CLEAR requirements: Improving validity using cognitive linguistic elicitation and representation.

摘要

Effectively communicating requirements is necessary to the success of every software system, especially those that are safety-critical or otherwise high-consequence. However, the validity of these systems is compromised by human cognitive habits and limits. Certain habits and limits interfere with the creation and transfer of knowledge necessary to produce such systems, making valid requirements difficult to achieve. This work presents and evaluates a methodology for improving communicative fidelity in requirements environments.; Cognitive Linguistics provides results that can be used to both explain undesired phenomena observed in requirements environments and direct how they might be avoided. In everyday life, necessary cognitive states arise spontaneously. In environments that require the accuracy and precision necessary for the correct implementation of high-consequence systems, and where communication is often across domain boundaries, they generally do not.; Cognitive Linguistic Elicitation and Representation (CLEAR) proposes that these necessary states can be brought into existence in environments where their spontaneous appearance is not the common case. CLEAR accommodates natural human communicative habits and limits by defining, and guiding the construction of, explicit processes and artifacts that allow necessary cognitive states to be achieved.; CLEAR was evaluated using a two-part approach designed to achieve both experimental control and environmental realism. A formal experiment first demonstrated the value of central CLEAR activities and products with statistical significance. A larger-scale case study of its application to an safety-critical medical device in development then both replicated earlier results as well as extended them.; CLEAR makes three main contributions to the study and practice of requirements engineering. First, it provides a linguistically-grounded analysis of communication deficiencies and their sources. Second, it provides a mechanism for systematic detection and correction of several classes of such deficiency. Finally, analysis of the products resulting from CLEAR provides for the first known mechanism for quantitatively estimating the miscommunication risks of a project's critical notions, thereby offering guidance in the allocation of resources for the preservation of their integrity throughout a project's duration. Together, these contributions advance the state of the art in achieving high-integrity communication for requirements engineering.