Extensible software agents: Improving the development of extensible software.

摘要

Software systems must be more extensible to more quickly react to organizational and environmental changes. Extensible software agents can be used in applications to improve their development and their ability to handle subsequent changes. A high level of extensibility provides design-time and run-time flexibility to handle new and changed functionality. It is important for managers, users, and developers to understand the benefits, principles, and processes relating to extensible software agents. However, few extensible software agents have been designed and developed. There is a gap in controlled experimental research to demonstrate agents are easy to use and useful for development.; This research provides a design for an extensible agent, and studies whether a higher level of agent extensibility can be easy for developers to use, and can help them perform their job duties to develop applications and handle changes. Three hypotheses were experimentally tested based upon the Technology Acceptance Model, the Theory of Plausible Designs, and the Architecture Tradeoff Analysis Method (ATAM) recently developed for architecture evaluation. The hypotheses stated that a higher level of agent extensibility would positively affect ease of use, usefulness, and the ability to handle changes. The methodology was a randomized groups repeated measures design using demonstrations and designs for two levels of agent extensibility. Participants were 52 VCU Information Systems students. Instrumentation was an adapted TAM instrument supplemented with reasons and comments, and responses to participant-requested changes for each agent. Statistical analysis and content analysis of responses provided very strong support for the hypotheses.; This research demonstrates a useful application of TAM for agent development, and useful additions to the TAM scales for content analysis. The 53 design patterns used for an extensible agent, including 15 J2EE design patterns, how they are combined, and several novel approaches are described. Recommendations for researchers and practitioners are provided to guide future research and work with extensible software agents.