This article presents a cognitive model of the auditor's going-concern judgment, called the GCX model, that is based on the analysis of transcripts of interviews with expert auditors. It proposes a specific set of knowledge and reasoning skills to model the auditor's performance of the going-concern judgment. the GCX model has been implemented and tested in a computer program, called “GCX,” using data drawn from a realworld company, and its performance qualitatively matches that of expert auditors. © 1992 John Wiley & Sons, Inc.

Natural language interaction with robots is an important goal since it promises access to robots by non-experts. Specifically, natural language interaction with robots would allow robots to be used easily and to be taught new skills. To produce such interaction requires, however, a choice among a variety of technologies in several areas. A working natural language interface to a robotic assembly system is described that allows a user to converse with the system about both vision and manipulation and to teach the system new vision knowledge and new assembly plans. The success of this approach suggest that its technologies are an appropriate choice for future real-world natural language interfaces to robot assembly systems.

Robots are increasingly being used in situations requiring them to exhibit flexible behavior. In such situations point-to-point motion sequences are inadequate. Instead, robots will probably use planning mechanisms, coupled with sensors of various kinds, to carry out their tasks. Such mechanisms allow robots to perform in situations where object position cannot be exactly specified in advance and in which unanticipated events can occur. One alternative is to teach the robot new plans. This paper describes a robot assembly system in which the robot is taught the steps in the new plan using natural language input.

Existing techniques for programming robots are either limited or require expert programmers. This paper describes an implemented robot system, TL, which is taught using a new technique which is easy to use yet produces powerful high-level plans. In this technique the user provides an example of the plan by moving the manipulator through the desired assembly. TL observes the assembly and infers the underlying plan. This paper describes TL, presents an example of it learning to assemble a simple electric switch, and discusses future directions for development.

The ability to explain something, e.g., the operation of a complex machine or program, is an important, but poorly understood, component of intelligent behavior. We discuss an artificial intelligence approach to the modeling of the explanation process within the framework of a graphics-based CAD system currently under development, which can describe its own use, including the common ways to make and recover from errors. With a coordinated textual and pictorial display, the system, CADHELP, simulates an expert demonstrating the operation of the graphical features of the CAD subsystem. It consults a knowledge base of feature scripts to explain a feature, generate prompts as the feature is being operated, and to give certain types of help when a feature is misused. CADHELP provides these services by summarizing a feature script in different ways depending upon what it has told the user previously. The summarization process is based upon a theory of natural-language generation, in which a concept to be expressed is replaced in a short-term memory by words spanning part of its meaning interspersed with subconcepts still to be expressed. The generation process is mediated by a series of "sketchification" strategies, which prescribe which parts of a knowledge structure, a causal chain, or a single concept need not be expressed, since the listener should be able to infer them.