Cognitive Science at present is highly fragmented across multiple disciplines and even within tho... more Cognitive Science at present is highly fragmented across multiple disciplines and even within those disciplines. It not only lacks an underlying explanatory framework but even a unifying theme or concept other than the not-very-helpful ones of mind and behavior. In this paper, we survey some ways in which a robust notion of self may be "waiting" to be fleshed out across the allied cognitive sciences in what we think may become a powerfully unifying and explanatory role.
In this paper we provide an account of the structural underpinnings of self-awareness. We offer b... more In this paper we provide an account of the structural underpinnings of self-awareness. We offer both an abstract, logical account-by way of suggestions for how to build a genuinely selfreferring artificial agent-and a biological account, via a discussion of the role of somatoception in supporting and structuring self-awareness more generally. Central to the account is a discussion of the necessary motivational properties of self-representing mental tokens, in light of which we offer a novel definition of self-representation. We also discuss the role of such tokens in organizing self-specifying information, which leads to a naturalized restatement of the guarantee that introspective awareness is immune to error due to mis-identification of the subject. 1. The essential prehension John Perry once noticed a trail of sugar on the supermarket floor (Perry, 1977). Thinking to himself, "someone is making a mess," he set out to find the person responsible and stop him. Thus, pushing his shopping cart before him, he began to follow the trail in search of the messmaking shopper while sugar continued to leak from the torn bag in his own cart. This amusing and self-defeating incoherence ends only when he realizes: (1) I am the one making the mess. Perry calls the "I" in this realization the "essential indexical"-essential because no belief of the form "X is the one making a mess" will cause a shopper to check his own cart except one in which X has, or can be made to have, the indexical and self-reflexive character of "I". Thus, for instance, the belief, (2) The only bearded philosopher in the market is the one making a mess. will only cause Perry to check his cart if he also believes that he himself (he* (Castañeda, 1966))
A standard model of mind will involve not only an architecture but also a set of capabilities. Id... more A standard model of mind will involve not only an architecture but also a set of capabilities. Ideally, the two should inform one another at a deep level, as an architecture is what both enables and constrains capabilities. In that spirit, we consider in some detail a routine and (deceptively) simple robotic task. From it, we build out a substantial list of capabilities that appear essential for a general-purpose execution of the task. We argue that this type of exercise is an indispensable step toward the establishment of a baseline for the comparison of cognitive architectures, and that the resulting taxonomy can inform the synthesis of a standard model of the mind.
This essay describes a general approach to building perturbation-tolerant autonomous systems, bas... more This essay describes a general approach to building perturbation-tolerant autonomous systems, based on the conviction that artificial agents should be able to notice when something is amiss, assess the anomaly, and guide a solution into place. This basic strategy of self-guided learning is termed the metacognitive loop; it involves the system monitoring, reasoning about, and, when necessary, altering its own decision-making components. This paper (a) argues that equipping agents with a metacognitive loop can help to overcome the brittleness problem, (b) details the metacognitive loop and its relation to our ongoing work on time-sensitive commonsense reasoning, (c) describes specific, implemented systems whose perturbation tolerance was improved by adding a metacognitive loop, and (d) outlines both short-term and long-term research agendas.
Although much e ort has been expended by researchers in trying to maintain a consistent belief ba... more Although much e ort has been expended by researchers in trying to maintain a consistent belief base in formalizing commonsense reasoning, there is some evidence that the nature of commonsense reasoning itself brings inconsistencies with it. I will outline a number of sources of such inconsistencies, and discuss why they appear unavoidable. I will also suggest that, far from being a roadblock to e ective commonsense, (detected) inconsistencies are often a reasoner's best guide to what to do next.
Default reasoning is analyzed as consisting (implicitly) of at least three further aspects, that ... more Default reasoning is analyzed as consisting (implicitly) of at least three further aspects, that we call oracles, jumps, and fixes, which in turn are related to the notion of a belief. Beliefs are then discussed in terms of their use in a reasoning agent. Next an idea of David Israel is embellished to showthat certain desiderata regarding these aspects of default reasoning lead to inconsistent belief sets, and that as a consequence the handling of inconsistencies must be taken as central to commonsense reasoning. Finally, these results are applied to standard cases of default reasoning formalisms in the literature (circumscription, default logic, and non-monotonic logic), where it turns out that evenw eaker hypotheses lead to failure to achieve commonsense default conclusions. descriptors: beliefs, consistency, introspection, knowledge representation, defaults, circumscription, nonmonotonic logic, commonsense reasoning, ornithology that in pursuing the former,one is naturally led to consider idealized situations in which features irrelevant to the particular phenomenon at hand are deliberately left out of consideration. Such an approach has been customary in much research in commonsense reasoning, most conspicuously in the assumption of logical omniscience: that an agent knows (and eveni nstantly) all logical consequences of his beliefs, generally regarded as part of the notion of epistemological adequacy. That is, although no one believest hat agents actually can reason this way,i th as seemed to be a convenient test-bed for ideas about what reasoning is like,apart from the ''noise''ofthe real world. While this has come under criticism lately,and while authors of default formalisms acknowledge the importance to their very topic of the process nature of defaults, 2 still the latter has remained conspicuously absent from the continuing development of such formalisms. Here we argue that the very essence of default reasoning, and of commonsense reasoning in general, derivesf rom its being embedded in the real world, and in agents evolved to deal with such by means of an appropriately introspective viewoftheir own fallibility and corrigibility overt ime. This in turn will be seen to pose problems for logical omniscience. The ''spec''v iewo fa ni deal thinker,w hich we are critiquing, we refer to as that of an ''omnithinker''(or OT for short). To facilitate this discussion, we first present an extended illustration of default reasoning along lines found in the literature. Reasoning by default involves reaching a conclusion C on the basis of lack of information that might rule out C. 3 Fore xample, givent hat Tweety is a bird, and no more, one might, if prompted, conclude (at least tentatively) that Tweety can fly. 4 Here C is the statement that Tweety can fly. Such a conclusion may be appropriate when ''typical''e lements of a givenc ategory (in this case, birds) have the property under consideration (ability to fly). 2 At least regarding their context of an overall process of reasoning going on overtime within a changing environment of inputs. E.g., McDermott and Doyle [1980, p 41] speak of ''...modelling the beliefs of active processes which, acting in the presence of incomplete information, must makeand subsequently revise assumptions in light of newobservations.'' Reiter [1980, p 86] mentions ''...the need for some kind of mechanism for re vising beliefs [his emphasis] in the presence of newinformation. What this amounts to is the need to record, with each derivedbelief, the default assumptions made in deriving that belief. Should subsequent observations invalidate the default assumptions supporting some belief, then it must be removedfrom the data base.'' 3 Doyle [1983,1985] has presented interesting views on this phenomenon, relating it to group decision-making. 4 Tentativity is the obvious key,w hich all default formalisms are designed to capture, as opposed to other more robust kinds of inferences. And it is this that our analysis will focus on most.
In recent years, there has been strong interest in both reasoning about goal-identification and s... more In recent years, there has been strong interest in both reasoning about goal-identification and selection and metacognitive handling of anomalous situations. These two concerns are usually framed in terms of making agents more autonomous and flexible in dynamic and complex domains. Here, we wish to argue that there is a natural unifying perspective that includes both concerns and that may point the way to a yet more powerful kind of autonomy.
A fully autonomous agent recognizes new problems, explains what causes such problems, and generat... more A fully autonomous agent recognizes new problems, explains what causes such problems, and generates its own goals to solve these problems. Our approach to this goal-driven model of autonomy uses a methodology called the Note-Assess-Guide procedure. It instantiates a monitoring process in which an agent notes an anomaly in the world, assesses the nature and cause of that anomaly, and guides appropriate modifications to behavior. This report describes a novel approach to the note phase of that procedure. A-distance, a sliding-window statistical distance metric, is applied to numerical vector representations of intermediate states from plans generated for two symbolic domains. Using these representations, the metric is able to detect anomalous world states caused by restricting the actions available to the planner. 1.
In recent years, there has been strong interest in both reasoning about goal-identification and s... more In recent years, there has been strong interest in both reasoning about goal-identification and selection and metacognitive handling of anomalous situations. These two concerns are usually framed in terms of making agents more autonomous and flexible in dynamic and complex domains. Here, we wish to argue that there is a natural unifying perspective that includes both concerns and that may point the way to a yet more powerful kind of autonomy.
A fully autonomous agent recognizes new problems, explains what causes such problems, and generat... more A fully autonomous agent recognizes new problems, explains what causes such problems, and generates its own goals to solve these problems. Our approach to this goal-driven model of autonomy uses a methodology called the Note-Assess-Guide procedure. It instantiates a monitoring process in which an agent notes an anomaly in the world, assesses the nature and cause of that anomaly, and guides appropriate modifications to behavior. This report describes a novel approach to the note phase of that procedure. A-distance, a sliding-window statistical distance metric, is applied to numerical vector representations of intermediate states from plans generated for two symbolic domains. Using these representations, the metric is able to detect anomalous world states caused by restricting the actions available to the planner. 1.
A symbol is a pattern (of physical marks, electromagnetic energy,etc.) which denotes, designates,... more A symbol is a pattern (of physical marks, electromagnetic energy,etc.) which denotes, designates, or otherwise has meaning. The notion that intelligence requires the use and manipulation of symbols, and that humans are therefore symbol systems, has been extremely influential in artificial intelligence.
A symbol is a pattern (of physical marks, electromagnetic energy,etc.) which denotes, designates,... more A symbol is a pattern (of physical marks, electromagnetic energy,etc.) which denotes, designates, or otherwise has meaning. The notion that intelligence requires the use and manipulation of symbols, and that humans are therefore symbol systems, has been extremely influential in artificial intelligence.
Beliefs play complex and sometimes confusing roles in AI. This paper survey (i) a variety of noti... more Beliefs play complex and sometimes confusing roles in AI. This paper survey (i) a variety of notions of belief, (ii) formal efforts to characterize beliefs, and (iii) how beliefs are related to action, to language, and to commonsense. In addition, we will consider certain logical tensions between beliefs and consistency.
The paper discusses the benefits of metacognitive guidance for a natural language dialog agent. T... more The paper discusses the benefits of metacognitive guidance for a natural language dialog agent. These capabilities may be included directly in the agent or through a general purpose external module. We report on the specific case of handling pause time in dialog, using a metacognitive loop within the agent, and discuss future experiments implementing guidance for this example also using the general module.
Cognitive Science at present is highly fragmented across multiple disciplines and even within tho... more Cognitive Science at present is highly fragmented across multiple disciplines and even within those disciplines. It not only lacks an underlying explanatory framework but even a unifying theme or concept other than the not-very-helpful ones of mind and behavior. In this paper, we survey some ways in which a robust notion of self may be "waiting" to be fleshed out across the allied cognitive sciences in what we think may become a powerfully unifying and explanatory role.
In this paper we provide an account of the structural underpinnings of self-awareness. We offer b... more In this paper we provide an account of the structural underpinnings of self-awareness. We offer both an abstract, logical account-by way of suggestions for how to build a genuinely selfreferring artificial agent-and a biological account, via a discussion of the role of somatoception in supporting and structuring self-awareness more generally. Central to the account is a discussion of the necessary motivational properties of self-representing mental tokens, in light of which we offer a novel definition of self-representation. We also discuss the role of such tokens in organizing self-specifying information, which leads to a naturalized restatement of the guarantee that introspective awareness is immune to error due to mis-identification of the subject. 1. The essential prehension John Perry once noticed a trail of sugar on the supermarket floor (Perry, 1977). Thinking to himself, "someone is making a mess," he set out to find the person responsible and stop him. Thus, pushing his shopping cart before him, he began to follow the trail in search of the messmaking shopper while sugar continued to leak from the torn bag in his own cart. This amusing and self-defeating incoherence ends only when he realizes: (1) I am the one making the mess. Perry calls the "I" in this realization the "essential indexical"-essential because no belief of the form "X is the one making a mess" will cause a shopper to check his own cart except one in which X has, or can be made to have, the indexical and self-reflexive character of "I". Thus, for instance, the belief, (2) The only bearded philosopher in the market is the one making a mess. will only cause Perry to check his cart if he also believes that he himself (he* (Castañeda, 1966))
A standard model of mind will involve not only an architecture but also a set of capabilities. Id... more A standard model of mind will involve not only an architecture but also a set of capabilities. Ideally, the two should inform one another at a deep level, as an architecture is what both enables and constrains capabilities. In that spirit, we consider in some detail a routine and (deceptively) simple robotic task. From it, we build out a substantial list of capabilities that appear essential for a general-purpose execution of the task. We argue that this type of exercise is an indispensable step toward the establishment of a baseline for the comparison of cognitive architectures, and that the resulting taxonomy can inform the synthesis of a standard model of the mind.
This essay describes a general approach to building perturbation-tolerant autonomous systems, bas... more This essay describes a general approach to building perturbation-tolerant autonomous systems, based on the conviction that artificial agents should be able to notice when something is amiss, assess the anomaly, and guide a solution into place. This basic strategy of self-guided learning is termed the metacognitive loop; it involves the system monitoring, reasoning about, and, when necessary, altering its own decision-making components. This paper (a) argues that equipping agents with a metacognitive loop can help to overcome the brittleness problem, (b) details the metacognitive loop and its relation to our ongoing work on time-sensitive commonsense reasoning, (c) describes specific, implemented systems whose perturbation tolerance was improved by adding a metacognitive loop, and (d) outlines both short-term and long-term research agendas.
Although much e ort has been expended by researchers in trying to maintain a consistent belief ba... more Although much e ort has been expended by researchers in trying to maintain a consistent belief base in formalizing commonsense reasoning, there is some evidence that the nature of commonsense reasoning itself brings inconsistencies with it. I will outline a number of sources of such inconsistencies, and discuss why they appear unavoidable. I will also suggest that, far from being a roadblock to e ective commonsense, (detected) inconsistencies are often a reasoner's best guide to what to do next.
Default reasoning is analyzed as consisting (implicitly) of at least three further aspects, that ... more Default reasoning is analyzed as consisting (implicitly) of at least three further aspects, that we call oracles, jumps, and fixes, which in turn are related to the notion of a belief. Beliefs are then discussed in terms of their use in a reasoning agent. Next an idea of David Israel is embellished to showthat certain desiderata regarding these aspects of default reasoning lead to inconsistent belief sets, and that as a consequence the handling of inconsistencies must be taken as central to commonsense reasoning. Finally, these results are applied to standard cases of default reasoning formalisms in the literature (circumscription, default logic, and non-monotonic logic), where it turns out that evenw eaker hypotheses lead to failure to achieve commonsense default conclusions. descriptors: beliefs, consistency, introspection, knowledge representation, defaults, circumscription, nonmonotonic logic, commonsense reasoning, ornithology that in pursuing the former,one is naturally led to consider idealized situations in which features irrelevant to the particular phenomenon at hand are deliberately left out of consideration. Such an approach has been customary in much research in commonsense reasoning, most conspicuously in the assumption of logical omniscience: that an agent knows (and eveni nstantly) all logical consequences of his beliefs, generally regarded as part of the notion of epistemological adequacy. That is, although no one believest hat agents actually can reason this way,i th as seemed to be a convenient test-bed for ideas about what reasoning is like,apart from the ''noise''ofthe real world. While this has come under criticism lately,and while authors of default formalisms acknowledge the importance to their very topic of the process nature of defaults, 2 still the latter has remained conspicuously absent from the continuing development of such formalisms. Here we argue that the very essence of default reasoning, and of commonsense reasoning in general, derivesf rom its being embedded in the real world, and in agents evolved to deal with such by means of an appropriately introspective viewoftheir own fallibility and corrigibility overt ime. This in turn will be seen to pose problems for logical omniscience. The ''spec''v iewo fa ni deal thinker,w hich we are critiquing, we refer to as that of an ''omnithinker''(or OT for short). To facilitate this discussion, we first present an extended illustration of default reasoning along lines found in the literature. Reasoning by default involves reaching a conclusion C on the basis of lack of information that might rule out C. 3 Fore xample, givent hat Tweety is a bird, and no more, one might, if prompted, conclude (at least tentatively) that Tweety can fly. 4 Here C is the statement that Tweety can fly. Such a conclusion may be appropriate when ''typical''e lements of a givenc ategory (in this case, birds) have the property under consideration (ability to fly). 2 At least regarding their context of an overall process of reasoning going on overtime within a changing environment of inputs. E.g., McDermott and Doyle [1980, p 41] speak of ''...modelling the beliefs of active processes which, acting in the presence of incomplete information, must makeand subsequently revise assumptions in light of newobservations.'' Reiter [1980, p 86] mentions ''...the need for some kind of mechanism for re vising beliefs [his emphasis] in the presence of newinformation. What this amounts to is the need to record, with each derivedbelief, the default assumptions made in deriving that belief. Should subsequent observations invalidate the default assumptions supporting some belief, then it must be removedfrom the data base.'' 3 Doyle [1983,1985] has presented interesting views on this phenomenon, relating it to group decision-making. 4 Tentativity is the obvious key,w hich all default formalisms are designed to capture, as opposed to other more robust kinds of inferences. And it is this that our analysis will focus on most.
In recent years, there has been strong interest in both reasoning about goal-identification and s... more In recent years, there has been strong interest in both reasoning about goal-identification and selection and metacognitive handling of anomalous situations. These two concerns are usually framed in terms of making agents more autonomous and flexible in dynamic and complex domains. Here, we wish to argue that there is a natural unifying perspective that includes both concerns and that may point the way to a yet more powerful kind of autonomy.
A fully autonomous agent recognizes new problems, explains what causes such problems, and generat... more A fully autonomous agent recognizes new problems, explains what causes such problems, and generates its own goals to solve these problems. Our approach to this goal-driven model of autonomy uses a methodology called the Note-Assess-Guide procedure. It instantiates a monitoring process in which an agent notes an anomaly in the world, assesses the nature and cause of that anomaly, and guides appropriate modifications to behavior. This report describes a novel approach to the note phase of that procedure. A-distance, a sliding-window statistical distance metric, is applied to numerical vector representations of intermediate states from plans generated for two symbolic domains. Using these representations, the metric is able to detect anomalous world states caused by restricting the actions available to the planner. 1.
In recent years, there has been strong interest in both reasoning about goal-identification and s... more In recent years, there has been strong interest in both reasoning about goal-identification and selection and metacognitive handling of anomalous situations. These two concerns are usually framed in terms of making agents more autonomous and flexible in dynamic and complex domains. Here, we wish to argue that there is a natural unifying perspective that includes both concerns and that may point the way to a yet more powerful kind of autonomy.
A fully autonomous agent recognizes new problems, explains what causes such problems, and generat... more A fully autonomous agent recognizes new problems, explains what causes such problems, and generates its own goals to solve these problems. Our approach to this goal-driven model of autonomy uses a methodology called the Note-Assess-Guide procedure. It instantiates a monitoring process in which an agent notes an anomaly in the world, assesses the nature and cause of that anomaly, and guides appropriate modifications to behavior. This report describes a novel approach to the note phase of that procedure. A-distance, a sliding-window statistical distance metric, is applied to numerical vector representations of intermediate states from plans generated for two symbolic domains. Using these representations, the metric is able to detect anomalous world states caused by restricting the actions available to the planner. 1.
A symbol is a pattern (of physical marks, electromagnetic energy,etc.) which denotes, designates,... more A symbol is a pattern (of physical marks, electromagnetic energy,etc.) which denotes, designates, or otherwise has meaning. The notion that intelligence requires the use and manipulation of symbols, and that humans are therefore symbol systems, has been extremely influential in artificial intelligence.
A symbol is a pattern (of physical marks, electromagnetic energy,etc.) which denotes, designates,... more A symbol is a pattern (of physical marks, electromagnetic energy,etc.) which denotes, designates, or otherwise has meaning. The notion that intelligence requires the use and manipulation of symbols, and that humans are therefore symbol systems, has been extremely influential in artificial intelligence.
Beliefs play complex and sometimes confusing roles in AI. This paper survey (i) a variety of noti... more Beliefs play complex and sometimes confusing roles in AI. This paper survey (i) a variety of notions of belief, (ii) formal efforts to characterize beliefs, and (iii) how beliefs are related to action, to language, and to commonsense. In addition, we will consider certain logical tensions between beliefs and consistency.
The paper discusses the benefits of metacognitive guidance for a natural language dialog agent. T... more The paper discusses the benefits of metacognitive guidance for a natural language dialog agent. These capabilities may be included directly in the agent or through a general purpose external module. We report on the specific case of handling pause time in dialog, using a metacognitive loop within the agent, and discuss future experiments implementing guidance for this example also using the general module.
A standard model of mind will involve not only an architecture but also a set of capabilities. Id... more A standard model of mind will involve not only an architecture but also a set of capabilities. Ideally, the two should inform one another at a deep level, as an architecture is what both enables and constrains capabilities. In that spirit, we consider in some detail a routine and (deceptively) simple robotic task. From it, we build out a substantial list of capabilities that appear essential for a general-purpose execution of the task. We argue that this type of exercise is an indispensable step toward the establishment of a baseline for the comparison of cognitive architectures, and that the resulting taxonomy can inform the synthesis of a standard model of the mind.
Uploads
Papers by Don Perlis