Variation in Human-Intensive Systems: a Conceptual Framework for Characterizing, Modeling, and Analyzing Families of Systems

03 Mar
Tuesday, 03/03/2015 7:30am to 9:30am
Ph.D. Thesis Defense

Borislava Simidchieva

Computer Science Building, Room 151

A system model---namely a formal definition of the coordination of people and software components performing activities, using resources and artifacts, and producing various outputs---can aid understanding of the real-world system it models. Complex real-world systems, however, exhibit considerable amounts of variation that can be difficult or impossible to represent within a single model. This dissertation evaluates the hypothesis that the careful characterization and representation of system variation can aid in the generation and analysis of concrete system instances related to one another in specified ways and manifesting different kinds of variation. When a set of closely related systems can be characterized by a compelling set-membership criterion, it is often useful and appropriate to characterize the set as a family of systems. In this dissertation, a variety of needs for system variation and for family criteria are identified. We focus on two specific kinds of variation, namely functional and agent variation, and suggest an approach for meeting these needs both at the level of requirements specification (problem-level variation), as well as at the level of implementation specification (solution-level variation). 

We present a framework for generating and analyzing new system instances, using the Little-JIL process definition language as an experimental vehicle to study what process definition language capabilities are necessary to support the explicit modeling of variation at the solution level, and thereby to address needs at the problem level. We define a formal notation for specifying several different scenarios of functional and agent variation in human-intensive processes and describe a prototype system to accommodate this specification within an existing modeling framework. Once a family of systems is formally defined and characterized at the solution level, different analysis techniques can be applied to make assurances that all members of the family share certain kinds of properties. These analysis results can then be used to inform variation needs at the problem level. To evaluate the applicability of the approach, we study and model the variation observed in two real-world, human-intensive systems from the domains of conflict resolution and elections. Both case study domains have been observed to employ functional variants of their processes, and, given their complex coordination of human and software agents, both domains require agent variation, therefore fostering a fruitful application of our approach.

Advisor: Leon Osterweil