Trust and Reliance in Human Socio-Cultural Behavior Models

Farry, M., Pfautz, J., Carlson, E., and Koelle, D.

Proceedings of 3rd International Conference on Applied Human Factors and Ergonomics, Miami, FL (2010)

Human socio-cultural behavior (HSCB) modeling technologies are gaining traction in operational communities as a means to analyze and predict human behavior at various scales. The National Research Council’s study on behavior modeling and simulation has identified salient challenges to the acceptance of HSCB modeling technologies, including the lack of interoperability between models, the need for support of data collection, standards for verification and validation, and the establishment of usefulness to users. Other efforts have identified specific factors that lead to distrust of models and their results (e.g., model authorship, the scientific basis behind a model), and suggest the presentation of taxonomic factors to support model selection and build trust (e.g., authorship history, catalogs of the successful application of models). While these efforts directly address operational issues, we can achieve a more rigorous understanding of these complex design issues by studying the mature body of work in trust in automation.

In this paper, we review the leading principles of automation design to inform HSCB modeling efforts, extending and modifying principles to fit the unique challenges faced in the HSCB modeling domain. Most critically, classic design principles for automation seek to support a linear process of information acquisition, information analysis, decision and action selection, and action implementation. We must modify this framework for HSCB modeling operational use cases, where automation is usually only brought to bear to analyze information and support decision making. Furthermore, planning tasks that exploit HSCB modeling are frequently iterative, and constantly reviewed over long time scales as they are applied to a dynamic environment. Therefore, HSCB design principles must support these iterative processes rather than the traditional linear decision-action processes. We review relevant extensions to traditional principles for automation design as they relate to the problem of HSCB modeling.

Building on well-established principles of trust in automated systems and experience developing operational HSCB modeling systems, we present definitions of the concepts of trust and reliance specific to HSCB models, and discuss the implications of those definitions. We expand upon traditional perspectives on the universal bases of trust (e.g., performance, process, and purpose) to provide a perspective unique and beneficial to the development of HSCB modeling systems. We characterize common tasks that exploit HSCB modeling and bolster the set of human performance factors commonly considered in automation design (e.g., mental workload, situational awareness, complacency, skill degradation) with factors that are more task-specific and detailed. Based on those performance factors, we provide a set of design guidelines to engender appropriate trust and maximize the effectiveness of HSCB modeling systems.

For More Information

To learn more or request a copy of a paper (if available), contact Farry, M.

(Please include your name, address, organization, and the paper reference. Requests without this information will not be honored.)