Naturalistic Model of Causal Reasoning: Developing an Experiential User Guide (EUG) to Understand Fusion Algorithms and Simulation Models

Abstract

Abstract : For this effort a naturalistic study of causal reasoning was performed. One strand of this project examined causal reasoning in military and other contexts. A second strand designed a support concept to help people understand the causal reasoning of computer decision aids they use. For the first strand, we determined that most causal reasoning involve indeterminate situations. We demonstrated that people see a variety of causal reasoning formats in naturalistic settings, ranging from simple, single cause attributions to complex events, decisions, and forces. We further demonstrated that we could manipulate preferences for explanation types by varying features of the target audience for an explanation. We also found cultural differences in preferences for simple vs. complex causal reasoning formats. For the second strand, we developed the concept of an Experiential User Guide (EUG), to enable users of sophisticated decision aids to better understand the causal reasoning embedded in the algorithms. We demonstrated this EUG concept with JCAT (Java Causal Analysis Toolkit), a Bayesian reasoning support tool, and formulated recommendations for similar EUG applications with other types of decision aids. Finally, we formulated recommendations for helping people become more effective in causal reasoning within naturalistic settings.