Knowledge that Transforms

We present a formal approach for modeling complex commands characterized by heavy overloading of function, large numbers of parameters, dependencies among parameters, subtle side effects, and lack of abstraction. Complex commands arise in a variety of business settings such as requesting a brokerage order, enrolling in a course, and specifying a product order. In addition, complex commands are also prevalent where specification of commands is strictly separated from multiple, independent implementations as in open software standards. Our approach is based on an inheritance structure known as a command lattice. Like other forms of inheritance, command lattices support incremental definition and abbreviation of specifications. Because a complete command lattice can have a large number of specifications, we develop another structure known as a minimal command tree in which a command lattice is derived from a much smaller number of independent specifications. To map from a minimal command tree to a command lattice, we present algorithms that materialize an arbitrary node of a command lattice and compactly generate the behavior of a command lattice. To demonstrate the potential of command lattices, we have implemented a set of tools that provide convenient specification and powerful reasoning capabilities. Our tool collection includes the Command Specification Language that supports a precise and rich specification of the structural and behavioral properties of commands, the incremental definition tool that ensures consistency of command lattices, the browsing tool that displays a command's inheritance structure, the type checker that ensures structural consistency of commands in expressions, and the target system tracer that simulates a sequence of command executions. We discuss our experiences applying the tools to IBM's Distributed Data Management, a large scale specification of data access on remote and heterogeneous IBM systems.

Benchmarking is the quantitative method most commonly used when managers contemplate procuring a large business information system. It consists of running a group of representative applications on the systems offered by vendors to validate their claims. The implementation of benchmarking can be very costly, as users need to convert, run, and test applications on several partially compatible computer systems. Benchmarking works well in modern database management systems (DBMS)-oriented applications because the system performance is more a function of the database structure and activities than of the complexity of the application code. Earlier research focused primarily on designing various benchmarks for database systems; the decision problem associated with finding an optimal mix of benchmarks has largely been overlooked. In this paper, we examine the problem of defining the most economical process for generating and evaluating the appropriate mix of benchmarks to be used across the contending information systems. Our analytical approach considers information-gathering priorities, acquisition and execution costs, resource consumption, and overall time requirements. We present a multiobjective decision-making approach for deriving the optimal mix of benchmarks; this approach reflects the major organizational objectives in more than simple one-dimensional numerical terms. A practical example illustrates the utility of this approach for evaluating a client-server relational database system.

Control is being increasingly recognized as having a critical role in the automation of the model formulation process. This paper describes an approach to understanding this role of control by observing experts' behavior and studying their verbalizations during the process of formulating models. Control concerns were noted at two levels: strategic and tactical. At the strategic level, control behavior was found to be opportunistic, i.e., the modelers did not follow a prespecified approach. This paper focuses on the tactical level, where the emphasis was on scheduling the tasks used to construct the model (referred to as formulation tasks). The tactical controls were demonstrated by our AEROBA system implementation.

Decisions in large corporations continually evolve from several group processes, shaping the focus of business activities over time. These decisions arise out of a combination of formal analyses and less formal interactions among decision makers. We address the pragmatics of group decision processes from the perspective of argumentation and analysis. We develop formalisms for the representation of argumentative knowledge, gaming the argumentation process and the coordination of the games. The representation formalism provides a framework for organizing the logic underlying the claims and arguments in a group. The gaming formalism provides a framework for conducting and regulating the group interactions. The coordination formalism addresses the issues of scheduling the games and the resolution process. The three formalisms together constitute the basis for designing computer-assisted systems that support argumentation processes in groups. We introduce the term Argumentative Reasoning Facilitation Systems (ARFS), and develop a framework for their design. These systems would serve to record, organize, regulate and coordinate argumentative decision processes in organizations. The formalisms provide new windows for research on novel applications of decision support systems in organizations. Some of the systemic, organizational and behavioral research issues identified from this work are also presented.

This paper evaluates the free-access policy as a control mechanism for internal networks. We derive the optimal message pricing scheme, compare it to the free-access policy, and study the associated net-value loss. We derive uniform upper bounds on this value loss, and apply our results to the polar implementations of ethernet and token ring networks. The results show that the free-access policy is often attractive.

Collopy, Adya and Armstrong (1994) (CAA) advocate the use of atheoretical “black box” extrapolation techniques to forecast information systems spending. In this paper, we contrast this approach with the positive modeling approach of Gurbaxani and Mendelson (1990), where the primary focus is on explanation based on economics and innovation diffusion theory. We argue that the objectives and premises of extrapolation techniques are so fundamentally different from those of positive modeling that the evaluation of positive models using the criteria of “black box” forecasting approaches is inadequate. We further show that even if one were to accept CAA's premises, their results are still inferior. Our results refute CAA's claim that linear trend extrapolations are appropriate for forecasting future IS spending and demonstrate the risks of ignoring the guidance of theory.

Research over two decades has advanced the knowledge of how to assess predictive validity. We believe this has value to information systems (IS) researchers. To demonstrate, we used a widely cited study of IS spending. In that study, price-adjusted diffusion models were proposed to explain and to forecast aggregate U.S. information systems spending. That study concluded that such models would produce more accurate forecasts than would simple linear trend extrapolation. However, one can argue that the validation procedure provided an advantage to the diffusion models. We reexamined the results using an alternative validation procedure based on three principles extracted from forecasting research: (1) use ex ante (out-of-sample) performance rather than the fit to the historical data, (2) use well-accepted models as a basis for comparison, and (3) use an adequate sample of forecasts. Validation using this alternative procedure did confirm the importance of the price-adjustment, but simple trend extrapolations were found to be more accurate than the price-adjusted diffusion models.

This article offers commentary on the paper by Barki and Hartwick (1994), to replicate and extend the model of conflict during systems development reported in previous research by Robey and his colleagues (Robey and Farrow 1982, Robey et al. 1989, Robey et al. 1993). Because of differences in the approach to measurement and data analysis, Barki and Hartwick's contribution is more properly considered as an extension of the model rather than a replication. Barki and Hartwick's strategy of model fitting is appropriate for such an extension, but it is not clear what role their “hypothesized model” plays in this exploratory work. A more careful distinction between hypothesis testing and data exploration is suggested. Finally, all of the studies using the original model or its variants are limited in their ability to support theoretical reasoning about the process of system development. The direct use of process research strategies is encouraged as a means of overcoming this limitation.

The availability of a large and diverse collection of stored modules such as data relations and decision models is a desirable feature in a decision support system (DSS). However, it is usually infeasible to design a DSS in which every problem instance can be solved using a single module. Instead, it may be necessary to combine several stored modules into an integrated model that is sufficient to solve the given problem. We show that modules such as data files and decision models in a DSS can be usefully represented by a metagraph, a graph-theoretic construct that captures relationships between pairs of sets of elements. In addition to the visualization benefits that graphical representation offers, we show that many useful questions faced by the designers and users of DSS can be addressed by exploiting analytical properties of metagraphs. In particular, we show that the process of model integration can be significantly facilitated by exploiting certain connectivity properties in metagraphs.