Many intellectual disciplines contribute to the field of cognitive science: psychology, linguistics, anthropology, and artificial intelligence, to name just a few. Cognitive science itself

The origins of the Soar architecture can be traced back to the seminal research of Allen Newell and Herbert Simon on symbol systems, heuristic search, goals, problem spaces, and production systems. Since its official inception in 1982, Soar has evolved through six major releases, as both an AI architecture and as the basis for a unified theory of cognition. This paper traces this evolutionary path, starting with Soar's intellectual roots, and then proceeding through the stages defined by the six major system releases. Each stage is characterized with respect to a hierarchy of four levels of analysis: the knowledge level, the problem space level, the symbolic architecture level, and the implementation level.

In contrast to current intelligent systems, which must be laboriously programmed for each task they are meant to perform, instructable agents can be taught new tasks and associated knowledge. This thesis presents a general theory of learning from tutorial instruction and its use to produce an instructable agent. Tutorial instruction is a particularly powerful form of instruction, because it allows the instructor to communicate whatever kind of knowledge a student needs at whatever point it is needed. To exploit this broad flexibility, however, a tutorable agent must support a full range of interaction with its instructor to learn a full range of knowledge. Thus, unlike most machine learning tasks, which target deep learning of a single kind of knowledge from a single kind of input, tutorability requires a breadth of learning from a broad range of instructional interactions. The theory of learning from tutorial...

Abstract not found

In complex, dynamic environments, an agent's domain knowledge will rarely be complete and correct. Existing deliberate approaches to domain theory correction are significantly restricted in the environments where they can be used. These systems are typically not used in agent-based tasks and rely on declarative representations to support non-incremental learning. This research investigates the use of procedural knowledge to support deliberate incremental error correction in complex environments. We describe a series of domain properties that constrain the error correction process and that are violated by existing approaches. We then present a procedural representation for domain knowledge which is sufficiently expressive, yet tractable. We develop a general framework for error detection and correction and then describe an error correction system, IMPROV, that uses our procedural representation to meet the constraints imposed by complex environments. Finally, we test the syst...

This paper attempts to rectify this by offering an initial comparison of two of the most well-known cognitive architectures: Act-R (Anderson, 1993; Lebiere, 1996) and Soar (Laird, Rosenbloom & Newell, 1986; Laird, Newell & Rosenbloom, 1987; Newell, 1990), two of the most wellknown cognitive architectures.

interaction, artificial intelligence. People make use of hidden external information, first recalling that it exists and then finding it. This dissertation investigates the memory phenomena involved in recalling that external information exists. We present data in which a programmer navigates to hidden features in a real-world task environment. We then present a model that accounts for this navigation by encoding and using simple episodic memories for having seen a feature. The model inherits constraints from its underlying cognitive architecture, which specify that learning is passive and pervasive, and that it creates simple memories that depend on the feature itself being present as a cue. The nature of these memories requires the model to recall features to its mind’s eye as cues in order to retrieve them. This retrieval process requires domain knowledge: familiarity with features in order to imagine them, and an idea of when it would be useful to recall having seen them. Recalling that a hidden feature exists prompts the model to scroll to that feature. Thus the model’s access to external information is a function of passively-encoded episodic memories, and retrieval of these memories using knowledge. As a claim applied to people, this appears to overlap with a recently-

Introduction Autonomous learning agents acquire knowledge from a variety of sources over their lifetimes. Some knowledge is given to the agent "at birth"; the rest is learned, through methods like autonomous exploration /observation of the environment, interactive dialogues with an instructor, etc. Unfortunately, there is no guarantee that any of this knowledge is correct. Thus, a learning agent needs the ability to recover from incorrect knowledge when that knowledge leads to performance failures. Recovering from incorrect knowledge that has caused a performance failure is difficult for two reasons. First, the agent must discover how to remedy the performance failure -- how to get out of a bad situation and back on the path towards its goals. Second, the agent must determine what knowledge caused the failure (a difficult credit assignment problem) and how to correct that knowledge. Thus, techniques for recovery from incorrect knowledge generally rely on weak inductive methods

Symbolic Performance & Learning In Continuous-valued Environments by Seth Olds Rogers Co-Chairs: John Laird and Paul Nielsen Real-world and simulated real-world domains, such as flying and driving, commonly have the characteristics of continuous-valued (CV) environments. These environments are frequently complex and difficult to control, requiring a great deal of specific, detailed knowledge. Although past approaches to learning control policies employed various forms of numerical processing, symbolic agents can also perform and learn in CV environments. There are both functional and theoretical motivations for choosing symbolic processing. SPLICE (Symbolic Performance & Learning In Continuous-Valued Environments) is a symbolic agent for adaptive control implemented in the Soar architecture. SPLICE uses a threelevel framework to first classify its sensory information into symbolic regions, then map the set of regions to a local model, then use the local model to determine an action tha...

this paper had a very simple linear domain law, the agent can easily handle more complex, possibly nonlinear or discontinuous, domain laws. These types of domains are very difficult for systems which explicitly represent domain laws like Bacon [3]. I intend to improve the agent to handle progressively more difficult domains. I specifically hope to investigate multiple state parameters, multiple simultaneous goals, multiple effectors, and real-time performance. My ultimate goal is to allow the agent to learn to control a simulated airplane from its own experience, given a simple qualitative model of flight physics. Although many active learning systems are not explicitly goal-directed, all certainly maintain implicit goals which drive their experimentation. This implies that all active learning systems dynamically select goals, and an investigation of explicit goal selection for an agent is relevant to the active learning community. From this symposium, I hope to clarify the relationship of active goal selection to active learning, and get some ideas on how to add more traditional active learning into the current agent. References