On how constraints shape action

Abstract: Understanding howjoint systems of people and artifacts perform goal-directed actionin a dynamic environment is essential for improving and supportingwork. The literature on cognitive systems engineering and itsrelated disciplines identifies the concept of constraint as anessential element in the analysis of systems in dynamicenvironments. These disciplines however offer differing definitionsand ascribe different roles to constraints in (cognitive) systemsanalysis. This thesis distills from these perspectives onconstraints the hypothesis that constraints shape action. Itinvestigates empirically how constraints shape action in studies oftwo domains characterized by dynamic environments: driving andcommand and control.The first study investigates the effect of the constraint of pathwidth on steering behavior. It reviews research on the -lawof steering- - the quantitative relationship betweenhuman temporal performance and the spatial characteristics of themovement path. It then extends the study of the law of steering todriving in virtual environments, within the field of human-computerinteraction. Participants drove a virtual vehicle in a virtualenvironment on paths whose shape and width were systematicallymanipulated. Results showed that the law of steering indeed appliesto locomotion in virtual environments. On average both the mean andthe maximum speeds of the participants were linearly proportionalto path width. Such regularity in human performance provides aquantitative tool for (3D) human machine interface design andevaluation.The second study describes a method for the recognition ofconstraints in network-based command and control, and illustratesits application in an experimental study in a command and controlmicroworld. The method uses goals-means task analysis to extractthe essential variables that describe the behavior of a command andcontrol team. It juxtaposes these variables in state spacerepresentations illustrating constraints and regions foropportunities for action. A series of examples shows how statespaces plots of experimental data can aid in the description ofbehavior vis-à-vis spatial and temporal resource constraints, anddiscusses how state space representations may be used to improvecontrol in network-based command and controlsettings.