From Work Practice Models and Simulation To Implementation of Human-Centered Agent Systems Maarten Sierhuis Ron van Hoof William J. Clancey Mike Scott RIACS/USRA QSS NASA/Ames Research QSS NASA/Ames Research NASA/Ames Research Center NASA/Ames Research Center Center Computational Sciences Center Mail Stop 19-39 Mail Stop 269-3 Mail Stop 269-3 Mail Stop 269-3 Moffett Field, CA 90435 Moffett Field, CA 90435 Moffett Field, CA 90435 Moffett Field, CA 90435 USA USA USA USA (+01) 650-604-4917 (+01) 203-531-4741 (+01) 650-604-2526 (+01) (212) 941-1845 msierhuis@mail.arc.nasa.gov rvanhoof@mail.arc.nasa.gov bclancey@mail.arc.nasa.gov mscott@ptolemy.arc.nasa.gov is well defined. Sometimes however, it is more difficult to ABSTRACT describe the input and output of the work. For example, consider In this paper we describe an agent-based software development a soccer match between two professional soccer teams. It is environment for developing agent systems that are fundamentally difficult to define the input and output of this type of work, based on a holistic analysis of the human and software agent although most of us would agree that professional soccer players organization and work practices for which the agent system is to are working. To describe the work of a soccer team we quickly be developed. Brahms is an agent modeling, simulation and fall into descriptions about teamwork and collaboration on the development environment developed at NASA Ames Research field. Center. Brahms stems from a decade of research on modeling and We claim that the individual activities that make up the work simulating human work practices. As a result of this research, not only have to do with the transformation of input to output, but partly at NASA Ames and partly in industry, we are now working more importantly with the collaboration between individuals in on extending the environment to allow for the design and action, in pursuit of a goal. Imagine soccer players who implementation of software systems that are fundamentally based collaborate in their activities of kicking a soccer ball, in pursuit of on work practice and include software agents that have a dynamic scoring a goal. Just focusing on the in- and output of each representation of the human- and software agents it is individual activity of a soccer player would not only be very collaborating with. difficult, if not impossible, it would also miss the opportunity to Keywords understand what is really going on in this work. However, in the past century work has been defined as the transformation of input multi-agent, modeling, simulation, intelligent agents, work practice, activity to output, starting with Frederick W. Taylor’s view of work to Michael Hammer’s view of business processes [6]. 1. Modeling and simulating work practice We take a different view, and are interested in describing Work practice is a concept that originates in socio-technical work as a practice, a collection of psychologically and socially systems, business anthropology, work systems design, and situated collaborative activities between members of a group. We management science. try to understand how, when, where, and why collaborative The notion of “practice” is central to work systems design, activities are performed, and identify the effects of these which has its roots in the design of socio-technical systems, a activities, as well as to understand the reasons why these method developed in the 1950s by Eric Trist and Fred Emery [1]. activities occur in the way they do. Therefore, the central theme is Socio-technical systems design sought to analyze the relationship to find a representation for modeling work practice. Many of the social system and the technical system, such as researchers in the social sciences use the word practice as if it is a manufacturing machinery, and then design a “socio-technical well-defined concept that everyone understands. However, it is system” that leveraged the advantages of each. Work systems difficult to describe what a practice is. People notice when design extends this tradition by focusing on both the formal something is not a practice, and can often describe why. Although features of work (explicit, intentional) and the informal features it can be said that a group of people has developed a practice, of work (as it is actually carried out “in practice,” analyzed with when asked to describe what that practice is, and what it consists the use of ethnographic techniques) [2] [3] [4] [5, chapter 16]. of, we find it difficult to describe in words. As such, practice is A work practice is defined as the collective activities of a part of our tacit knowledge [7]. group of people who collaborate and communicate, while An ad hoc definition of the word practice is: The performing these activities synchronously or asynchronously. (collaborative) performance of situated activities in real life Most often, people view work merely as the process of situations, by making use of knowledge previously gained through transforming input to output, i.e. a Tayloristic view of work. For experience in performing similar activities. example, when building a house the input and output of the work 1 In short, practice is doing in action [8]. Scientists have are mostly represented at the group-level or class-level described how a practice develops, like Wenger, who defines the respectively, but can also be represented down at the agent and creation of a practice as follows [9]: object level. Activities at the group- and class-level are inherited Being alive as human beings means that we are constantly at the lower levels. Related to this, we describe the constraints of engaged in the pursuit of enterprises of all kinds, from ensuring when these activities can be performed in the Timing Model. Such our physical survival to seeking the most lofty pleasures. As we activity constraints are represented in the form of preconditions of define these enterprises and engage in their pursuit together, we situation-activity rules. We call such situation-activity rules interact with each other and with the world and we tune our workframes [16]. A workframe executes an agent's activity when relations with each other and with the world accordingly. In other its preconditions match against the agent's individual belief-set. words, we learn. Over time, this collective learning results in Because activities take time, a workframe instantiation has a practices, which reflect both the pursuit of our enterprises and the duration. However, activities, and thus workframes, can be attendant social relations. These practices are thus the property of interrupted and resumed, making the actual length of an activity a kind of community created over time by the sustained pursuit of performance situation dependent. a shared enterprise. Next, we can represent an agent's reasoning behavior as Everybody knows what Wenger means when he says, “this forward-chaining production-rules in the Knowledge Model. Such collective learning results in practices”, but what is it that results? production rules can also be represented at the group-level, and is Can it be described? Can it be modeled? To do this we need to be also inherited at lower levels. In Brahms, production-rules are able to describe practice at an epistemological level we call the called thoughtframes [16]. work practice level. In the rest of this paper, we will discuss a The last model we distinguish is the Communication Model. representational language to represent models of work practice. In this model we represent the agent and object communication. These models can be simulated in order to show the effects of the In Brahms we represent communication as speech-acts, i.e. activities of people and their communication, being situated in a situation-specific communication-actions of agents' beliefs geographical environment, and using tools and artifacts to to/from other agents or artifacts [17]. perform their collaborative work. 2. From simulation to agent systems Work practice includes those aspects of the work process We have recently reimplemented the Brahms environment in that make people behave a certain way in a specific situation, at a Java. As part of this effort we have made three enhancements, specific moment in time, in the real world. To describe people’s which will allow us to use the Brahms language as a full-fledged situation-specific behavior we need to include those aspects of the agent language for developing intelligent human-centered agent- situation that explain the influence on the activity behavior of based systems. The first enhancement is the creation of a real- individuals (in contrast with problem-solving behavior), such as time Brahms execution engine— the Brahms Virtual Machine people’s collaboration, “off-task” behaviors, multi-tasking, (VM). The VM is similar to the simulation engine, but does not interrupted and resumed activities, informal interaction, include time synchronization between agents and objects by a knowledge and geography [10] [11]. centralized scheduler. Each agent and object operates Brahms is a modeling and simulation environment for autonomously using its own discrete event engine. This allows representing work practice in a rule-based agent language, which each agent to execute as an independent Java-thread and run as can be simulated using the Brahms rule-based, multi-agent fast as possible, without having to be synchronized by the simulation engine. At NASA we have used Brahms to model and scheduler. simulate the work practices of the Apollo astronauts, as well as The second enhancement is a Brahms Java-activity type, the human-robot collaboration for a semi-autonomous robotic with which a modeler can implement any agent or object activity mission to a planetary surface [12] [13] [14]. in Java. This allows moving execution from a Brahms agent to To model a work practice we develop seven models, as Java. The third enhancement is that of a Java-API for the described by the World Modeling Framework [15]. First, we development of Brahms proxy-agents in Java. Brahms proxy- design the Agent Model in which we represent the group-agent agents allow moving execution from Java to Brahms. membership hierarchy of all the agents in the work system. The Agent Model describes to which groups the agents belong and 3. References how these groups are related to each other. After the Agent [1] F. E. Emery and E. L. Trist, “Socio-Technical Model, the next model that needs to be designed is the Object Systems,” in Management Sciences, Models and Model. In this model we design the class-hierarchy of all the Techniques, C. W. C. a. o. (eds.), Ed. London: domain objects and artifacts. Pergamon, 1960. Now that the agents, objects and real-world artifacts are [2] P. Ehn, Work-Oriented Design of Computer Artifacts. represented, the next model is the Geography Model in which the Stockholm, Sweden: Arbetslivcentrum, 1988. agents and artifacts are located during the simulation. In Brahms we model geographical locations using two concepts, area- [3] J. Greenbaum and M. Kyng, “Design at Work: definitions and areas [16]. Area-definitions are user-defined types Cooperative design of computer systems,” . Hillsdale, of areas. Areas are instances of area-definitions. An area is an NJ.: Lawrence Erlbaum, 1991. instance of a specific location in the real world that is being [4] W. A. Pasmore, Creating Strategic Change: Designing modeled. Furthermore, areas can be part-of other areas. With this the Fliexible, High Performing Organization. New representation scheme we can represent any location at any level York: Wiley, 1993. of detail. [5] M. R. Weisbord, Productive Workplaces: Organizing The fourth model is the Activity Model. In the Activity and Managing for Dignity, Meaning, and Community. Model we describe the behavior of agents and objects in terms of San Francisco, CA: Jossey-Bass Inc., Publishers, 1987. the activities they perform over time. Agent or object activities 2 [6] M. Hammer and J. Champy, Re-engineering the Corporation. New York, NY: Harper Collins Publishers, Inc., 1993. [7] M. Polanyi, The Tacit Dimension. Magnolia, MA: Peter Smith, 1983. [8] L. A. Suchman, Plans and Situated Action: The Problem of Human Machine Communication. Cambridge, MA: Cambridge University Press, 1987. [9] E. Wenger, Communities of Practice; Learning, meaning, and identity, (Draft) ed: Cambridge University Press, 1997. [10] M. Sierhuis, “Modeling and Simulating Work Practice,” in Social Science and Informatics. Amsterdam, The Netherlands: University of Amsterdam, 2001 forthcoming, pp. 350. [11] W. J. Clancey, P. Sachs, M. Sierhuis, and R. van Hoof, “Brahms: Simulating practice for work systems design,” International Journal on Human-Computer Studies, vol. 49, pp. 831-865, 1998. [12] M. Sierhuis, “Modeling and Simulation of Work Practices on the Moon,” presented at Computational Analysis of Social and Organizational Systems 2000, Carnegie Mellon University, Pittsburgh, PA, 2000. [13] M. Sierhuis, W. J. Clancey, R. v. Hoof, and R. d. Hoog, “Modeling and Simulating Human Activity,” presented at Autonomous Agents 2000 workshop on Intelligent Agents for Computer Supported Co-operative Work: Technology and Risks, Barcelona. Spain, 2000. [14] M. Sierhuis, W. J. Clancey, R. v. Hoof, and R. d. Hoog, “Modeling and Simulating Human Activity,” presented at 2000 AAAI Fall Symposium on Simulating Human Agents, North Falmouth, MA, 2000. [15] M. Sierhuis and A. M. Selvin, “Towards a framework for collaborative modeling and simulation,” presented at presented at the Workshop on Strategies for Collaborative Modeling and Simulation, CSCW '96, Boston, MA, 1996. [16] R. van Hoof and M. Sierhuis, “Brahms Language Reference,”: http://www.agentisolutions.com/documentation/langua ge/ls_title.htm, 2000. [17] J. Searle, R., Speech Acts. Cambridge, UK: Cambridge University Press, 1969. 3