Rule-Based Expert Systems and Linear Models: An Empirical Comparison of Learning-By-Examples Methods*

Building models of expert decision-making behavior from examples of experts’ decisions continues to receive considerable research attention. In the 1960's and 70's, linear models derived by statistical methods were studied extensively. More recently, rule-based expert systems derived by induction algorithms have been the focus of attention. Few studies compare the two approaches. This paper reports on a study that compared linear models derived by logistic regression with rule-based systems produced by two induction algorithms—ID3 and the genetic algorithm. The techniques performed comparably in modeling the experts at one task, graduate admissions, but differed significantly at a second task, bidder selection.     

Decision Models for Robot Selection: A Comparison of Ordinary Least Squares and Linear Goal Programming Methods

The profusion of robot designs, the cost of testing, and the fact that robot operational parameter maximums are often mutually exclusive are factors that create a complex selection decision for the potential user. While formal robot testing standards are now in place, formal techniques to select robots for the testing process have not been addressed. A linear goal programming model is an effective tool for the decision maker for optimizing the robot selection process in terms of requirement priorities. It is also shown that this model provides a more stable result than the ordinary least squares estimator in the presence of statistical outliers of robot parameters. The methodology is illustrated through the use of current robot specifications.     

The Effects of Memory Structure on Using Rule-Based Expert Systems for Training: A Framework and an Empirical Test*

One justification for eliciting and representing the judgment strategy of an expert in a rule-based expert system (RBES) is to facilitate knowledge transfer to individuals with less expertise. However, prior research suggests complexities and potential problems when using RBESs for training. In this paper, a conceptual framework of user learning from RBESs is presented. It is proposed that learning may be ineffective when the problem representation of the RBES is inconsistent with the user's mental representation of the task-domain knowledge. An experiment was conducted to examine the effects of consistency (inconsistency) between the problem-solving strategy of RBESs and individuals' memory structures. Groups of subjects whose memory structure either matched or did not match two RBESs' problem-solving strategies were examined using an internal control evaluation task. The results suggest that learning was facilitated only for groups with congruence between the RBES's problem-solving strategy and a subject's memory structure.     

Combining Neural Networks and Statistical Predictions to Solve the Classification Problem in Discriminant Analysis*

A number of recent studies have compared the performance of neural networks (NNs) to a variety of statistical techniques for the classification problem in discriminant analysis. The empirical results of these comparative studies indicate that while NNs often outperform the more traditional statistical approaches to classification, this is not always the case. Thus, decision makers interested in solving classification problems are left in a quandary as to what tool to use on a particular data set. We present a new approach to solving classification problems by combining the predictions of a well-known statistical tool with those of an NN to create composite predictions that are more accurate than either of the individual techniques used in isolation.     

Some Empirical Findings on Short-Term Forecasting: Technique Complexity and Combinations

The purpose of this research is to determine if prior findings that favor simple forecasting techniques and technique combinations hold true in a short-term forecasting environment, where demand data can be quite volatile. Twenty-two time series of daily data from a real business setting are used to test one-period ahead forecasts, the epitome of short-term forecasting. The time series vary systematically as to data volatility and forecast difficulty. Forecast accuracy is measured in terms of both mean absolute percentage error (MAPE) and mean percentage error (MPE).     

The Effects of Group Attitudes Toward Alternative GDSS Designs on the Decision-making Performance of Computer-Supported Groups*

Empirical investigations of the effects of group decision support systems (GDSS) accord relatively little attention to the role of attitudes developed by groups toward the GDSS. This study draws upon the theory of social influence to examine the role played by group attitudes in mediating the impact of GDSS designs on group decision-making performance. We found that, in addition to capabilities of GDSS designs, group attitudes toward GDSS were influential in affecting group decision-making performance. The managerial implications of the study are: (1) effective GDSS designs for supporting groups in equivocal decision-making contexts should include structures for communication and consensus support and (2) attempts should be made to enhance user attitudes toward the GDSS through design features of the GDSS, training, and facilitation of positive social influences within the group. The study also provides encouraging evidence about the value of social theories, such as the social influence model and the adaptive structuration theory, in investigating GDSS effects. Further, this study illustrates the value of the partial least squares (PLS) analysis method for testing holistic structural models of GDSS effects.     

Decisional Conflict and User Acceptance of Multicriteria Decision-Making Aids*

Despite the development of increasingly sophisticated and refined multicriteria decision-making (MCDM) methods, an examination of the experimental evidence indicates that users most often prefer relatively unsophisticated methods. In this paper, we synthesize theories and empirical findings from the psychology of judgment and choice to provide a new theoretical explanation for such user preferences. Our argument centers on the assertion that the MCDM method preferred by decision makers is a function of the degree to which the method tends to introduce decisional conflict. The model we develop relates response mode, decision strategy, and the salience of decisional conflict to user preferences among decision aids. We then show that the model is consistent with empirical results in MCDM studies. Next, the role of decisional conflict in problem formulation aids is briefly discussed. Finally, we outline future research needed to thoroughly test the theoretical mechanisms we have proposed.     

The Application of Neural Networks and a Qualitative Response Model to the Auditor's Going Concern Uncertainty Decision*

An auditor gives a going concern uncertainty opinion when the client company is at risk of failure or exhibits other signs of distress that threaten its ability to continue as a going concern. The decision to issue a going concern opinion is an unstructured task that requires the use of the auditor's judgment. In cases where judgment is required, the auditor may benefit from the use of statistical analysis or other forms of decision models to support the final decision. This study uses the generalized reduced gradient (GRG2) optimizer for neural network learning, a backpropagation neural network, and a logit model to predict which firms would receive audit reports reflecting a going concern uncertainty modification. The GRG2 optimizer has previously been used as a more efficient optimizer for solving business problems. The neural network model formulated using GRG2 has the highest prediction accuracy of 95 percent. It performs best when tested with a small number of variables on a group of data sets, each containing 70 observations. While the logit procedure fails to converge when using our eight variable model, the GRG2 based neural network analysis provides consistent results using either eight or four variable models. The GRG2 based neural network is proposed as a robust alternative model for auditors to support their assessment of going concern uncertainty affecting the client company.     

An Empirical Investigation of the Effects of Decision Aids on Problem-Solving Processes*

Despite the growing number of investigations examining decision support systems (DSS), relatively few empirical studies have evaluated the effects of DSS on problem-solving processes. This study uses protocol analysis to investigate the impact of a specific decision aid on problem-solving processes in a semi-structured problem. Results indicate that decision aids influence the problem-solving processes of decision makers. The effect of a decision aid is found to be contingent on familiarity with the decision aid, task familiarity, and the interaction of these two factors. Suggestions for incorporating interaction effects and problem-solving processes into future examinations are proposed.     

Estimation of Nonlinear Models in the Presence of Measurement Error*

Techniques used in decision sciences and business research to estimate interactions between latent variables are limited in controlling for measurement error. This article uses a latent structure modeling approach that substantially controls for measurement error in nonlinear relationships. The results of this technique are compared to the results obtained applying hierarchical regression analysis and the impact of measurement error is assessed. The paper provides a unique assessment of the validity of the multi-attribute attitude model. The validity of the multiplicative rule in the model is supported.     

A Note on Input Consistency in the Application of AHP

This is a response to the points raised by Wu [3] concerning my paper on the quantitative evaluation of data bases [4].     

A Multiperiod Audit Staff Planning Model Using Multiple Objectives: Development and Evaluation*

The audit staff planning problem, a specific type of manpower planning problem, has been modeled using goal programming and, more recently, multiple objective linear programming. Prior studies developed single-period models and did not go beyond the model building stage. This study develops a multiperiod audit staff planning model and evaluates the model using a test application involving actual decision makers (partners in public accounting firms). The multiperiod model includes seven objectives to be optimized: profit (to be maximized), late completion of work, work declined, staff augmentation, staff reduction, underutilization of the work force, and shortfall in meeting professional development targets (all to be minimized). Over a four-quarter planning horizon with one “busy season,” the model is subject to constraints with respect to the projected audit work load, ability to substitute personnel and to perform interim audit work, available staff hours (including overtime limitations), supervisory requirements, and professional development targets. Results of the test application showed that the model was capable of producing a range of values for each objective. The participants were exposed to much of that range when making their decisions. The results also showed that all objectives were important and that participants were consistent in choosing their preferred level of each objective over several runs of the model. These results and the reactions of the participants demonstrate that the model is usable by actual decision makers and has potential for a number of specific applications.     

An Empirical Investigation of Information Search Strategies with Implications for Decision Support System Design

Understanding how people search through and combine information before making decisions is an important concern in the study of decision making and in the design of decision support systems (DSS). The purpose of this study is to examine DSS search strategies in relation to the body of empirical research on information load. Prior research is examined in relation to basic empirically-testable hypotheses and compared to the results of this study as a way of validating the procedures used here. The transfer of information load empirical research to a DSS scenario is confirmed statistically. Then, results of an analysis of information search patterns under DSS conditions are described. Finally, implications for DSS design are discussed in terms of possible search support mechanisms for specific search strategies.     

PERCEIVED USEFULNESS OF INFORMATION: A PSYCHOMETRIC EXAMINATION*

The perceived usefulness of information is an important construct for the design of management information systems. Yet an examination of existing measures of perceived usefulness shows that the instruments developed have not been validated nor has their reliability been verified. In this paper a new instrument for measuring two dimensions of perceived usefulness is developed. The results of an empirical study designed to test the reliability and construct validity of this instrument in a capital-budgeting setting are presented.     

The Emergence of Artificial Intelligence as a Reference Discipline for Decision Support Systems Research

It is timely and appropriate to examine both philosophical and pragmatical issues associated with formalizing the adoption of artificial intelligence as a reference discipline for decision support systems research. This paper reflects on where we were when the first special issue of Decision Sciences on expert systems and decision support systems was published, addresses the dynamics of what has taken place subsequent to the publication of that first special issue, sets forth a proposition to stimulate ongoing dialog with respect to synergies between the decision support system research agenda and the research agenda of the artificial intelligence discipline, and demonstrates how the papers appearing in this follow-up special issue of Decision Sciences are representative of an emerging, challenging, and exciting new decision support systems era.     

Production System Development for Expert Systems Using a Recursive Partitioning Induction Approach: An Application to Mortgage,Commercial, and Consumer Lending

The concepts of expert systems and decision support systems have received considerable attention recently. While systems have been proposed for various problem areas in business, difficulties still exist in the knowledge acquisition phase of development. This paper presents a recursive partitioning analysis (RPA) approach to knowledge acquisition. The RPA production system approach was applied to data sets representing the mortgage, commercial, and consumer lending problems. Comparison of the classification rates across these problems to the results of a generalized inductive inference production system (Quinlan's ID3 algorithm) and across the mortgage and commercial lending problems to traditional statistical modeling approaches indicated that the RPA approach provided superior results while using fewer variables.     

Mexican and Swedish Managers' Perceptions of the Impact of EIS on Organizational Intelligence,Decision Making,and Structure*

Although information technologies in business organizations around the world may be very similar, the meanings conveyed through the technologies may be dependent on managerial values and national culture. Cultural differences need to be understood before information technology developed for organizations in one country can be effectively implemented in organizations in another country. Drawing on survey responses from managers using Executive Information Systems (EIS) across many organizations in Mexico, Sweden, and the United States, the current study examined whether cultural differences influence perceptions of the relationship between Executive Information Systems' use and various outcomes related to decision-making behaviors and processes. The study found significant differences, predicted by cultural factors, in the impact of EIS use on senior management decision making. The findings confirm the notion that IT is used by executives to reinforce the decision-making behaviors valued in their culture.     

Subordinate Participation in Audit Budgeting Decisions: A Comparison of Decisions Influenced by Organizational Factors to Decisions Conforming with the Vroom-Jago Model*

This study examines the role of subordinate participation in the effectiveness of audit hours budget decisions in accounting firms. The study compares decisions influenced by organizational factors (such as organizational policies or the preferences of the superior) to decisions that conform to the Vroom-Jago (VJ) model. Our results indicate that actual level of participation used in budget decisions appears to be based primarily on perceived preferences of superiors and participation styles available under the circumstancesof the individual audit. However, results also indicate that decisions consistent with the Vroom-Jago model are characterized by higher decision quality and increased subordinatedevelopment. Since managers are apparently reluctant to use decision styles that conflictwith organizational factors, accounting firms may wish to gather information regarding the attributes of the decision context and use the VJ model to establish their available decision styles and determine the preferences of superiors. Doing so has the advantage of considering decision-specific attributes while enhancing the probability of manager compliance with a desired decision style.     

An Examination of Two Screen/Report Design Variables in an Information Recall Context*

Screen/report design for output organization is still very much a common sense-driven activity. This paper identifies a preliminary set of factors beyond display format for examination as screen/report design variables. A laboratory experiment was conducted to assess the influence of one of these factors, spatial layout of information, along with display format in an information recall context. Individual imagery orientation (verbalizer/visualizer) was used as a moderating variable in the experimental design. The findings indicate that spatial layout of information significantly influenced performance when the recall task involved pattern detection. In fact, a central location of data in a document was the most conducive to pattern recall performance. Further, terminal items facilitated pattern recall better than early input items. With respect to display format, the results support previous findings that a tabular display enhances performance for point-value recall tasks while graphs exhibit an edge over tables for pattern types of recall tasks. Overall, the verbalizers demonstrated a complete supremacy in performance. Interaction between individual imagery orientation and display format indicates that, while the verbalizers and the mixed type were immune to display format variations, the graphical stimulus enhanced the pattern recall performance of the visualizers.