Harry E. Pople

INTERNIST-I, An Experimental Computer-Based Diagnostic Consultant for General Internal Medicine

INTERNIST-I is an experimental computer program capable of making multiple and complex diagnoses in internal medicine. It differs from most other programs for computer-assisted diagnosis in the generality of its approach and the size and diversity of its knowledge base. To document the strengths and weaknesses of the program we performed a systematic evaluation of the capabilities of INTERNIST-I. Its performance on a series of 19 clinicopathological exercises (Case Records of the Massachusetts General Hospital) published in the Journal appeared qualitatively similar to that of the hospital clinicians but inferior to that of the case discussants. The evaluation demonstrated that the present form of the program is not sufficiently reliable for clinical applications. Specific deficiencies that must be overcome include the program’s inability to reason anatomically or temporally, its inability to construct differential diagnoses spanning multiple problem areas, its occasional attribution of findings to improper causes, and its inability to explain its “thinking.” (N Engl J Med. 1982; 307:468–76.

Modeling human intention formation for human reliability assessment

Abstract This paper describes a dynamic simulation capability for modelling how people form intentions to act in nuclear power plant emergency situations. This modeling tool, Cognitive Environment Simulation or CES, was developed based on techniques from artificial intelligence. It simulates the cognitive processes that determine situation assessment and intention formation. It can be used to investigate analytically what situations and factors lead to intention failures, what actions follow from intention failures (e.g. errors of omission, errors of commission, common mode errors), the ability to recover from errors or additional machine failures, and the effects of changes in the NPP person-machine system. One application of the CES modeling environment is to enhance the measurement of the human contribution to risk in probabilistic risk assessment (PRA) studies.

Applications of artificial intelligence

One of the aims of artificial intelligence has been to make machines behave intelligently as humans do. Thus the study of artificial intelligence includes the study of how humans acquire and apply knowledge, reason under uncertainty and in complex environments, and how they do planning and solve problems. Knowledge-based automation deals with the application of artificial intelligence to a production environment in order to reduce the involvement of human beings to a minimum.

An information processing approach to theory formation in biomedical research

The extensive literature on modeling of biological systems published in the past decade reflects the growing expectation that theories of biological functions can be subject to more exacting tests of consistency with the natural system, and yield more powerful predictions if embodied in the formal structure of computer programs.

Enhancing the training of cognitive skills for improved human reliability: lessons learned from the cognitive environment simulation project

The authors have been involved in a project to model the cognitive activities that underlie operator intention formation during complex nuclear power plant emergencies. The project has involved observation of crew performance in simulated emergencies and development of a computer simulation of operator cognitive performance on the same events. The results of the research point to a number of cognitive skills that are important in handling complex multifault events of the sort that often arise in actual incidents. These cognitive skills are summarized, and an approach to developing training scenarios to enhance these skills and improve human reliability is suggested.<<ETX>>

EAGOL: an artificial intelligence system for process monitoring, situation assessment and response planning

EAGOL is an artificial intelligence system for process monitoring, situation assessment, and response planning in the management of complex, engineered systems in real time. Understanding the behavior of complex systems requires two basic types of analysis, both of which are incorporated within the EAGOL model: (1) first-principles cause-and-effect analysis of the engineered system, and (2) analysis of the types of interventions that may introduced into the engineered system from (a) built-in automatic safeguard mechanisms, and (b) human operators, who are often guided by pre-defined written procedures. EAGOL includes a goal-based model of procedure generation which allows the program (1) to generate procedures based on its assessment of real or potential system states and events, and (2) to use its internal representation of procedures and goals to reason along with human operators in pursuit of an emergency resolution.<<ETX>>

Modeling Operator Performance in Emergencies

Human performance frequently has been shown to be a substantial contributor to the overall reliability of complex dynamic systems. As a consequence, there has been a great emphasis in human factors to develop models of human cognition relevant to these situations as a tool to better predict human performance and to help in the design of improved person-machine systems. This paper describes the Cognitive Environment Simulation strategy towards human performance modeling and summarizes one comparison between system behavior and operator behavior in nuclear power emergency operations.