When we need to answer questions about the real world, we build models. Models are necessarily simplifications, which introduce departures from the real world. Over time we have developed a number modeling paradigms that have been found to work well for certain classes of questions. Each modeling paradigm's simplifying assumptions are appropriate for its class of questions but would not necessarily be appropriate for another class of questions. The simplifying assumptions of one modeling paradigm may be inconsistent with or even contradict the simplifying assumptions of another modeling paradigm. These contradictions are not present in the real world, only in our models. They are introduced by the simplifications. The gaps among the ontologies underlying different paradigms are not bridgeable by a unified ontology that enables addressing questions that cross complex physical, human, economic and social phenomena. We argue instead that the best strategy for addressing complex paradigm-spanning questions is a four-level formulation of optimize if you can; if not, adapt if you can; if not, hedge if you can; if not, accept the consequences.

Link-based analysis of the Web provides the basis for many important applications-like Web search, Web-based data mining, and Web page categorization-that bring order to the massive amount of distributed Web content. Due to the overwhelming reliance on these important applications, there is a rise in efforts to manipulate (or spam) the link structure of the Web. In this manuscript, we present a parameterized framework for link analysis of the Web that promotes spam resilience through a source-centric view of the Web. We provide a rigorous study of the set of critical parameters that can impact source-centric link analysis and propose the novel notion of influence throttling for countering the influence of link-based manipulation. Through formal analysis and a large-scale experimental study, we show how different parameter settings may impact the time complexity, stability, and spam resilience of Web link analysis. Concretely, we find that the source-centric model supports more effective and robust rankings in comparison with existing Web algorithms such as PageRank.

Web ranking is one of the most successful and widely used collaborative computing applications, in which Web pages collaborate in the form of varying degree of relationships to assess their relative quality. Though many observe that links display strong source-centric locality, for example, in terms of administrative domains and hosts, most Web ranking analysis to date has focused on the flat page-level Web linkage structure. In this paper we develop a framework for link-based collaborative ranking of the Web by utilizing the strong Web link structure. We argue that this source-centric link analysis is promising since it captures the natural link-locality structure of the Web, can provide more appealing and efficient Web applications, and reflects many natural types of structured human collaborations. Concretely, we propose a generic framework for source-centric collaborative ranking of the Web. This paper makes two unique contributions. First, we provide a rigorous study of the set of critical parameters that can impact source-centric link analysis, such as source size, the presence of self-links, and different source-citation link weighting schemes (e.g., uniform, link count, source consensus). Second, we conduct a large-scale experimental study to understand how different parameter settings may impact the time complexity, stability, and spam-resilience of Web ranking. We find that careful tuning of these parameters is vital to ensure success over each objective and to balance the performance across all objectives

Business processes continue to play an important role in today's service-oriented enterprise computing systems. Mining, discovering, and integrating process-oriented services has attracted growing attention in the recent year. In this paper we present a quantitative approach to modeling and capturing the similarity and dissimilarity between different process designs. We derive the similarity measures by analyzing the process dependency graphs of the participating workflow processes. We first convert each process dependency graph into a normalized process matrix. Then we calculate the metric space distance between the normalized matrices. This distance measure can be used as a quantitative and qualitative tool in process mining, process merging, and process clustering, and ultimately it can reduce or minimize the costs involved in design, analysis, and evolution of workflow systems

The information technology revolution has driven the pace of competition and rapid globalization. Consequently, enterprises increasingly need to consider and pursue fundamental change - transformation - to maintain or gain competitive advantage. This need raises important research issues concerning how transformation is best understood and pursued. This paper outlines a theory of enterprise transformation to guide research on these issues. The theory focuses on why and how transformation happens, as well as ways in which transformation is addressed and pursued in terms of work processes and the architecture of these processes.

This paper addresses the nature of teams and teamwork in the performing arts, including symphony, chamber orchestra, chorus, and jazz, as well as musical theater, straight theatre, improv, ballet, and puppetry. The results of an interview study of performing arts leaders in these domains are reported. These results suggest an "ecology" of performance. The characteristics of this ecology strongly influence the nature and roles of teams, as well as how teams are created and supported. Potential relationships among the attributes of this characterization are discussed.

A discussion of the scope of systems engineering, which seems to cover everything and therefore, in a way, to cover nothing. Systems engineering crosses the boundaries of other disciplines. It is the exploration, understanding, and design of how everything fits together.

The success of information and knowledge management depends on understanding and supporting the user's need to know. This requires understanding humans' abilities, limitations, and inclinations in seeking of information and knowledge. This paper explores these phenomena in the context of two decades of studying human decision making in the domains of research, design, and management. The findings summarized are discussed in terms of implications for design, development, and deployment of information and knowledge support systems.

This article discusses the insights gained from of a series of observations of cross-functional planning teams in a wide variety of enterprises. These observations involved over 2,000 participants in 100 planning efforts over a period of 6 years. These efforts focused on planning of new products or processes, development of strategic business plans, and consensus building around planned organizational changes. The insights reported in this article emphasize the impact of computer support on planning processes. © 1998 John Wiley & Sons, Inc.

The article summarizes and integrates the state of knowledge of R&D/technology management. The goal is to clarify what is known and discuss how this knowledge can be leveraged. The overall conclusion is that understanding at multiple levels is key to deeply understanding R&D/technology management. Further, these multiple levels of understanding can be integrated by considering information access and utilization throughout these processes. Specific conclusions are summarized in terms of key issues, central questions, and tradeoffs.

Most of the world's archive of knowledge sources remains untouched by those who could greatly benefit from accessing and utilizing this information. The fundamental barriers to access and utilization by these users involve the content and format of this information. These users are not particularly interested in data compilations, research methods, and discussions of related literature. Instead, they are interested in the knowledge available from these sources in terms of concepts and principles that can enhance their planning, decision making, and allocation of resources. The article focuses on meeting these needs via a six step knowledge mining methodology that relies upon user oriented knowledge maps. This concept is illustrated in the context of R&D/technology management.

A methodology is presented for assessing the economic and noneconomic cost/benefits of investments in research and development. It is argued that the most difficult aspect of cost/benefit analysis is not calculation, but representation of stakeholders, products, benefits, and costs. A stakeholder-based methodology is described that focuses on these representation issues. Use of this methodology is illustrated with a case study of a long-term government research and development (R&D) program.

Describes experiments that compare air traffic controller performance with two different types of simulated airport radar display formats. The formats differed in the method of presenting aircraft radar returns: analog or digital symbols. Pre-experimental studies examined potential airport tower automation systems and the opinions of various system stakeholders about these automated systems, and then modeled airport operations to discern the impact of these systems on controller workload and airport congestion. Results from the earlier studies helped focus the experiments which concluded that the digital icon format is likely to be better than the analog format at helping controllers accomplish their duties in a timely manner.

The relationship between subjective workload and human behaviors is considered. A model of the dynamics of this relationship is proposed. Results of three simulation experiments are reported that show simple linear identification algorithms are surprisingly robust in online identification of noisy, nonlinear versions of this model. The implications for developing adaptive aiding and decision support systems are noted.>

Traditional maintenance information includes large graphical drawings that are difficult to portray on the small screens of computer-based maintenance information systems. The research described investigates the design of graphical displays using display abstraction and aggregation as design parameters. A display's aggregation level reflects the field of view of a display, such as component, assembly, or system-level diagrams. A display's abstraction level reflects the representation contained in a diagram, such as a component's form, function, or purpose in an assembly. The results from two experiments with experienced maintenance personnel are presented. Usage of the display abstraction levels was found to vary with both the nature of the maintenance task and the experience level of the maintenance personnel. Usage of the display aggregation levels was found to change as task performance proceeded. The results support the application of these concepts to computer-based maintenance information systems.>

The authors focus on team performance in complex systems. Representative empirical literature is reviewed and models of team performance are discussed. The role of mental models in team performance is considered and several propositions developed that focus on mental models as mechanisms for forming expectations and explanations of team behaviors. The implications of these propositions for team performance and training are elaborated, particularly in terms of likely performance problems if mechanisms for forming expectations and explanations are deficient. The results of two initial studies that support the plausibility of the propositions are reported.>

The authors discuss validating the knowledge bases underlying intelligent systems. In particular, the concern is with identifying and remediating modeling limits that are likely to lead to unacceptable consequences in terms of system performance or safety. A general conceptual framework for considering the roles in intelligent systems of models of physical, behavioral, and operational phenomena is given. Ten types of modeling limits are identified, and their likely consequences discussed. A methodology for identifying limits in particular intelligent systems is described and illustrated via an evaluation of the pilot-vehicle interface within the Pilot's Associate. The essence of this methodology involves interactive probing and tracing of concept graph representations of knowledge bases.>

Presents a review of some of the areas to which supervisory control theory has been applied and indicate some of the ideas which have as yet not been followed up but which were introduced by Sheridan and his students. Topics covered include: modeling manual and supervisory control tasks; supervisory control of deep-ocean vehicles; technology problems and technology solutions; human integration in a complex world; supervisory control in distributed decision-making; and human supervisor control models in advanced manufacturing systems.>

Algorithmic identification of models of human skills and knowledge is considered. Alternative representational forms are discussed in terms of variables, relationships among variables, and parameters within relationships. A key distinction is between models of signal processing and models of symbol processing. Methods of identification for these two classes of models are discussed and contrasted. Identifiability, or the uniqueness of models identified, is considered for both classes of model, and a variety of fundamental limits relating to existence and computability are reviewed. Practical issues and results associated with identification are considered in the context of three examples of identifying signal processing and symbol processing models. The discussion of available methods and known limits concludes with consideration of the general implications for endeavors aimed at describing and explaining human skills and knowledge.>

Results of two experiments in dynamic task allocation are discussed. Subjects performed two concurrent computer-based tasks: visual target identification and subcritical compensatory tracking. Target identification could be allocated dynamically to human or computer aid. Three aiding conditions were investigated: no aid, manual aid (with subjects making the allocation decision), and automatic aid (with allocation decisions based on models of human performance). The results indicated that: (1) overall performance was better with the aid available; (2) need for the aid depended on both current and previous task states; (3) unaided performance was benefited by having an aid available, but only if subjects were in charge of task allocation; and (4) although overall performance was better with he automatic aid, subjects preferred the manual aid. The implications of these and other results are discussed.>

The conceptual design of a comprehensive support system for operators of complex systems is presented. Key functions within the support system architecture include information management, error monitoring, and adaptive aiding. One of the central knowledge sources underlying this functionality is an operator model that involves a combination of algorithmic and symbolic models for assessing and predicting an operator's activities, awareness, intentions, resources, and performance. Functional block diagrams are presented for the overall architecture as well as the key elements within this architecture. A variety of difficult design issues are discussed, and ongoing efforts aimed at resolving these issues are noted.

The complexity of monitoring and controlling a large-scale system, such as a communication network, is considered. Relevant literature is reviewed, with emphasis on both behavioral and nonbehavioral approaches to measuring complexity. A simulated large-scale network is described that is used in an experiment to assess the effect of network redundancy and number of systems levels on human fault-diagnosis performance. Experimental data are also used to evaluate two time-varying measures of task complexity (using analysis of variance and time-series analysis). The first measure is dependent on the structure of the system; the second measure is dependent on the strategy of the person controlling the system. Results suggest that this distinction is appropriate. In addition, results emphasize the different implications that complexity can have for normal system operation and human failure diagnosis performance. Although system design characteristics such as redundancy may help to avoid the short-term effects of failures, these same characteristics may have the dual effect of making the human supervisory controller's task more difficult.

Technology-driven efforts to implement computer technology often encounter problems due to lack of acceptance or begrudging acceptance of the personnel involved. It is argued that individuals' acceptance of automation, in terms of either computerization or computer aiding, is heavily influenced by their perceptions of the impact of the automation on their discretion in performing their jobs. It is suggested that desired levels of discretion reflect needs to feel in control and achieve self-satisfaction in task performance, as well as perceptions of inadequacies of computer technology. Discussion of these factors leads to a structured set of considerations for performing front-end analysis, deciding what to automate, and implementing the resulting changes.

The central ideas and results, as seen by the authors, of a workshop, sponsored by the Army Research Institute. The workshop concerned aiding and understanding the human decisionmaker through the knowledge-based sciences and was held in Northern Virginia in 1985. The purpose of the workshop was to explore research issues in human decisionmaking in terms of how they might be addressed by several disciplinary areas concerned with various aspects of knowledge-based systems. The goal was to develop a research agenda that takes advantage of the commonality and differences in these areas so as to create research programs with substantially greater potential than possible if only a single discipline was involved. Such an agenda and potentially high payoff research in knowledge-based support to human information processing and judgment and choice are described.

Human problem solving is considered with emphasis on situations involving human-machine interaction in detecting, diagnosing, and compensating for failures in engineering systems. An overall model is presented which considers the breath and robustness of human problem-solving behavior in dynamic environments typical of engineering systems. A realization of the general structure of this model within a particular rule-based computer program is discussed. In this program the human behavior in controlling a dynamic process is represented by a set of production rules. The selection of appropriate production rules for a given situation is performed by ordering the rules for specific tasks and by a control mechanism. The latter provides a means to access only a small part of the knowledge base at a time in order to derive decisions as opposed to searching the whole knowledge base. This program was applied to model human problem solving in a process control task. Results from comparing model and subject behavior are presented and discussed.

The question of what the operator of a dynamic system needs to know was investigated in an experiment using PLANT, a simulation of a generic dynamic production process. Knowledge of PLANT was manipulated via different types of instruction, so that four different groups were created: (1) minimal instructions only; (2) minimal instructions and guidelines for operation (procedures); (3) minimal instructions and dynamic relationships (principles), and (4) minimal instructions and procedures and principles. Subjects controlled PLANT in a variety of situations that required maintaining production while also diagnosing familiar and unfamiliar failures. Despite the fact that these manipulations resulted in differences in subjects' knowledge, as assessed via a written test at the end of the experiment, instructions had no effect on achievement of the primary goal of production or on subjects' ability to diagnose unfamiliar failures. However, those groups receiving procedures controlled the system in a more stable manner. Possible reasons for the failure to find an effect are presented, and the implications of these results for operator training and aiding are discussed.

An experimental tool for the investigation of human problem-solving behavior is introduced. Production levels and network troubleshooting (PLANT) is a computer-based process-control task which may be used to provide opportunities for subjects to control a dynamic system and diagnose, repair, and compensate for system failures. The task is described in detail, and experiments which have been conducted using PLANT are briefly discussed.

Issues surrounding the relationship between the basic research community and the applied world are discussed. Considerations include the contrasting perspectives of the two worlds and the resulting differences in the ways in which design and evaluation are approached. These contrasts are used as a basis for suggesting mechanisms for improving transfer of basic research results to real world applications.

A need exists to reduce the frequencies of consequential errors and/or develop systems that are error-tolerant in the sense that the undesirable consequences of errors do not propagate. Reduction and/or tolerance can be accomplished with a variety of mechanisms, including selection, training, equipment design, job design, and aiding. Since no single mechanism is sufficient, a mix of mechanisms is needed. An important question concerns how one should allocate resources among these mechanisms to achieve acceptable frequencies of consequential errors. This resource allocation problem is formulated, first in terms of a conceptual model, and them in terms of a mathematical model of the effects of resources on error reduction/tolerance mechanisms. Sensitivity analysis is used to assess the effects of parameter variations on the model and resulting optimal allocation. It is suggested that this approach provides a comprehensive framework within which the problem of human error can be studied and ameliorated.

The literature relating to human performance in fault diagnosis tasks is reviewed with emphasis on measures of performance. Based on this review, 30 measures of ability, aptitude, cognitive style, and task performance are evaluated, using data from two experiments that involved diagnosing faults in both simulated and live equipment. Results based on correlation, regression, and factor analysis are presented that indicate only three unique dimensions of performance: errors, inefficiency, and time. In addition, cognitive style appears to be a reasonable predictor of performance. Ability, measured by standard precollege tests, is also a reasonable predictor when combined with measures of cognitive style.

Human performance in monitoring and controlling activities in a hierarchical large-scale network, such as a communications system is considered. A scenario is described that is used in an experiment to examine three factors affecting humans functioning as network supervisors: cluster size (number of elements per display page), number of levels of pages in the hierarchy and failure rate per elements. The results indicate that increasing cluster size improves performance, increasing number of levels degrades performance, and failure rate affects only subjects' strategies.

A fuzzy rule-based model of human problem-solving is described. The model is presented in its general form and then adapted to fit data from a simulated fault diagnosis task. The model was able to match 50% of human subjects' actions exactly while using the same rules approximately 70% of the time. Problem-solving rules were selected by the model according to measures of recall, applicability, usefulness, and simplicity. Rules were further discriminated by their use of symptomatic information for pattern recognition or topographic information for information seeking.

Almost 200 documents reporting results of evaluative studies of complex man-machine systems are classified in terms of domain of study, method, approach to measurement, type of data, data collection methods, measures assessed, and nature of results. An analysis of the degree to which definitive evaluation results were produced as a function of a study's classification is discussed. The results are used to suggest several guidelines for design of evaluation efforts.

Planning will become an increasingly central function for the human operator as automation takes over more and more of the manual tasks associated with man-machine systems. In order to understand and aid the human in this role, an experimental methodology for the study of human planning behaviour is needed. Further, a model of the process by which the human plans is desirable. A methodology is presented for studying planning, together with the results of applying the methodology within two experimental investigations of planning behaviour of aircraft pilots in normal, abnormal, and emergency situations. Beyond showing that the methodology yields consistent results, these experiments also produced new concepts in terms of a dichotomy between event-driven and time-driven planning, subtle effects of automation on planning, and the relationship of planning to workload and light performance.

The literature on human error is reviewed with emphasis on theories of error and classification schemes. A methodology for analysis and classification of human error is then proposed which includes a general approach to classification. Identification of possible causes and factors that contribute to the occurrence of errors is also considered. An application of the methodology to the use of checklists in the aviation domain is presented for illustrative purposes.

It is proposed that human decisionmaking performance in multiple process monitoring situations can be modeled in terms of the detection of process related events and the allocation of attention among processes once events are felt to have occurred. An elementary pattern recognition technique, discriminant analysis, is used to generate estimates of event occurrence probability. A queueing theory framework is then utilized to incorporate these probabilities as well as other task characteristics into the solution of the attention allocation problem. The performance of the model is compared with that of subjects in two experiments.

A model of human performance on intraline text editing, the modification of a line of text, is presented. This model is based on an optimal minimum keystroke solution to the text modification plus a description of those keystrokes that are not easily predicted: feedback, command selection idiosyncrasies, and errors. To better match the human, who is assumed to employ an optimal sequence of editing commands, the optimal solution-was constrained to use the same types of command each human had been observed to use. The model's constraints and descriptions of keystrokes both give rise to an increased understanding of text editing and recommendations for editor design improvements.

Information seeking by human operators of technical systems is considered. Types of information and forms of presentation are discussed and important issues reviewed. This broad discussion provides a framework within which flight management is considered. The design of an onboard computer-based information system for aircraft is discussed. The aiding possibilities of a computer-based system are emphasized. Results of an experimental evaluation of a prototype system are presented. It is concluded that a computer-based information system can substantially lessen the frequency of human errors.

In complex environments where the human operator is a supervisor, he must allocate his attention between different kinds of tasks for satisfactory overall performance. When a portion of the future reference trajectory for a continuous control task is available for preview, scheduling various other discrete activities is possible. A model has been developed for this situation using dynamic programming to solve an optimal control problem. An experiment was conducted where subjects controlled an airplane symbol over a map, shown a fixed distance into the future. Discrete tasks were introduced as data entry tasks. Results showed that the model compared favorably with experimental results.

Abstract The problem of forecasting monthly demands for library network services is considered, especially in terms of using forecasts as inputs to policy analysis models and in terms of the use of forecasts as an aid to budgeting and staffing decisions. Forecasting methods considered include Box‐Jenkins time‐series methodology, adaptive filtering, and linear regression. Using demand data from the Illinois Library and Information Network for 1971–1978, it is shown that fading‐memory regression is the most appropriate method, in terms of both accuracy and ease of use.

Human problem-solving is considered with emphasis on situations involving human-machine interaction in detecting, diagnosing, and compensating for failures in engineering systems. An overall model is presented that considers the breadth and robustness of human problem solving behavior in dynamic environments typical of engineering systems. A realization of the general structure of this model within a particular rule-based computer program is discussed. In this program the human behavior in controlling a dynamic process is represented by a set of production rules. The selection of appropriate production rules for a given situation is performed by ordering the rules for specific tasks and by a control mechanism. The latter provides a means to access only a small part of the knowledge base at a time in order to derive decisions as opposed to searching the whole knowledge base. This program was applied to model human-problem solving in a process control task. Results from comparing model and subject behavior are presented and discussed.

Abstract The purpose of mathematical models is discussed and the phases of the modeling process are reviewed. These phases include: defining performance, representing the problem, predicting performance, estimating parameters, defining an optimization criterion, determining the optimal solution, and implementing the results. A selected review of models applied to library problems is presented. Models for book use, resource allocation, and library networks are considered.

As human and computer come to have overlapping decisionmaking abilities, a dynamic or adaptive allocation of responsibilities may be the best mode of human-computer interaction. It is suggested that the computer serve as a backup decisionmaker, accepting responsibility when human workload becomes excessive and relinquishing responsibility when workload becomes acceptable. A queueing theory formulation of multitask decisionmaking is used and a threshold policy for turning the computer on/off is proposed. This policy minimizes event-waiting cost subject to human workload constraints. An experiment was conducted with a balanced design of several subject runs within a computer-aided multitask flight management situation with different task demand levels. It was found that computer aiding enhanced subsystem performance as well as subjective ratings. The queueing model appears to be an adequate representation of the multitask decisionmaking situation, and to be capable of predicting system performance in terms of average waiting time and server occupancy. Server occupancy was further found to correlate highly with the subjective effort ratings.

Library networks are becoming increasingly complex public systems. Because of this trend, the management of library networks can perhaps be enhanced by the use of policy analysis models. Within this paper, such a model is briefly reviewed. Three case studies have been performed with this model. The recommendations resulting from these studies are discussed. Special emphasis is placed on what appear to be key ingredients to successfully applying the modeling approach. Further, the management implications of model-based policy analyses are considered.

Abstract A library network model is reviewed and a new derivation of the model is presented that requires fewer assumptions and allows more robust representations of library networks. It is then shown how uncertainty in model parameters, due to estimation on the basis of finite sample sizes, propagates through the model and thereby introduces uncertainty in the model's predictions. Using these results, it is shown how one can determine the appropriate amount of data to collect and/or the appropriate level of aggregation of data.

It is proposed that humans in automated systems will be asked to assume the role of troubleshooter or problem solver and that the problems which they will be asked to solve in such systems will not be amenable to rote solution. The design of visual displays for problem solving in such situations is considered, and the results of two experimental investigations of human problem solving performance in the diagnosis of faults in graphically displayed network problems are discussed. The effects of problem size, forced-pacing, computer aiding, and training are considered. Results indicate that human performance deviates from optimality as problem size increases. Forced-pacing appears to cause the human to adopt fairly brute force strategies, as compared to those adopted in self-paced situations. Computer aiding substantially lessens the number of mistaken diagnoses by performing the bookkeeping portions of the task.

Utilizing elementary concepts from the theory of fuzzy sets as well as several nonfuzzy heuristics, a model is presented of human decisionmaking in the task of troubleshooting graphically displayed networks. The performance of the model is compared to the results of two previously reported experimental studies. The ability of the model to represent human decisionmaking as a function of network size, forced-pacing, and computer aiding is considered.

Allocation of decisionmaking responsibility between pilot and computer is considered, and a flight management task, designed for the study of pilot-computer interaction, is discussed. A queueing theory model of pilot decisionmaking in this multitask control and monitoring situation is presented. An experimental investigation of pilot decisionmaking and the resulting model parameters are discussed.

Abstract Performance of interlibrary loan networks in terms of probability of success and average time to satisfy a request is enhanced when location and availability information can be accessed. Existing computer technologies such as shared cataloging networks and automated circulation systems can be of use in obtaining this information. A procedure is presented for quantitative assessment of the impact of these technologies and their various combinations on interlibrary loan activities. As an example, the procedure is utilized for predicting the impact of these computer technologies on the Illinois Library and Information Network (ILLINET). Results show that the value of location information as obtained from a shared cataloging network or similar technology is highly dependent on the information being specific enough to free the lending library from searching their own main catalog. The value of availability information is shown to be related to the processing time that can be avoided by having prior information about the circulation status of the desired item. These results are dependent on the policies employed in Illinois. However, the assessment procedures presented here have general applicability to interlibrary loan networks.

A computer algorithm employing fading-memory system identification and linear discriminant analysis is proposed for real-time detection of human shifts of attention in a control and monitoring situation. Experimental results are presented that validate the usefulness of the method. Application of the method to computer-aided decisionmaking in multitask situations is discussed.

Human decisionmaking in stochastic estimation tasks is considered. Theoretical and experimental results from mathematical psychology are reviewed. Concepts from stochastic estimation theory are used to develop a theory of human decisionmaking that employs optimal stochastic estimators with a short-term memory model, at least one long-term model, and a method of trading off estimates derived from each model. Approaches to testing the theory as well as the theory's implications are discussed. Also, the limitations of a linear theory are considered.

Human-computer interaction in multitask decisionmaking situations is considered, and it is proposed that humans and computers have overlapping responsibilities. Queueing theory is employed to model this dynamic approach to the allocation of responsibility between human and computer. Results of simulation experiments are used to illustrate the effects of several system variables including number of tasks, mean time between arrivals of action-evoking events, human-computer speed mismatch, probability of computer error, probability of human error, and the level of feedback between human and computer. Current experimental efforts are discussed and the practical issues involved in designing human-computer systems for multitask situations are considered.

Abstract The performance of an interlibrary loan network is characterized by the probability of satisfying a request, delay in satisfying a request, total and unit costs, and processing load on each network member. In this article, a distributed queueing network is used to model this situation. The general structure can be specialized to a ring structure or a centralized structure. A hypothetical network situation is used to discuss the effects of network dispersion and distribution of resources on several policy issues. These issues include request routing, reimbursement and possible investments in computer technology.

Estimation theory is used to develop a model of a human decision maker in a self-paced, visual data smoothing task. The model is basically a noisy fixed-point smoother combined with a weighting function that discounts data in relation to its distance (in time) from the point being smoothed. Experimental data is used to estimate the parameters of the model. Applications of the model to the design of information displays and man-computer interactive decision making systems are considered.

The human's ability to estimate the mean and standard deviation of discrete time series is discussed. The effect of polynomial interpolation formulas on these two statistics is considered both mathematically and experimentally. Derivations show that the piece-wise continuous functions resulting from polynomial interpolation have unbiased expected values and biased variances. Results of two experiments are presented that support the idea that humans base their perceptions on the piece-wise continuous result of interpolation and thereby transmit the statistical bias. The practical implications of these results are discussed.

Abstract A procedure is developed for optimal allocation of resources among the many processes of a library system. Queueing theory is used to model processes as either waiting or balking processes. The optimal allocation of resources to these processes is defined as that which maximizes the expected value of the decision‐maker's utility function. An application of the procedure to a specific library system is discussed.

Abstract The problem addressed is that of optimally choosing the source from which to order a book. Each acquisition source is characterized by frequency distributions for discount and service time. A technique called decision analysis is used to determine the tradeoff between discount and service time most consistent with the librarian's preferences.

Abstract A mathematical model is presented that can be used to predict a library's circulation. Variables considered include changing user population, decreasing demand for older items in the collection, purchasing and circulation policies, weeding and other losses, and inflation. The model requires little data other than what libraries normally collect. An example of applying the model to a particular library is discussed. Possible extensions of the model are noted.

The human decision maker's behavior when predicting future states of discrete linear dynamic systems driven by zero-mean Gaussian processes is modeled. The task is on a slow enough time scale that physiological constraints are insignificant compared with cognitive limitations. The model is basically a linear regression system identifier with a limited memory and noisy observations. Experimental data are presented and compared to the model.