No abstract

No abstract

The Generator of INterface Agents (GINA) is a software workbench that supports the development of interface agents. Viewing agent development as a cognitive engineering process, GINA supports development of agents from cognitive models of the human whose functions are being automated andlor supported by the agent. Built around the COGNET cognitive task analysis language, GINA provides software tools to support building and editing cognitive models, translating the models into executable forms that provide the reasoning kernel of an intelligent agent, building a software shell that allows this kernel to communicate with the targeted application environment, and testing and debugging the resulting agent.

PDQ and CANCERLIT are databases that represent potentially valuable resources for oncologists, but there are significant obstacles to their effective use in the clinical context. The clinician must learn where each databasecan be found, what it contains, and how to query it. More importantly, the clinician must also determine when and how each information resource is appropriate to the (clinical) task at hand, and integrate their general information with patient-specific data contained in patient records and clinical information systems. This research sought to remove these barriers to information use by developing a software agent in the human-computer interface that automates the information retrieval and integration process. The agent, developed from a cognitive model of how human experts perform these tasks, provides both a specific solution to the cancer information access problem, and a template for solving similar problems that occur with increasing frequency in many other domains.

Public reporting burden for this collection of infornation is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing this collection of informration. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden to Department of Defense, Washington Headquarters Services, Directorate for Information Operations and Reports (0704-0188), 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to any penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number. PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS. REPORT DATE (DD-MM-YYYY)2 SUPPLEMENTARY NOTES ABSTRACTThis report describes CHI Systems' efforts during Phase III and IV of the model comparison process of the Agent-based Modeling and Behavioral Representation (AMBR) program conducted by the Air Force Research Laboratory. Phases III and IV focused on modeling and simulating concept learning within the context of a simulated Air Traffic Control (ATC) work environment.CHI Systems extended the COGNET/iGEN framework used in prior AMBR phases, implementing a general capability to learn the conditions under which each of a disjunctive set of goals or actions should be taken. Deeper extensions to the system enabled the representation of memory decay, rehearsal, and proactive interference needed to model human learning performance.The Phase III and IV COGNET/iGEN model execution results were compared with the results of three other models (ACT-R, D-Cog, EPIC-Soar) and results collected from human trials.On most measures (accuracy, response time, workload) iGEN model results were indistinguishable from human performance and, overall, iGEN results provided a better fit to the human data than the other models tested.