Remote Evaluation of the Coherence of Indirect Manipulation Interface Systems For Agent-Mediated Legacy Data

摘要

Many information systems depend heavily on distributed legacy data sources. These data sources introduce a number of significant problems, especially when the sources must be combined and displayed to remote users. Many researchers have investigated various interface systems, however empirical studies have not been published that examine remote interfaces to distributed heterogeneous data. The purpose of this research is to determine the efficacy of a system that provides a more coherent representation of this distributed data in comparison to a more traditional system for users performing representative tasks. This dissertation presents the results of remote usability experiments in a specific, well-defined context. These web-based experiments empirically determine whether coherence is enhanced through application of the proposed methodology by presenting each interface system and a sequence of representative tasks. The remote evaluation system measures coherence based upon the subject's time to complete each task, the correctness of their answer, and their subjective confidence in that answer. When all tasks have been completed, the respondents complete a usability survey to express their satisfaction with the interface system. The specific research undertaking is to determine whether a system based upon this proposed methodology, the Visual Interface To Agent Mediated Information Networks (VITAMIN) system, is superior to a system based upon a traditional approach, the Java Indirect Manipulation Interface (JIMI) system. VITAMIN was developed to add coherence to the legacy data and JIMI was developed as a control treatment to represent a traditional legacy approach. Both VITAMIN and JIMI were implemented as indirect manipulation interface systems, or non-anthropomorphic interface agents. The interface systems provide intermediate query predicate acsystem in the IMPACT agent architecture thThe experiments used a within-subject randomly.