![]() |
|||
![]()
|
![]() |
![]() |
![]() |
5.3. VALIDATION AND VERIFICATIONJust as the effectiveness of the various instruments and drugs used by physicians must be validated, the accuracy, utility, and dependability of medical expert systems must eventually be assessed. Validation and verification (V & V) issues are more important in the field of medicine than in other fields because of potential problems that may result from an incorrect system. For example, an incorrect diagnosis or prescription provided by the system could threaten the life of a patient. Concepts of validation and verification are different, but often they are not clearly distinguished. They can be defined as follows (Ayel and Laurent, 1991):
These definitions are well founded in classical software with a life cycle. The V & V process can be distinguished according to the persons who use them. The "customer," the person placing the order and/or the future user, is concerned with the validation process. The customer looks for usefulness, competencies, performances, and the reliability of the product. The validation process is directly concerned with the behavior of the expert system. Unlike the validation process, the verification process is based on the designer's knowledge through development. This knowledge is mainly concerned with technical details about the product and with its components. The validation and verification of expert systems is different from that of other types of computer systems (O'Leary, 1991). These differences include: a focus on symbolic knowledge rather than numeric; an investigation of previously unstructured problems; inclusion of both symbolic and numerical information in the same program (e.g., rules of the form "IF ... THEN" and uncertainty factors on the weights); and the general lack of a means by which to determine the quality of a solution by means other than a human verification. To date, most of the methods for V & V have been developed for business fields, and they are often not suitable for the medical field. Accordingly, a trends that we expect to see increase with regard to V & V are the clinical trial methods that have widely been used in validating new drugs. Just as the effectiveness of new drugs must be validated by clinical trials, the accuracy and dependability of expert systems must undergo extensive clinical trials and be rated in experimental evaluation performing at human-expert levels in their respective domains. 5.4. ETHICAL AND LEGAL CONSIDERATIONSAs an increasing number of computer programs are being promoted for medical use, ethical and legal problems will also result from the application of such programs in clinical settings. While the legal problems associated with expert systems that provide medical advice have fyet to be addressed by the courts, there are several important ethical and legal questions related to the use of expert systems in clinical medicine. Who should be authorized to use such systems, and in what ways? How can doctors and patients evaluate whether an expert system is safe for use with humans? The major factor in determining the liabilities of a medical expert system is the classification of the system. If classified as a product, strict product liability will be imposed. But, if classified as a service, professional misconduct or negligence will be imposed (Cook and Whittaker, 1989). Since the physicians' acts of making a diagnosis and providing therapy have traditionally been classified as services, it is likely that adverse outcomes resulting from the use of such systems will be governed by the legal principle of negligence liability. As an increasing number of such systems are expected to be promoted for mecial use in the future, there should be proper research on these legal issue to prevent adverse outcomes. The issue of responsibility is closely tied to the assignment of liability when a poor decision is made. It should be apparent that liability be assigned to the designers of an expert system a well as to its users. This should be an important consideration for potential developers of expert systems. Furthermore, potential dangers exist in allowing open access to an expert system because the user might not be sufficiently trained to operate the system properly. Unauthorized or improper use of the system can result in legal problems. Moreover, the user must provide medically reliable information as input and be able to override a system's advice if the advice is in error. Access to some systems might therefore be limited to licensed professionals of specific categories. To use a program properly, the user would need the requisite educational background. Using expert systems for decision-making raises many important ethical ussues, especially in dealing with medical decisions. Current technological advances have further complicated the relationship between physician and patient, forcing medical professionals to review even the most basic concepts on which their discipline is based (Holtzman, 1989). Moreover, using expert systems implies that knowledge engineers and domain experts will indirectly participate in the medical decision process. This ultimately means that we must carefully consider the extent to which such participation is desirable. This discussion shows that using medical expert systems requires careful attention to ethical matters. Although the potential benefits of these systems can be great, they are also potentially subject to major abuse. Placing these systems in perspective within their human context is an essential factor in our ability to use them wisely.
|
![]() |
|
Use of this site is subject certain Terms & Conditions. Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details. |