Irving Good. Twenty-seven principles of rationality. In V Godambe and D Sprott, eds, Foundations of Statistical Inference. Toronto: Holt, Rinehart, Winston, 1971.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....Introduction In most real world settings, due to limited time, an agent cannot perform all potentially useful deliberation actions. As a result it will generally be unable to act rationally in the world. This phenomenon, known as bounded rationality, has been a long standing research topic (e.g. [3, 17] ) Most of that research has been descriptive: the goal has been to characterize how agents in particular, humans deal with this constraint. Another strand of bounded rationality research has the normative (prescriptive) goal of characterizing how agents should deal with this constraint. This ....

Irving Good. Twenty-seven principles of rationality. In V Godambe and D Sprott, eds, Foundations of Statistical Inference. Toronto: Holt, Rinehart, Winston, 1971.

....the agent with a perfect strategy is too strong and unrealistic. Therefore, artificial agents cannot be perfectly rational. They cannot manifest the best possible behavior even in relatively simple domains such as chess playing. The failure of classical decision theory led the statistician I. J. Good [1971] to distinguish between perfect rationality, which he called type I rationality, and type II rationality which acknowledges the fact that the agent must deliberate before it can act. This type of rationality requires that the agent maximize its expected utility, taking into account the cost of ....

I. J. Good. Twenty-seven principles of rationality. In Foundations of Statistical Inference, V. P. Godambe and D. A. Sprott (eds.), Toronto: Holt, Rinehart and Winston, 1971.

....solving by the agent itself is normally unavoidable, leading to a deliberative agent. Since the resources available to the agent (in terms of computational power, memory, etc. are bounded, the resulting behavior may be imperfect hence the distinction between bounded and perfect rationality [Good, 1971; Simon, 1982; Doyle, 1990; Russell and Wefald, 1991] Bounded rationality is a desired property of intelligent agents since it provides a good evaluation criteria and since it establishes a formal framework to analyze agents. This paper addresses several fundamental questions related to models ....

I. J. Good. Twenty-seven principles of ra- tionality. Foundations of Statistical Inference, V. P. Godambe and D. A. Sprott (eds.), Toronto: Holt, Rinehart and Winston, 1971.

....reasoning was Herbert Simon. In 1958, he claimed that the global optimization problem is to find the least cost or best return decision, net of computational costs [32] The statistician I.J. Good advocated a similar approach to decision making that takes deliberation costs into account [9]. # Corresponding author. E mail address: hansen cs.msstate.edu (E.A. Hansen) 0004 3702 01 see front matter 2001 Elsevier Science B.V. All rights reserved. PII: S0004 3702(00)00068 0 By the mid 1980s, AI researchers began to formalize these ideas and produce effective models for ....

I.J. Good, Twenty-seven principles of rationality, in: V.P. Godambe, D. Sprott (Eds.), Foundations of Statistical Inference, Holt, Rinehart, Winston, Toronto, 1971, pp. 108--141.

....bargaining occurs over whether or not to use a solution to the joint problem, and how to divide the associated value or cost. Early on, it was recognized that humans have bounded rationality, for example, due to cognitive limitations, so they do not act rationally as economic theory would predict [7, 35]. Since then, considerable work has focused on developing normative models that prescribe how a computationally limited agent should behave (see, for example, 5,9, 26] This is a highly nontrivial undertaking, encompassing numerous fundamental and technical difficulties. As a result most of ....

I. Good, Twenty-seven principles of rationality, in: V. Godambe, D. Sprott (Eds.), Foundations of Statistical Inference, Holt, Rinehart, Winston, Toronto, Ontario, 1971.

....1 Introduction In most real world settings, due to limited time, an agent cannot perform all potentially useful deliberation actions. As a result it will generally be unable to act rationally in the world. This phenomenon, known as bounded rationality, has been a long standing research topic [3, 17, 18]. Most of that research has been descriptive in the sense that the goal has been to characterize how agents in particular, humans deal with this constraint. Another strand of bounded rationality research has the normative (prescriptive) goal of characterizing how agents should deal with this ....

Irving Good. Twenty-seven principles of rationality. In V Godambe and D Sprott, editors, Foundations of Statistical Inference. Toronto: Holt, Rinehart, Winston, 1971.

....of the monitored process. Monitoring itself, however, also carries a cost. This brings up the interesting question of when and how monitoring should be performed to optimize the tradeoff between its costs and benefits. Monitoring decisions can be therefore viewed as a kind of type II rationality [7], and the difference between the performance with and without monitoring corresponds to the concept of intrinsic utility [22] Dean and Boddy [2,5] have worked on a more complicated setup with a sequence of anytime algorithms. They assumed that no run time monitoring is taking place and ....

I.J. Good, Twenty-seven principles of rationality, in: V.P. Godambe, D.A. Sprott (Eds.), Foundations of Statistical Inference, Holt, Rinehart, Winston, Toronto, 1971, pp. 108--141.

....for finding solutions to problems given in terms of goals, are making a tradeoff between computation and solution quality. A solution that satisfies the goals of a problem is a minimally acceptable solution. Good s type II rationality is closely related to Simon s ideas on bounded rationality [10]. Type II rationality, which is rationality that takes into account resource limits, is a concept that has its roots in mathematics and philosophy rather than psychology. Good creates a set of normative principles for rational behavior that take computational limits into account. He also considers ....

I. J. Good. Twenty-seven principles of rationality. In V. P. Godambe and D. A. Sprott, editors, Foundations of statistical inference, pages 108--141. Holt Rinehart Wilson, Toronto, 1971.

.... similar to the approach to bounded rationality in game theory, placing a static constraint on the complexity on agents [47] In comparison, economic models of metadeliberation select a level of deliberation within a decisiontheoretic framework, based on the expected value of further deliberation [22, 54, 48]. In contrast to the recent literature on bounded rational learning in games [32, 44] we assume in the portfolio selection problem that an agent s opponent (the market) plays the same strategy for all agent strategies. Prices do not depend on investment actions. Furthermore, there is no ....

I J Good. Twenty-seven principles of rationality. In V P Godambe and D A Sprott, editors, Foundations of Statistical Inference. Toronto: Holt, Rinehart and Winston, 1971.

.... by maximizing . In Figure 1, this happens at the rightmost edge of the graph. When costs are not negligible, the policy which maximizes may not maximize , as the cost 1 The subscripts and are employed here as a reference to the ideas of Good [Good, 1972] , who identified two types of rationality; the first, type , is without regard to computational costs, and the second, type , accounting for computational costs. may be too high. In general, i.e. a given policy never increases in value when costs are ....

Good, I. J. 1972. Twenty-seven principles of rationality. In Godambe, V. P., and Sprott, D., eds., Foundations of Statistical Inference. Toronto: Holt,Rinehart,Winston. 108--141.

....bargaining occurs over whether or not to use a solution to the joint problem, and how to divide the associated value or cost. Early on, it was recognized that humans have bounded rationality, for example, due to cognitive limitations, so they do not act rationally as economic theory would predict [35, 7]. Since then, considerable work has focused on developing normative models that prescribe how a computationally limited agent should behave (see, for example [9, 26, 5] This is a highly nontrivial undertaking, encompassing numerous fundamental and technical difficulties. As a result most of ....

Irving Good. Twenty-seven principles of rationality. In V Godambe and D Sprott, editors, Foundations of Statistical Inference. Toronto: Holt, Rinehart, Winston, 1971.

....or probably correct, where the algorithm will return the optimal solution with probability p. Research on these kinds of algorithms, generally referred to as bounded rationality, was initiated by Simon [Sim82] in the early 1950s. This work is described in this section. In the 1960s, Good [Goo71] distinguished a type II rationality from classical type I rationality. Type II rationality maximises expected utility taking into account deliberation costs. In this paper, the term bounded rationality is used for indicating the field of research that is concerned with the problems that ....

I. J. Good. Twenty-seven principles of rationality. In V. P. Godambe and D. A. Sprott, editors, Foundations of statistical inference, pages 108--141. Holt Rinehart Wilson, Toronto, 1971.

....maximizes EV II (ffi) by maximizing EV I (ffi) In Figure 1, this happens at the rightmost edge of the graph. When costs are not negligible, the policy which maximizes EV I (ffi) may not maximize EV II (ffi) as the cost 1 The subscripts I and II are employed here as a reference to the ideas of Good [Good, 1972] , who identified two types of rationality; the first, type I , is without regard to computational costs, and the second, type II , accounting for computational costs. may be too high. In general, EV II (ffi) EV I (ffi) i.e. a given policy never increases in value when costs are figured into ....

Good, I. J. 1972. Twenty-seven principles of rationality. In Godambe, V. P., and Sprott, D., eds., Foundations of Statistical Inference. Toronto: Holt,Rinehart,Winston. 108--141.

.... approaches which apply the principle of maximum expected utility to problems in these representations typically make the assumption of negligible computational costs [35, 21, 45] There is considerable interest in how the principle of maximum expected utility can be realized in practice [11, 41, 42, 18, 47], since the assumption of negligible computational costs is not always appropriate. In domains such as medical informatics, decision problems may be so large that computing an optimal policy is not possible in practice. In these cases, sub optimal policies might be computed off line. In the domain ....

I. J. Good. Twenty-seven principles of rationality. In V. P. Godambe and D.A. Sprott, editors, Foundations of Statistical Inference, pages 108--141. Holt,Rinehart,Winston, Toronto, 1972.

....power, memory, and the time available to make decisions. It follows that the e ective control of reasoning is a key factor in the success (or otherwise) of an agent system. Research on resource bounded decision making and the control of reasoning originated in economics and the decision sciences [16, 8]; in ai, such research falls under the banner of meta level reasoning [15] and in the agent literature, it falls under work on bounded optimality [14] In this paper, we examine the relationship between the properties of the environment in which an agent must operate, and the requirements for ....

I. J. Good. Twenty-seven principles of rationality. In V. P. Godambe and D. A. Sprott, editors, Foundations of statistical inference, pages 108-141. Holt Rinehart Wilson, Toronto, 1971.

....solve this combinatorial problem optimally without any deliberation costs such as CPU time costs or time delay costs. If the problem is hard and the instance is large, it is unrealistic to assume that it can be solved without deliberation costs. This paper adopts a model of bounded rationality [26, 10], where each agent has to pay for the computational resources (CPU cycles) that it uses for deliberation. A fixed computation cost c comp 0 per CPU time unit is assumed. 3 The domain cost associated with coalition S is denoted by c S (r S ) 0, i.e. it depends on (decreases with) the allocated ....

I. Good. Twenty-seven principles of rationality. In V. Godambe and D. Sprott, editors, Foundations of Statistical Inference. Toronto: Holt, Rinehart, Winston, 1971.

....combinatorial problem optimally without any deliberation costs such as CPU time costs or time delay costs. If the problem is hard and the instance is large, it is unrealistic to assume that it can be solved without deliberation costs. This paper adopts a model of bounded rationality [ Simon, 1982; Good, 1971 ] where each agent has to pay for the computational resources (CPU cycles) that it uses for deliberation. A fixed computation cost c comp 0 per CPU time unit is assumed. 2 The domain cost associated with coalition S is denoted by 1 In some problems, not all tasks have to be handled. This can ....

Irving Good. Twenty-seven principles of rationality. In V Godambe and D Sprott, editors, Foundations of Statistical Inference. Toronto: Holt, Rinehart, Winston, 1971.

....least worst case optimal. We also show that Min Max LRTA solves the goal directed navigation tasks fast, converges quickly, and requires only a small amount of memory. Introduction Situated agents (such as robots) have to take their planning time into account to solve planning tasks efficiently (Good 1971). For single instance planning tasks, for example, they should attempt to minimize the sum of planning and plan execution time. Finding plans that minimize the planexecution time is often intractable. Interleaving planning and plan execution is a general principle that can reduce the planning time ....

Good, I. 1971. Twenty-seven principles of rationality. In Godambe, V., and Sprott, D., eds., Foundations of Statistical Inference. Holt, Rinehart, Winston.

.... to model other agents in a game this can be somewhere between a full model (traditional game theory) and a static model (evolutionary game theory) There is a long history of the explicit use of decision theoretic models for meta reasoning (reasoning about reasoning) in artificial intelligence (Good, 1971; Simon, 1976; Boddy and Dean, 1989; Russell and Wefald, 1991) Recent models of multiagent learning provide a hierarchy of agent models and allow an analysis of the effect of strategic (non myopic) learning on the equilibrium outcomes in games (Gmytrasiewicz and Durfee, 1995; Wellman and Hu, ....

Good, I. J. (1971). "Twenty-seven principles of rationality," in Foundations of Statistical Inference (V. P. Godambe and D. A. Sprott, Eds.). Toronto: Holt, Rinehart and Winston.

....This observation has been a commonplace since the very beginning of AI. Yet systems that select suboptimal actions fall outside calculative rationality per se, and we need a better theory to understand them. 5 Metalevel Rationality Metalevel rationality, also called Type II rationality by I. J. Good [1971] , is based on the idea of finding an optimal tradeoff between computational costs and decision quality. Although Good never made his concept of Type II rationality very precise he defines it as the maximization of expected utility taking into account deliberation costs it is clear that the ....

I. J. Good. Twenty-seven principles of rationality. In V. P. Godambe and D. A. Sprott, editors, Foundations of Statistical Inference, pages 108--141. Holt, Rinehart, Winston, Toronto, 1971.

....Finally, in Section 5, we summarize our results, and 1 Elsewhere Pollack [1992] discusses this claim in detail. The idea of trading decision quality for timeliness dates back to Simon s critique of decision theory and his development of the notion of satisficing [Simon 1955,Simon 1957] Good [1983] similarly distinguishes between the idealized notion of deliberation that is central to decision theory, which he calls Type I Rationality, and Type II rationality, which takes into account the time cost of deliberation. The need to trade decision quality for timeliness is also been a concern in ....

I. J. Good. Twenty-seven principles of rationality. In Good Thinking: The Foundations of Probability and Its Applications, pages 15--19. University of Minnesota Press, Minneapolis, MN, 1983.

....the distribution information required to determine the expected case behavior, and then, what to do with such information, once it is available. We view our approach, as embodied in the palo system, as addressing exactly those questions. Moreover, our palo exhibits a type of Type II rationality [26], as it seeks an element whose expected utility is optimal, subject to the resource constraint of spending only a feasible amount of time to find such an element. Many other systems are similarly motivated by this issue of computational effectiveness i.e. what is the best performance the ....

I. J. Good. Twenty-seven principles of rationality. In V. P. Godambe and D. A. Sprott, editors, Foundations of Statistical Inference. Holt, Rinehart and Winston, Toronto, 1971.

....A suitable generalization of complexity theory has not yet been developed. A somewhat longer tradition of considering the effects of boundednesss on decisionmaking exists in economics and the decision sciences, where human characteristics must sometimes be considered. Since the 1960 s, I. J. Good [16] has emphasized the conceptual distinction between classical or type I rationality, and what he called type II rationality, or the maximization of expected utility taking into account deliberation costs. Researchers in decision analysis, especially Howard [28] have studied the problem of the ....

Good, I. J. (1971) Twenty-seven principles of rationality. In Godambe, V. P., and Sprott, D. A. (Eds.) Foundations of Statistical Inference. Toronto: Holt, Rinehart, Winston, 108-141.

....work focussed mainly on satisficing designs, which deliberate until reaching some solution satisfying a preset aspiration level. The results have descriptive value for modelling various actual entities and policies, but no prescriptive framework for bounded rationality was developed. I. J. Good [ 10 ] emphasized the conceptual distinction between classical or type I rationality, and what he called type II rationality, or the maximization of expected utility taking into account deliberation costs. 3 What this means is that an agent exhibits type II rationality if at the end of its ....

Good, I. J. (1971) Twenty-seven principles of rationality. In Godambe, V. P., and Sprott, D. A. (Eds.) Foundations of Statistical Inference. Toronto: Holt, Rinehart, Winston.

....Russell and Wefald use probability and decision theory to develop a general formula for the utility of computation. The utility of a computational action is derived from its effect on the agent s choice of action in the world. Their work is related to some early work in the decision sciences by Good (Good 1971) and Simon (Simon 1976) The problem is cast as search, with the marginal value of computation determining the optimal sequence of computations. Russell and Wefald present an application to competitive game playing, and are able to demonstrate a substantial improvement in search efficiency when ....

....of the effect that it has on actions performed by the agent, noting that both actions and computation have time value. 1.2 Views from the Decision Sciences Early discussion of bounded rationality within the decision sciences presumed that a normative approach to metareasoning would be too costly. Good (Good 1971) provided the earliest discussion of the explicit integration of the costs of inference within the framework of normative rationality. Good made a distinction between classical, or Type I rationality, and what he called Type II rationality, or the maximization of expected utility taking into ....

Good, I. J. 1971. Twenty-seven principles of rationality. In Godambe, V. P., and Sprott, D. A., eds., Foundations of Statistical Inference.

First 50 documents

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