A cognitive explanation of the correctness of information requirement specifications.

Huang, I-Lin

INTRODUCTION

It is widely recognized that incorrect requirement specifications are the major cause of system failures (Dardenne, van Lamsweerde, & Fickas, 1993; Davis, 1988; Dorfman, 1990; Greenspan & Mylopoulos, 1982; Scharer, 1981; Standish Group, 1995; Vessey & Conger, 1994). It has been reported that about two thirds of system failures can be attributed to the mistakes made in requirement specifications (Fraser, Kumar, & Vaishnavi, 1991; Shemer, 1987). According to an estimation, incorrect requirement specifications which are not discovered until system implementation may cost fifty to one hundred times more than what would have been required if the errors were discovered during information requirement analysis (Roman, April 1985; Shemer, 1987).

In order to identify the determinants for getting information requirements right, cognitive research has been conducted to study the process of information requirement analysis. One stream of research has investigated the differences in analytical behaviors between novice and expert information analysts. Richer domain knowledge and modeling knowledge have been recognized as the qualities of expert information analysts for better performance in information requirement analysis (Adelson & Soloway, 1985; Allwood, 1986; Koubek et al., 1989; Schenk, Vitalari, & Davis, 1998; Sutcliffe & Maiden, 1990; Vessey & Conger, 1993; Vitalari & Dickson, 1983). Another stream of research has focused on comparing the effectiveness of various requirement analysis techniques in specifying information requirements. However, the research results are contradictory (Kim & Lerch, 1992; Vessey & Cogner, 1994; Yadav, Bravoco, Chatfield, & Raikumar, 1988). No requirement analysis technique has shown consistently better performance than the others (Kim & Lerch, 1992; Poti & Ramesh, 2002; Vessey & Cogner, 1994; Yadav, Bravoco, Chatfield, & Raikumar, 1988). In addition, experienced information analysts use multiple requirement analysis techniques in analyzing complex information systems (Littman, 1989). In order to explain the contradictory results, some researchers have suggested that requirement analysis techniques should be matched to types of problem domains (Fitzgerald, 1996; Jackson, 1994; Vessey & Glass, 1994). Several frameworks have also been proposed to classify requirement analysis techniques on the basis of problem domains (Davis, 1988; Iivari, 1989; Marca & McGowan, 1993; Vessey & Glass, Fall 1994).

RESEARCH QUESTION

In order to identify and explain the important variables for the correctness of requirement specifications, various cognitive models have been built on the basis of different cognitive theories such as normative models, problem solving models, mental models, and comprehension models. However, they are inadequate because they are focused on some rather than all of the three determinants: the knowledge of information analysts, requirement analysis techniques, and problem domains. As a result, at least two issues related to requirement analysis techniques cannot be accounted for: First, no requirement analysis technique can consistently outperform the others. Second, experienced information analysts use multiple requirement analysis techniques in analyzing complex information systems.

Without an adequate model of information requirement analysis, research studies may miss influential variables in viewing and comparing the cognitive processes of information requirement analysis, resulting in erroneous findings. Therefore, the research question for this research is: Is there an adequate cognitive model that can not only explicate the interactive relationships among the three determinants for the correctness of requirement specifications, but also explain the two unsolved issues for requirement analysis techniques?

To answer the research question, this article proposes a cognitive model of information requirement analysis on the basis of the structure building theory of language comprehension (Gernsbacher, 1990). By modeling information requirement analysis as a process of text comprehension, the cognitive model can not only explicate the interactive relationships among the three important determinants for the correctness of requirement specifications, but also provide an explanation for the two unsolved issues on requirement analysis techniques.

The rest of this article is organized as follows. First, I review the characteristics of information requirement analysis, the cognitive variables for the performance of information requirement analysis, and the cognitive models used in current cognitive research of information requirement analysis. Then, I propose a more adequate cognitive model of information requirement analysis on the basis of the structure building theory of language comprehension. Third, in order to show the adequacy of the proposed cognitive model, I use the cognitive model to provide an explanation on the two unsolved issues related to requirement analysis techniques. Fourth, I validate the proposed cognitive model theoretically by the research findings from the related cognitive research studies. Sixth, I discuss the implications of this research. Finally, I make the conclusion in the final section.

LITERATURE REVIEW

In this section, I first give a brief overview on the characteristics of information requirement analysis. Then I review the cognitive variables important for the correctness of requirement specifications from the cognitive research in information requirement analysis. Finally, I review current cognitive models of information requirement analysis on the basis of their underlying cognitive theories. Due to the similarity of cognitive requirements between the upper-level system design and information requirement analysis, some cognitive models of the upper-level system design are also discussed.

Information Requirement Analysis

Information requirement analysis is the first stage of information systems development. Basically, information requirement analysis can be characterized by three features: the major outputs, the activities, and the roles involved in information requirement analysis. First, the major outputs of information requirement analysis are requirement specifications. Requirement specifications are mainly used as blueprints for creating the intended information systems. In order to support information system development effectively, requirement specifications should reflect an understanding of the intended information system, guide the subsequent design, and serve as a basis for all communications concerning the information system (Shemer, 1987). In order to facilitate an understanding of the target system, requirement specifications need to address user-level systelogical and infological issues. To serve as a communication tool during information system development, requirement specifications should be understandable to naive users, easy to modify, and maintainable. For the purpose of guiding the subsequent design, requirement specifications need to state the required functional and performance characteristics of information systems independent of any actual realization (Davis, 1988; Roman, April 1985).

Second, in order to generate requirement specifications, the activities of information analysts can be categorized into the following four purposes: "(1) identification and documentation of customer and user need, (2) creation of a document that describes the external behavior and its associated constraints which will satisfy those needs, (3) analysis and validation of the requirements document to ensure consistency, completeness, and feasibility, and (4) evolution of needs" (Hsia, Davis, & Kung, November 1993, p. 75).

Finally, there are three roles involved in the process of information requirement analysis: users, information analysts, and system designers (Fraser, Kumar, & Vaishnavi, 1991). Users have information requirements for the information systems under development. System designers have the responsibility of designing the information systems on the basis of the users' information requirements. Information analysts work in between users and system designers. Information analysts collect, understand, and document information requirements from users, and finally pass the requirement specifications to system designers.

Cognitive Variables

The research in the correctness of requirement specifications has been conducted along three dimensions: the knowledge of information analysts, requirement analysis techniques, and problem domains. First, richer domain knowledge and modeling knowledge have been suggested as important factors for better modeling performance of expert information analysts. Domain knowledge is drawn upon by both expert and novice information analysts in specifying information requirements (Sutcliffe & Maiden, 1990; Vessey & Conger, 1993). While understanding problem statements, information analysts use domain knowledge to mentally simulate a scenario of the system behavior in order to test the adequacy of the requirement specifications, to add assumptions to increase the completeness of the requirements, to test internal and external consistency of the requirements, and to abstract, summary, select and highlight important information in the problem statements (Guindon, Krasnar, & Curtis, 1987). Without domain knowledge, even expert information analysts can only specify high-level conceptual models without details (Adelson & Soloway, 1985). With the availability of domain knowledge, novice information can reuse the domain knowledge to achieve almost the same level of completeness of requirement specifications as expert information analysts do (Sutcliffe & Maiden, 1990). On the other hand, modeling knowledge has long been regarded as an important factor to differentiate expert from novice information analysts. Modeling knowledge can be divided into syntactic and semantic parts (Koubek et al., 1989). Syntactic knowledge consists of allowable syntax of a specific modeling language. Semantic knowledge, however, consists of modeling principles which are independent of a particular modeling language (Allwood, 1986). Compared to novice information analysts, expert information analysts with richer semantic knowledge can retrieve and apply more relevant modeling principles, make more critical testing of hypotheses, and finally achieve requirement specifications with better quality (Allwood, 1986; Koubek et al., 1989; Schenk, Vitalari, & Davis, 1998; Vitalari & Dickson, 1983). Modeling knowledge can also be divided into declarative and procedural aspects (Vessey & Conger, 1993). The procedural aspect of a requirement analysis technique is more difficult to learn than the declarative aspect. However, the procedural aspect of modeling knowledge is more important in determining the correctness of requirement specifications (Vessey & Coger, 1993).

Second, research into requirement analysis techniques has focused on comparing the effectiveness of various requirement analysis techniques in specifying information requirements. The purpose of requirement analysis techniques is to provide notation and procedures to help information analysts formalize the domain knowledge of problem domains during the process of information requirement analysis (Sutcliffe & Maiden, 1992). It was found that information analysts who specified information requirements by model-based reasoning based on requirement analysis techniques coud produce more complete solutions than those with partial or no model-based reasoning behavior (Sutcliffe & Maiden, 1992). However, it was also found that novice information analysts had difficulties in identifying important concepts of problem statements with requirement analysis techniques (Batra & Davis, 1992; Batra & Sein, 1994; Sutcliffe & Maiden, 1992). Yadav, Bravoco, Chatfield, & Raikumar (1988) compared the effectiveness of data flow diagrams and IDEF for supporting novice information analysts in specifying information requirements. They found that data flow diagrams were easier to learn and to use. However, neither of them produced significantly better specifications. While object orientation has become a new paradigm for information requirement analysis, there is a debate on which approach, the object orientation or the functional orientation, is a more "natural" way to specify information requirements (Firesmith, 1991; Loy, 1990; Shumate, 1991). The results on the basis of empirical studies are inconclusive. Kim and Lerch (1992) reported that expert information analysts with object-oriented techniques spent less time in analyzing problem domains and developed better understanding of the underlying problem structures than expert information analysts with functional-oriented techniques. However, Vessey and Cogner (1994) found that novice information analysts were better able to apply functional-oriented techniques than to apply object-oriented techniques. In addition, significant learning effects only occurred for functional-oriented techniques.

Finally, the research in the dimension of problem domains argues that the characteristics of problem domains should be the basis for the selection of requirement analysis techniques for information requirement analysis. Littman (1989) conducted several empirical studies to investigate the ways in which expert software designers constructed mental representations of problem domains. He found that expert software designers used multiple mental representations to model problem domains. Littman reported that expert software designers identified several modeling techniques that might be appropriate for a problem domain and then selected one that seemed most appropriate. Vessey and Glass (1994) argued that cognitive fit between problem domains and requirement analysis techniques was important for the effectiveness of information requirement analysis. They suggested that taxonomies of problem domains and taxonomies of requirement analysis techniques were needed to facilitate matching techniques to problem domains. To match requirement analysis techniques to problem domains, requirement analysis techniques have long been classified into functional orientation, data orientation, control orientation, or object orientation (Dorfman, 1990). Marca and McGowan (1993) classified requirement analysis techniques on the basis of the theory of world views in metaphysics. Iivari (1989) added levels of abstraction as another dimension to classify requirement analysis techniques. Sowa and Zachman (1992) provided a framework to categorize requirement analysis techniques on the basis of six dimensions: data, process, network, people, time, and motivation. Opdahl and Sindre (1995) proposed a facet-modeling structure to integrate various requirement analysis techniques. Jackson (November 1994) suggested that future requirement analysis techniques should be more problem-oriented to fit the structures of problem domains rather than solution-oriented.

In this section the cognitive variables determining the performance of information requirement analysis are discussed along three dimensions: requirement analysis techniques, the knowledge of information analysts, and types of problem domains. Research evidence shows that the three dimensions are highly correlated in determining the correctness of requirement specifications. However, the interactive relationships among the three dimensions are still unclear. As a result, there are at least two important issues related to requirement analysis techniques remaining unsolved: First, why no requirement analysis technique can have consistently better performance than others (Kim & Lerch, 1992; Poti & Ramesh, 2002; Vessey & Cogner, 1994; Yadav, Bravoco, Chatfield, & Raikumar, 1988)? Second, why expert information analysts use multiple requirement analysis techniques in analyzing information requirements (Littman, 1989)? On the basis of previous research evidence, we cannot predict the influence of requirement analysis techniques upon the correctness of requirement specifications without knowing first the types of problem domains and the knowledge of information analysts. In order to solve the two issues, I will explore the interaction among the three dimensions through the cognitive models proposed in cognitive research of information requirement analysis.

Cognitive Models

In order to identify and explain important variables for the correctness of requirement specifications, various cognitive models have been proposed on the basis of different cognitive theories. Basically, four approaches have been used to derive cognitive models of information requirement analysis: normative models, problem solving models, mental models, and comprehension models.

First, normative models are referred to as the models of information requirement analysis that are built on the basis of the researchers' experiences or opinions. Normative models are often built for comparing or evaluating requirement analysis techniques. Examples can be found in the research papers of Davis (1988); and Yadav, Bravoco, Chatfield, and Rajkumar (1988). This class of models provides a set of criteria about what should be achieved by good requirement analysis techniques. However, without an understanding of the cognitive behaviors of information analysts, those models provide no guideline for how to support information analysts in understanding problem domains and in specifying correct information requirements.

Second, some researchers believe that information requirement analysis is similar to the cognitive process of problem solving (Malhotra et al., 1980; Schenk, Vitalari, & Davis, 1998; Sutcliffe & Mainden, 1992; Vitalari & Dickson, 1983). They focus on the reasoning processes that information analysts use to analyze information requirements. Specifically, they focus on "the frequency, ordering, and association with analysts' performance of the clues, goals, strategies, heuristics, hypotheses, information, and knowledge manifested in the thought process of the information analysts" (Vitalari & Dickson, 1983, p. 949). Some examples can be found in the research papers of Malhotra et al. (1980); Vitalari and Dickson (1983); and Sutcliffe and Mainden (1992). On the basis of the problem solving paradigm, this class of cognitive models provides a framework for understanding the influence of knowledge and reasoning processes on the correctness of requirement specifications. However, these models cannot identify the interaction among cognitive variables in determining the correctness of requirement specifications. In addition, problem domains have not been identified as a variable in these models.

Third, mental models have also been used to identify important determinants for the correctness of requirement specifications. A mental model is a collection of interconnected autonomous objects (Williams, Hollan, & Stevens, 1983, p.133). According to Williams, Hollan, and Stevens, autonomous objects are mental objects that have their own internal rules to guide their behaviors. The interactions among autonomous objects achieve the task of human reasoning. Some mental models for information requirement analysis can be found in the research papers of Adelson and Soloway (1985); Guindon and Curtis (1988); Guindon, Krasner, and Curtis (1990); and Vessey and Conger (1993). On the basis of the mental models, problem statements, domain knowledge of information analysts, and methodology knowledge of information analysts are identified as three sources of knowledge for specifying information requirements. However, those models provide no theoretical basis for explaining the interactive relationships between problem domains and requirement analysis techniques.

Finally, some researchers view information requirement analysis as a comprehension process. They believe that information requirement analysis is a process of translating the knowledge of problem statements into that of requirement analysis techniques. Therefore, good requirement analysis techniques should be able to make the translation process easy for information analysts. For example, Kim and Lerch (1992) focus the required skills for the translation of information requirements. They believe that object-oriented techniques are better than functional-oriented techniques because the required skill for object oriented techniques (symbolic simulation) is easier than that for functional-oriented techniques (test case). Batra and Sein (1994), on the other hand, suggest that the constructs used by requirement analysis techniques should be close to those of problem domains. By modeling information requirement analysis as a comprehension process, the above two models provide some insights about the required features for good requirement analysis techniques. However, they do not explain the interactive relationships between requirement analysis techniques and problem domains. In addition, the knowledge of information analysts is not included in those models.

In sum, current cognitive models have focused on different aspects of information requirement analysis like knowledge of information analysts, requirement analysis techniques, or modeling behaviors of information analysts. However, due to the lack of an integrated view of information requirement analysis, it is unknown what the interactive relationships between the different cognitive variables are. Even worse, some research studies may reach conflicting or contradictory conclusions because of the negligence of confounding variables.

A STRUCTURE BUILDING MODEL FOR INFORMATION REQUIREMENT ANALYSIS

In this section, I will propose a more adequate cognitive model of information requirement analysis that meets two requirements: first, it explicates the interactive relationships among the three determinants for the correctness of requirement specifications; and second, it provides an explanation for the two unsolved issues related to requirement analysis techniques. In the rest of this section I will first discuss the rationale of using text comprehension as a basis for the proposed cognitive model. Then I will discuss the details of the proposed cognitive model.

Information requirement analysis has been recognized as a process of understanding the domains of information systems and then specifying the information requirements for the information systems (Borgida, Greenspan, & Mylopoulos, 1985; Fraser, Kumar, & Vaishnavi, 1991; Huang & Burns, 1997; Yadav, 1983). Empirical studies on the analytical behavior of information analysts also showed a strong association among the activities of gathering information, identifying relevant facts, and conceptual modeling (Batra & Davis, 1992; Sutcliffe & Maiden, 1992). This strong association reflects that information requirement analysis is basically an understanding process.

From the perspective of human cognition, understanding is a process of building a coherent mental representation of the information being comprehended (Gernsbacher, 1990; Graesser, 1995; Kintsch, 1988; Ortony, 1978). On the basis of the structure building model of language comprehension (Gernsbacher, 1990), this article proposes a more adequate cognitive model of information requirement analysis that can explicate the interactive relationships among the knowledge of information analysts, requirement analysis techniques, and problem domains as shown in Figure 1. In addition, the cognitive model can provide a cognitive explanation for the two unsolved issues about requirement analysis techniques.

[FIGURE 1 OMITTED]

According to the structure building model of information requirement analysis, there are three important features that explain the modeling behavior of information analysts. First, coherence is the goal of information analysts in specifying information requirements. According to Collins Cobuild English Dictionary (1995), coherence is "a state or situation in which all the parts or ideas fit together well so that they form a united whole." While information analysts analyze problem domains on the basis of a particular requirement analysis technique, they may find that the problem statements can not be fitted into the conceptual model very well. Information analysts then perform intensive coherence inferences to resolve the discrepancies. Although correctness may be the goal for information requirement analysis, information analysts believe that understanding is achieved only if the requirement specifications are coherent.

Second, the coherence of a requirement specification can be achieved only when the cognitive capability of the information analyst is big enough to fill the cognitive gap between the problem domain and the requirement analysis technique being used. The cognitive capability of an information analyst can be evaluated by the domain knowledge and modeling knowledge owned by the information analyst. It is well recognized that richer domain knowledge and modeling knowledge are the qualities of expert information analysts for better performance in specifying information requirements. On the other hand, the cognitive gap between a problem domain and a requirement analysis technique can be evaluated by the difference of ontologies used by the problem domain and the requirement analysis technique. An ontology is a conceptual system (Guarino & Giaretta, 1995; Regoczei & Plantinga, 1987; Wand, Monarchi, Parsons, & Woo, 1995) that includes two parts: (1) a set of concepts for describing a problem domain such as entities, relationships, data flows, and agents; and (2) a cognitive structure for organizing the concepts (Marca & McGowan, 1993; Pepper, 1942) such as functional orientation, object orientation, data orientation, and control orientation. The difference in ontologies determines how difficult it is for information analysts to model a problem domain by a particular requirement analysis technique. Requirement analysis techniques act like schemata in text comprehension (Rumelhard, 1980). A schema governs a body of inferences which can help information analysts understand the conceptual structures of problem domains easily (Abelson, 1981). However, if the concepts and cognitive structures between problem domains and requirement analysis techniques are very different, information analysts may abandon the requirement analysis techniques, or even resulting in a failure of comprehension.

Third, understanding is a structure-building process that translates the ontologies of problem domains into those of requirement analysis techniques. Three operations are important for the translation process: mapping, shifting, and integrating. Mapping occurs while an information analyst studies a substructure of a problem domain. The information analyst first selects a requirement analysis technique that is best suited to the substructure of the problem statements. On the basis of modeling knowledge and domain knowledge, the information analyst then maps the substructure of the problem dsomain into the conceptual model of the selected requirement analysis technique. Shifting occurs while an information analyst decides that the selected technique cannot achieve a locally coherent conceptual model for a substructure of a problem domain. In this situation, the information analyst will try other requirement analysis techniques to build a conceptual model with local coherence for the substructure of the problem domain. Finally, integrating is the operation that integrates the conceptual models for the substructures of a problem domain into a conceptual model for the whole problem domain with global coherence.

A COGNITIVE EXPLANATION OF THE TWO UNSOLVED ISSUES ABOUT REQUIREMENT ANALYSIS TECHNIQUES

The cognitive gaps and cognitive capabilities in the proposed cognitive model can explain the two unsolved issues related to requirement analysis techniques: First, why is there no requirement analysis technique can consistently outperform the others? On the basis of the structure building model of information requirement analysis, the best requirement analysis technique should be the one that can not only minimize the cognitive gap by matching the ontology of the requirement analysis technique with the ontology of the particular problem domain, but also maximize the cognitive capabilities by matching the ontology of the requirement analysis technique and the ontology of the information analyst's knowledge. Therefore, the performance of a requirement analysis technique depends on its interactions with the type of problem domain under investigation, and the knowledge of information analysts. As a result, there is no single requirement analysis technique that can outperform other techniques without considering the types of problem domains and the knowledge of information analysts.

Second, why do experienced information analysts use multiple requirement analysis techniques in analyzing complex information systems? This is to say that experienced information analysts are eclectic in terms of their choice of techniques. This phenomenon reflects that in order to achieve high-level coherence of requirement specifications, experts select a particular requirement analysis technique not only based on their experience and knowledge (cognitive capabilities), but also the match between requirement analysis techniques and types of problems domains (cognitive gaps). As a result, experts use different requirement analysis techniques to fit different aspects of a problem domain and to fit their knowledge about the requirement analysis technique and the problem domain.

A THEORETICAL VALIDATION FOR THE STRUCTURE BUILDING MODEL

The proposition of the structure building model of information requirement analysis is based on the findings of empirical research in text comprehension and information requirement analysis. This section will discuss the previous research findings that support the concepts of coherence, cognitive gap and structure building process in the structure building model of information requirement analysis.

First, coherence has been recognized as a measurement for concept comprehension (Murphy & Medin, 1985). Komatsu (1992) suggested that an adequate theory about concept comprehension should be able to explain how coherence is achieved. Gernsbacher (1990) identified four sources of coherence: referential coherence, temporal coherence, location coherence, and causal coherence. He found that sentences with the above four types of coherence were easier to understand. In investigating the inference behavior of readers in text comprehension, Mckoon and Ratcliff (1992) found that inferences for local coherence were the basic inferences that were automatically performed by readers. Zwaan, Graesser, and Magliano (1995) investigated the effects of temporal, spatial, and causal discontinuities on sentence reading times in naturalistic story comprehension. They found that readers simultaneously monitored multiple dimensions of text coherence under a normal reading instruction. In addition, readers' goals would direct readers to allocate cognitive resources to specific dimensions of coherence.

Second, the cognitive gap between a problem domain and a requirement analysis technique is also recognized as a determinant for comprehension. The research in education found that the concepts in science were difficult to learn because of the ontology incompatibility between students' cognition and science (Chi, Scotta, & de Leeuw, 1994). It was believed that students' concepts of physics were mostly matter-based. However, the concepts in the theories of physics were mostly process-based. In an experiment to test users' ability to validate information requirements, Nosek and Ahrens (1986) found that compared to data flow diagrams, task oriented menus, which were believed to be closer to users' cognitive models, were better understood by users. Batra, Hoffer, and Bostrom (1990) compared the performance of novice analysts in data modeling by relational data models and entity relationship models. They found that entity relationship models, which were believed to be closer to analysts' cognitive models, scored higher in correctness.

Finally, empirical evidence also supports the structure building process of information requirement analysis. Model-based reasoning has been regarded as an important factor for the effectiveness of information requirement analysis (Sutcliffe & Maiden, 1990). Littman (1989) found that expert software designers used multiple mental representations to model a problem domain. In addition, expert software designers engaged in the process of selecting requirement analysis techniques to best fit the problem domain. Wijers and Heijes (1990) found that although expert information analysts had their own preference on requirement analysis techniques, they all had their own strategies to select different techniques in information requirement analysis. Finally, Gentner (1983) suggested that structure mapping was the way that people use to associate domain knowledge to problem domains for comprehension.

IMPLICATIONS OF THIS RESEARCH

This research has developed a cognitive model of information requirement analysis on the basis of the structure building model of language comprehension. This cognitive model contributes to the research in information requirement analysis in four aspects as follows.

First, this cognitive model provides a more adequate theory of information requirement analysis. In addition to explicating the interactive relationships among the knowledge of information analysts, requirement analysis techniques, and problem domains, the cognitive model uses cognitive gaps and cognitive capabilities to explain two important phenomena in information requirement analysis: (1) no requirement analysis technique can consistently outperform the others; and (2) experienced information analysts use multiple requirement analysis techniques in analyzing complex information systems.

Second, this cognitive model provides a basis for empirical validation. On the basis of this cognitive model, empirical research can be conducted to test the influence of the knowledge of information analysts, requirement analysis techniques, problem domains, and the interactive relationships among the three determinants on the correctness of requirement specifications.

Third, the cognitive model provides a framework to view and to compare important aspects of information requirements. According to the cognitive model, the knowledge of information analysts, requirement analysis techniques, and problem domains are interactive in determining the correctness of requirement specifications. Therefore, effective research studies in the cognitive processes of information requirement analysis should consider interactive relationships among these three determinants.

Finally, this cognitive model provides a theoretical basis for the development of computer-aided software engineering (CASE) tools. According to this cognitive model, how to reduce cognitive gaps and how to enhance cognitive capabilities are two important issues for the research in CASE tools.

CONCLUSION

Current research into the process of information requirement analysis has focused on answering the question of what factors influence the performance of information analysts in specifying information requirements. However, this research tries to answer why the knowledge of information analysts, requirement analysis techniques, and problem domains can influence the coherence of information requirement specifications.

On the basis of the theories of text comprehension, coherence is assumed to be the goal of information analysts in specifying information requirements. Cognitive gaps are proposed to measure the fitness between requirement analysis techniques and problem domains. Cognitive capabilities are used to reflect the knowledge of information analysts for transforming the ontologies. On the basis of coherence, cognitive gaps, and cognitive capabilities, a cognitive model of information requirement analysis is then proposed to explicate the interactive relationships among knowledge of information analysts, requirement analysis techniques, and problem domains. In addition, the cognitive model also provides an explanation for the two issues related to the performance of requirement analysis techniques.

In order to reduce the cognitive gaps between problem domains and requirement analysis techniques, three directions for future research in requirement analysis techniques can be identified: First, the ontologies of requirement analysis techniques should be close to those of problem domains. Therefore, domain-specific requirement analysis techniques deserve the attention of future research. Second, requirement analysis techniques should be able to cover and integrate more than one cognitive structure to reduce the need to shift among requirement analysis techniques with different cognitive structures. Third, in order to avoid the cognitive gaps, requirement specification reuse should be a major concern for the research in information requirement analysis.

REFERENCES

Abelson, R. P. (1981). Psychological status of the script concept. American Psychologist, 36 (7), 715-729.

Adelson, B., & Soloway, E. (1985). The role of domain experience in software design. IEEE Transactions On Software Engineering, se-11 (11), 1351-1360.

Allwood, C. M. (1986). Novices on the computer: a review of the literature. International Journal of Man-Machine Studies, 25, 633-658.

Batra, D., & Davis, J. G. (1992). Conceptual data modeling in database design: Similarities and differences between expert and novice designers. International Journal of Man-Machine Studies, 37, 83-101.

Batra, D., Hoffer, J. A., & Bostrom, R. P. (1990). Comparing representations with relational and EER models. Communications of the ACM, 33 (2), 126-139.

Batra, D., & Sein, M. K. (1994). Improving conceptual database design through feedback. International Journal of Human-Computer Studies, 40, 653-676.

Borgida, A., Greenspan, S., & Mylopoulos, J. (1985, April) Knowledge representation as the basis for requirements specifications. IEEE Computer, 82-90.

Chi, M.T. H., Slotta, J. D., & de Leeuw, N. (1994). From things to process: A theory of conceptual change for learning science concepts. Learning and Instruction, 4, 27-43.

Collins Cobuild English Dictionary (1995). Harper Collins Publishers.

Dardenne, A., van Lamsweerde, A., & Fickas, S. (1993). Goal-directed requirements acquisition. Science of Computer Programming, 20, 3-50.

Davis, A. M. (1988). A comparison of techniques for the specification of external system behavior. Communications of the ACM, 31(9), 1098-1115.

Dorfman, M. (1990). System and software requirements engineering. In Thayer, R. H., & Dorfman, M. (eds.). System and Software Requirements Engineering, IEEE Computer Society Press. 4-16.

Firesmith, D. (1991). Structured analysis and object-oriented development are not compatible. Ada Letters, XI (9), 5666.

Fitzgerald, B. (1996). Formalized systems development methodologies: A critical perspective. Information System Journal, 6, 3-23.

Fraser, M. D., Kumar, K., & Vaishnavi, V. K. (1991). Informal and formal requirements specification languages: Bridging the gap. IEEE Transaction On Software Engineering, 17(5), 454-465.

Gentner, D. (1983). Structure-mapping: A theoretical framework for analogy. Cognitive Science, 7, 155-170.

Gernsbacher, M. A. (1990). Language Comprehension As Structure Building. Hillsdale, New Jersey: Lawrence Erlbaum Associates, Publishers.

Graesser, A. C. (1995). Inference generation and the construction of situation models. In Weaver, C. A., Mannes, S., & Hetcher , C. R. (eds.). Discourse Comprehension. Hillsdale, New Jersey: Lawrence Erbaum Associates, Publishers. 117-139.

Greenspan, S. J., & Mylopoulos, J. (1982). Capturing more world knowledge in the requirements specification. Proceedings of International Conference in Software Engineering, IEEE Computer Society Press 225-234.

Guarino, N., & Giaretta, P. (1995). Ontologies and knowledge bases towards a terminological clarification. Proceedings of the Second Conference on Building and Sharing of Very Large-scale Knowledge Base, 25-32.

Guindon, R., & Curtis, B. (1988). Control of cognitive process during software design: what tools are needed? Proceedings of CHI'88. 263-268.

Guindon, R., Krasner, H., & Curtis, B. (1987). Cognitive process in software design: Activities in early, upstream design. in Bullinger, H. J., & Shackel, B. (eds). Human-computer Interaction-INTERACT'97. North-Holland, 383-388.

Hsia, P., Davis, A., & Kung, D. C. (1993). Status Report: Requirements Engineering, IEEE Software, 75-79.

Huang, I. L., & Burns, J. R. (1997). Organizational information requirement topology as the basis for organizational information requirement specification reuse. Proceedings of AIS Americas Conference on Information systems, 36-37.

Iivari, J. (1989). Levels of abstraction as a conceptual framework for an information system. in Falkenberg, E. D., & Lindgreen, P. (eds.). Information System Concepts: An In-depth Analysis. North-Holland Co., 323-349.

Jackson, M. (1994, November). Problems, methods and specialization. IEEE Software, 57-62.

Kim J., & Lerch, F. J. (1992). Towards a model of cognitive process in logical design: Comparing object-oriented and traditional functional decomposition software methodologies. Proceedings of CHI'92, 489-498

Kintsch, W. The role of knowledge in discourse comprehension: A construction-integration model. Psychological Review, 95(2), 163-182.

Komatsu, L. K. (1992) Recent views of conceptual structure. Psychological Bulletin, 112(3), 500-526.

Koubek, R. J., Salvendy, G., Dunsmor, H. E., & Lebold, W. K. (1989). Cognitive issues in the process of software development: review and reappraisal. International Journal of Man-Machine Studies, 30, 171-191.

Littman, D. (1989). Constructing expert systems as building mental models. Knowledge Representation and Organization in Machine Learning. West German: Springer-Verlag, 88-106.

Loy, P. H. (1990). A comparison of object-oriented and structured development methods. Software Engineering Notes, 15(1), 44-48.

Malhotra, A., Thomas, J. C., Carroll, J. M., & Miller, L. A. (1980). Cognitive processes in design. International Journal of Man-Machine Studies, 12, 119-140.

Marca, D. A., & McGowan, C. L. (1993). Specification approaches express different world hypotheses. Proceedings of the Seventh International Workshop on Software Specification and Design. IEEE Computer Society Press, 214-223.

McKoon, G., & Ratcliff, R. (1992). Inference during reading. Psychological Review, 99(3), 440-466.

Murphy, G. L. & Medin, D. L. (1985, July). The role of theories in conceptual coherence. Psychological Review, 92(3), 289-316.

Nosek, J. T., & Ahrens, J. D. (1986). An experiment to test user validation of requirements: Data-flow diagrams vs task-oriented menus. International Journal of Man-Machine Studies, 25, 675-684.

Opdahl, A. L., & Sindre, G. (1995). Facet models for problem analysis. Advanced Information System Engineering, 54-67.

Ortony, A. (1978). Remembering, understanding, and representation. Cognitive Science, 2, 53-69.

Pepper, S. C. (1942). World Hypotheses: A Study in Evidence. Berkeley, CA: University Of California Press.

Regoczei, S., & Plantinga, E. P. O. (1987). Creating the domain of discourse: Ontology and inventory. International Journal of Man-Machine Studies, 27, 235-250.

Roman, G. (1985, April). A taxonomy of current issues in requirements engineering. IEEE Computer, 14-22.

Rumelhart, D. E. (1980). Schemata: The building blocks of cognition. In Spiro, R. Bruce, B., & Brewer, W. (eds.). Theoretical Issues in Reading Comprehension. Hillsdale, New Jersey: Lawrence Erlbaum Associates.

Scharer, L. (1981, April). Pinpointing requirements. Datamation, 17-22.

Schenk, K.D., Vitalari, N.P., & Davis, K.S. (1998). Differences between novice and expert system analysts: What do we know and what do we do? Journal of Management Information Systems, 15(1), 9-50.

Shemer, I. (1987). System analysis: A systematic analysis of a conceptual model. Communications of the ACM, 30(6), 506-512.

Shumate, K. (1991). Structured analysis and object-oriented design are compatible. Ada Letters, XI(4), 78-90.

Sowa, J. F., & Zachman, J. A. (1992). Extending and formalizing the framework for information systems architecture. IBM Systems Journal, 31(3), 590-616.

Standish Group (1995). Software Chaos, retrieved November 20, 2004 from http://www.standishgroup.com/chaos.html Sutcliffe, A., & Maiden, N. (1990). Cognitive studies in software engineering. Proceedings of the Fifth Conference on Cognitive Ergonomics, European Association of Cognitive Ergonomics, ECCE-5, 159-171.

Sutcliffe, A., & Maiden, N. (1992). Analysing the novice analyst: cognitive model in software engineering. International Journal of Man-Machine Studies, 36, 719-740.

Topi, H., & Ramesh, V. (2002). Human factors research on data modeling: A review of prior research, an extended framework and future research directions, Journal of Database Management, 13(2), 3-19.

Vessey, I., & Conger, S. (1993). Learning to specify information requirements: the relationship between application and methodology. Journal of Management Information Systems, 10 (2), 177-201.

Vessey, I. & Conger, S. A. (1994). Requirement specification: Learning object, process, and data methodologies. Communications of the ACM, 37(5), 102-113.

Vessey, I., & Glass, R. L. (Fall 1994). Application-based methodologies: Development by application domain. Information System Management, 53-57.

Vitalari, N. P., & Dickson, G. W. (1983). Problem solving for effective system analysis: An experimental exploration. Communications of the ACM, 26(11), 948-956.

Wand, Y., Monarchi, D.E., Parsons, J., Woo, C.C. (1995). Theoretical foundations for conceptual modeling in information system development, Decision Support Systems 15, 285-304.

Wijers, G. M., & Heijes, H. (1990). Automated support of the modeling process: A view based on experiments with expert information engineers. Lecture Notes in Computer Science, Spring-Verlag. 88-106.

Williams, M. D., Hollan, J. D., & Stevens, A. L. (1983). Human reasoning about a simple physical system. in Gentner, D., & Stevens, A. L. (eds). Mental Models. Lawrence Erlbaum Associates, Publishers, Hillsdale, New Jersey. 131-153.

Yadav, S. B. (1983). Determining an organization's information requirements: A state of the art survey. DATA BASE, 14(3), 3-20.

Yadav, S. B., Bravoco, R. R., Chatfield, A. T., & Rajkumar, T. M. (1988) Comparison of analysis techniques for information requirement determination. Communication of the ACM, 31(9), 1090-1097.

Zwaan, R. A., Graesser, A. C., & Magliano, J. P. (1995). Dimensions of situation model construction in narrative comprehension. Journal of Experimental Psychology, 21(2), 386-397.

I-Lin Huang, Langston University