L. Myers and D.E. Smith. The persistence of derived information. In AAAI88, pages 496--500, 1988. 58

....consistent with the action consequences. The most classical problem in this context concerns the non persistence of fluents, which have been caused by the execution of an action and disappear together with the action result which caused them. Consider the following example due to Myers and Smith [28]. The world is characterized by two fluents specifying whether some object (the heavy hammer) is safe (from the baby) and whether it is on the table. There is one constraint, namely, whenever an object is on the table, then it is safe. There are two actions, namely putting the hammer on the ....

K. Myers and D. Smith. The persistence of derived information. In Proceedings of the 8th National Conference on Artificial Intelligence, pages 496--500, 1988.

....logic was introduced in [Tur98] UCL extends the recently introduced causal theories formalism of McCain and Turner [MT97] which shares its underlying motivations. The fundamental distinction in UCL 3 Inertia rules of essentially this form have been entertained previously in the literature [Rei80, MS88], but without substantial success. The specific proposals are not complete enough to analyze in any detail. We do discuss throughout the first part of the dissertation several interacting factors that contribute to the success of our approach. One such factor is that, while default logic is ....

....and theory update. As described previously, work on defining possible next states is still explicitly directed toward the problem of reasoning about action, but in a setting where some of the complexities of action domains are ignored in order to focus more directly on the frame problem itself [GS88, MS88, Win88, BH93, Bar95, MT95b, Thi95a]. The frame problem is also confronted in the still more abstract setting of theory update [KM91, MT93b, Bar94, MT94, MT95a, PT95, PT97, MT98a] Recall that one example is the formalism of revision programming [MT94, MT98a] which can be understood as a precurser to the causal approaches presented ....

Karen Myers and David Smith. The persistence of derived information. In Proc. AAAI-88, pages 496--500, 1988.

....with simple actions and constraints. By simple actions we mean actions whose effects can be described by a set of literals. Analogously for the constraints. As a consequence at the formal level, we have not to deal with all the complications caused by disjunctive information (see for example (Myers and Smith 1988; McCain and Turner 1995) e.g. ramifications are not possible. As a consequence at the application level, the frame problem can be and indeed it is solved by simply updating the knowledge base of the scenario. About the formalization, it seems that much of the work in theories of action ....

K. Myers and D. Smith. The persistence of derived information. In Proc. of the 7th National Conference on Artificial Intelligence, pages 496-- 500, 1988.

....the constraints imposed by any of the main options discussed, and thus from the strictures of existing proposals in the literature; instead, we are free to choose the persistence policy for a given non deterministic action that best captures its expected behavior. 5 Varying persistence policies Myers and Smith [1988] present a number of examples in which the reasonableness of treating a fact as persistent by default depends on the way in which we came to know that fact. Similarly, it is easy to find examples in which some fact should be treated as persistent or not in virtue of what causes it. For example, ....

Myers, Karen L. and Smith, David E. 1988. The persistence of derived information. In Proceedings of the Seventh Conference of the AAAI.

....information into a knowledge base while preserving consistency. Recently, belief revision has received a lot of attention in AI, 1 which led to a number of different proposals for different applications (Ginsberg 1986; Ginsberg, Smith 1987; Dalal 1988; Gardenfors, Makinson 1988; Winslett 1988; Myers, Smith 1988; Rao, Foo 1989; Nebel 1989; Winslett 1989; Katsuno, Mendelzon 1989; Katsuno, Mendelzon 1990; Doyle 1990) Most of this research has been considerably influenced by approaches in philosophical logic, in particular by Gardenfors and his colleagues (Alchourr on, Gardenfors, Makinson 1985; Gardenfors ....

....base revision coincides with revision generated by epistemic relevance in the sense that the class of prioritized base revisions is identical with the class of revisions generated by epistemic relevance. This abstract view on syntax based revision may also answer some of the questions raised by Myers and Smith (1988). They observed that sometimes base revision does not seem to be the appropriate operation because some derived information turns out to be more relevant than the syntactically represented sentences in a belief base, and we get the wrong results when using base revision. However, there is no magic ....

Myers, K. L., Smith, D. E. (1988): The Persistence of Derived Information, in: Proceedings of the 7th National Conference of the American Association for Artificial Intelligence, Saint Paul, Minn., 496--500.

....standing at location L 1 , where this robot is holding an ice cream cone. Suppose the robot now moves to location L 2 . Does the ice cream cone s absolute position persist, or does the fact that it is being held by the robot persist (equivalent scenarios have been discussed by many authors, e.g. [4, 16, 18]) Based on our understanding of the domain, we would probably guess that after the robot moves, it will continue to hold the ice cream cone. Yet, if we simply write down the domain constraint: Holds(Holding(x; y) s) Holds(At(x; l) s) oe Holds(At(y; l) s) then both persistence possibilities ....

K.L. Myers and D.E. Smith, The persistence of derived information, in: Proceedings AAAI-88, St. Paul, MN (1988) 496--500.

....may depend on their history, and not only on their current values. 1 Introduction This paper is about the ramification problem in the theory of commonsense reasoning, that is, about the problem of determining the indirect effects of an action. More specifically, we are interested in what Myers and Smith [1988] called the persistence of derived information in deciding whether the indirect effects of an action should be presumed to persist, or, in other words, whether they are subject to the commonsense law of inertia. Compare two examples: Example 1 [ Crawford, 1994 ] If you are in the lake then ....

....effects of an action should be presumed to persist, or, in other words, whether they are subject to the commonsense law of inertia. Compare two examples: Example 1 [ Crawford, 1994 ] If you are in the lake then you are wet. Jumping in the lake has an indirect effect getting wet. Example 2 [ Myers and Smith, 1988 ] If an object is on the table then it is not dangerous for the baby crawling on the floor. Putting an object on the table has an indirect effect making it safe. The examples look similar. Consider, however, what happens in each case if the action under consideration is followed by an action ....

[Article contains additional citation context not shown here]

Karen Myers and David Smith. The persistence of derived information. In Proc. AAAI-88, pages 496--500, 1988.

....a database 2 , and after a series of updates the database ends up at time t in a state in which it can be represented by some formula t which is equivalent to 1 ; according to (U4) updating t with should result in a database equivalent to 2 . Clearly, this is not always reasonable. Myers and Smith [ 1988 ] present a number of examples in which the reasonableness of treating a fact as persistent by default depends on the way in which we came to know that fact. Similarly, it is easy to find examples in which some fact should be treated as persistent or not in virtue of what causes it. For example, ....

Karen L. Myers and David E. Smith. The persistence of derived information. In Proceedings of the Seventh Conference of the AAAI, 1988.

....the following constraints: ontable(cup) oe safe(cup) 6) ontable(doll) oe safe(doll) 7) Intuitively, 6) and (7) say that both the cup and doll on the table are safe to the baby. This domain reveals a problem so called the persistence of derived information first examined by Myers and Smith [15]. Specifically, safe(cup) and safe(doll) is derived information from ontable(cup) and ontable(doll) respectively) But there are different requirements for the persistence of safe(cup) and safe(doll) with respect to the change of ontable(cup) and ontable(doll) respectively. For example, ....

K. Myers and D. Smith, The persistence of derived information. In Proceedings of the Seventh National Conference on Artificial Intelligence (AAAI'88). Morgan Kaufmann Publishers, Inc. (1988) 496--500.

....into a knowledge base while preserving consistency. Recently, belief revision has received a lot of attention in AI, 1 which led to a number of different proposals for different applications [ Ginsberg, 1986; Ginsberg and Smith, 1987; Dalal, 1988; Gardenfors and Makinson, 1988; Winslett, 1988; Myers and Smith, 1988; Rao and Foo, 1989; Nebel, 1989; Winslett, 1989; Katsuno and Mendelzon, 1989; Katsuno and Mendelzon, 1990; Doyle, 1990 ] Most of this research has been considerably influenced by approaches in philosophical logic, in particular by Gardenfors and his colleagues [ Alchourr on et al. 1985; ....

....based on the relevance ordering defined above leads to the same results as the prioritized base revision. This means that prioritized base revision coincides with revision based on epistemic relevance. This abstract view on syntax based revision may also answer some of the questions raised by Myers and Smith [ 1988 ] They observed that sometimes base revision does not seem to be the appropriate operation because some derived information turns out to be more relevant than the syntactically represented sentences in a belief base, and we get the wrong results when using base revision. However, there is no ....

Karen L. Myers and David E. Smith. The persistence of derived information. In Proceedings of the 7th National Conference of the American Association for Artificial Intelligence, pages 496--500, Saint Paul, Minn., August 1988.

....update jlli.tex Date: May 22, 1994 Time: 18:17 Deriving Properties of Belief Update from Theories of Action 15 by some formula t which is equivalent to 1 ; according to (U4) updating t with should result in a database equivalent to 2 . Clearly, this is not always reasonable. Myers and Smith (1988) present a number of examples in which the reasonableness of treating a fact as persistent by default depends on the way in which we came to know that fact. Similarly, it is easy to find examples in which some fact should be treated as persistent or not in virtue of what causes it. For example, ....

Myers, Karen L. and Smith, David E. 1988. The persistence of derived information. In Proceedings of the Seventh Conference of the AAAI.

....t 1 g) where t 1 = Moving(B; C) t 0 . The above example depicts the general idea of using default logic to construct an action theory. However, this method has limitations in some aspects. Myers and Smith indicated that the above method failed to solve the persistence of derived information [Myers and Smith, 1988]. More seriously, it was shown that Reiter s default logic, as well as other classical non monotonic logics, could not solve the temporal projection problem [Hanks and McDermott, 1987] We will address this problem later. Circumscription and Action Theory Circumscription was offered by McCarthy ....

....get the preferred resulting state S 1 if we define predicate layers L(ontable) 2 and L(occupied) 1. The following example concerns the persistence of derived information. Example 2.4 The baby room domain. Let us consider a simple baby room domain that was first considered by Myers and Smith [Myers and Smith, 1988]. The set of domain constraints about this world is as follows: 8x:ontable(x) Phi onf loor(x) 8x:ontable(x) oe liftable(x) 8x:ontable(x) oe safefrombaby(x) Intuitively, we have the following preference: if something is liftable, then it should be still liftable when it is moved out from the ....

K.L. Myers and D.E. Smith. The persistence of derived information. In Proceedings of the Seventh National Conference on Artificial Intelligence, pages 496--500, 1988.

.... on lamp, sw1, sw2 as a result of switching switch 1 to its negative position. PSA makes on lamp and sw2 uncertain. Another type of constraint is only quasi definitional in that C takes the form d l 1 . l n . This is illustrated by the examples below. a. Myers and Smith s example [MS88] has, among others, constraints of the form on table(x) liftable(x) and on table(x) safe to baby(x) Here the problem is illustrated by the impossibility of distinguishing the proper update based on syntax alone. For instance, assuming only one object to reduce atoms to propositional form, ....

Myers, K.L. and Smith, D.E., "The Persistence of Derived Information", Proceedings of the Seventh National Conference on Artificial Intelligence, AAAI-88, 1988.

....[ Simon, 1953; Druzdzel Simon, 1993; Pearl, 1996 ] but does not appeal to probabilities or rankings of any sort. It also combines (propositional) action rules and causal rules in a clear manner, provides a simple way for determining when derived information can be assumed to persist [ Myers Smith, 1988; Mailing address from US and Europe: Hector Geffner, Bamco CCS 144 00, P.O.BOX 02 5322, Miami Florida 331025322, USA. E mail: hector usb.ve. Giunchiglia Lifschitz, 1995 ] and provides another perspective on the distinction between qualification and ramification constraints [ Lin ....

....(e.g. wet oe hard to grasp) and one shot relations in which the effect remains true even after the cause that triggered the effect has vanished (e.g. in water oe wet) In our model we have not made this distinction, yet this distinction is often needed. Consider for instance a scenario from [ Myers Smith, 1988 ] cited in [ Giunchiglia Lifschitz, 1995 ] if an object is put on the table it becomes safe for the baby, yet if removed from the table it won t necessarily remain safe. Such a scenario can be modeled by the temporal rules: put Gamma on table remove Gamma :on table and the causal rule: ....

Myers, K., and Smith, D. 1988. The persistence of derived information. In Proceedings AAAI-88, 496--500.

....and f:on lamp; sw1; sw2g as a result of switching switch 1 to its negative position. PSA makes on lamp and sw2 uncertain. Another type of constraint is only quasi definitional in that C takes the form d l 1 : l n . This is illustrated by the examples below. a) Myers and Smith s example [12] has, among others, constraints like on table(x) liftable(x) andon table(x) safe to baby(x) Here the problem is illustrated by the impossibility of distinguishing the proper update based on syntax alone. For instance, assuming there is only one object to reduce atoms to propositional form, ....

K. L. Myers and D. E. Smith. The persistence of derived information. Proc. of the 7th National Conference on Artificial Intelligence, 1988.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC