Representing the Social Practices of Communities as the Basis for Agent-based Simulation


   This paper presents a new approach, the Human Community Model (HCM), to simulating complex human systems.  Rather than treat human actions in a manner analogous to biological processes, such as the predator-prey model, we consider all human behavior to be a case of an actor engaging in a practice of a community, and use formal articulations of the concepts of community and practice.  The rigorous articulations allow us to represent, in machine-usable form, a variety of human action in human communities, without doing violence to the fundamental realities of human choice and values.  The paper explains the model and how it differs from other approaches, and shows how it can be used in agent-based simulations of human communities. 
    The practices of a community, which are the possible behaviors a member of the community can engage in, have a complex logical structure including:  a) Other, smaller, practices; b) Choices the actor can or must make in carrying out a practice; c) Choice Principles that govern the choices a member makes; and d) Individual preferences. Each practice has versions, allowable sets of steps for carrying out the practice. (“Allowable” not in the physical sense but in the sense of being known to community members as ways of doing that practice.) An actor engages in a behavior by carrying out a version of a practice.  Each step is itself a practice, and is carried out by engaging in one of its versions, down to the level of detail at which there are no behaviorally significant variations (behaviorally significant in that community).  The practices of a community thus form a hierarchically organized set of logically complex units representing what can be done, incorporating actions, choices, community values, and individual values.  The formal model we employ allows us to represent these structures, and build software based on them.
    Carrying out a practice is rarely, if ever, analogous to a machine carrying out an algorithm; a key aspect of human practices is that, in the paradigm case, there are several kinds of choices to be made: which practice to carry out; which of the steps, or options for a step, to do; and which actual individuals, objects, or materials to use for the logical roles in the steps of the version of the practice.
    This model has been used to build a number of software systems in different communities of practice, including several question-answering systems (employing from 50 to 150 practice descriptions) and a system to actually carry out practices in a large multi-office bank.  A number of the applications were to practices not usually considered to be formally describable, such as the practice of a manager attempting to improve a subordinate’s performance by appealing the subordinate’s intrinsic motivations. 
    Since carrying out an action is a case of carrying out a set of steps, and each of the steps is itself done by carrying out a set of sub-steps, it is easy to see how HCM-based simulation can proceed:  Each actor carries out Step 1 of the version of the practice he/she has chosen; completion of that Stage is some outcome (by definition), so the facts known to the actor are now changed; Each actor now re-assesses the situation (the known facts) and again chooses what to do, based on the new facts, which include (the fact that) the actor has engaged part of the version of the practice and therefore has behavioral reason to continue it.  Theories about an actor, a kind of actor, or even whole societies, can be tested by representing them as priorities for making choices among behaviors, and observing the results of the choices.
    The complex logical structure of each practice, combined with the explicit representation of both the “values” or “preferences” related to the practice and the precise places in the practice where the choices reflect the values, provides a rich set of resources for modeling the varieties of human action.  Since the formulation of community has been deliberately created to cover a wide range of organized human groups, from work teams and businesses to societies and whole cultures, the applicability seems widespread.
    In the paper, we present the full formal structure of Communities and Practices, and give examples of descriptions of practices. In addition, we discuss limitations of the current work with the model, and several planned applications.



Joe Jeffrey
Northern Illinois University
Department of Computer Science
jeffrey@cs.niu.edu