COHSI Consortium Meeting

February 12, 2009

The Committee on Human-Systems Integration is in the process of organizing a consortium of agencies with interest in HSI. At this meeting Committee members gave two presentations on topics of interest identified by Consortium members at a previous meeting.

Human-Mediated Anomaly Detection Systems

View the PowerPoint presentation used by Tom Sanquist

Anomaly detection refers to finding patterns in data that deviate from expected behavior, and is a fundamental element in safety and security systems, as well as diagnostics in a variety of fields (medicine, equipment maintenance). This session will focus on general issues associated with anomaly detection across a range of applications in which human operators serve as monitors and adjudicators of a system output, such as an intrusion, threat, or biological state detection (e.g., fatigue). These systems involve varying levels of automation, from those that require human sensory discrimination, to systems in which the human role is to determine if sub-sensory events that are categorized as anomalous by system sensors and decision criteria, are indeed worthy of intervention. Numerous different sensor types and data processing approaches are used in anomaly detection systems, including radiation sensors, visual scene analysis, terrain feature analysis, etc. The session will review the general characteristics of anomaly detection systems and applications, discuss various quantitative approaches, and illustrate the common human factors questions across these applications. A key part of the session will be illustrations of specific application examples from on-going government system developments. Committee members will discuss how anomalies are of interest in particular sub-disciplines, such as social and organizational psychology, and neuro-ergonomics, and the potential value-added of these perspectives to system development. We will synthesize at a very high level (1) how what we know now can be applied to system development in the near-term, and (2) the longer-term fundamental research issues that could be addressed by programmatic study.

Cognitive Processes and Decision Making in Healthcare

Potential areas of interest

• Patient decision making: Risk assessment/risk translation

• Team-level cognitive processes in healthcare

• Telemedicine:  Supporting distributed work

• Medical error: Individual, team, organizational and technological factors

• The design of resilient systems in healthcare

• Medical decision making: Human-centered design of tools to support diagnosis

and treatment

• Healthcare information systems:  Tracking for early detection and prevention and for

and for continuous improvement of practices and procedures

View Nancy Cooke’s PowerPoint on Team-Level Cognitive Processes

View Phil Smith’s PowerPoint on Human-Centered Design of Decision Support Systems to Enhance Medical Decision Making

View Debbie Boehm-Davis’s PowerPoint on Analysis Techniques for Gaining Insights into Medical Error

This session emphasized the importance of taking a systems approach to human-centered design. In order to understand the factors that affect performance, it is necessary to consider not only the person performing a particular task, but to consider the broader group, organizational, social and physical environments, as well as the broader task environment. A wide range of analysis methods were discussed, ranging from the use of cognitive task analyses to the mapping of emergent patterns of interaction in a team. Human-centered approaches to design were also illustrated, including the use of tools to support better visualization and the design of knowledge-based systems to provide decision support. Results from detailed human factors studies of such design served to further illustrate the expertise available from the human factors community. Discussions with the full Committee and attending representatives from healthcare agencies and organizations further enriched the meeting, providing a variety of examples where this field of study can be applied.