How controls influence organizational information processing: insights from a computational modeling investigation

Abstract

In this study, we use a series of computational models to investigate an information processing perspective on organizational control use. We evaluate and compare the information processing capabilities of various formal and informal control configurations under different information uncertainty conditions. We find that a wide range of formal controls can be used to direct subordinates performing interdependent tasks while a more narrow range of informal controls are most effective for directing subordinates who perform complex tasks. Results of this study provide a basis for formalizing an information processing perspective on organizational control implementation that differs but is complementary to the current emphasis on agency in organizational control research.

Appendix

This appendix provides an overview of the VITE’ program

1.1 EC. 1. Overview

The commercial software version 2.2 of the Vite’Project (also VITE’) discrete event computational model is comprised of an agent-based computational modeling platform. Within the parameters specified by the modeler, boundedly-rational computational agents stochastically perform tasks and make decisions while communicating and coordinating their work on projects containing other boundedly-rational computational agents.

According to Levitt et al. (1999, p. 1483), VITE’ “models the total information-processing capacity of an organization as the aggregate information-processing capacities of its nodes, modified by the efficiency of the communication network-comprised of vertical relationships defined by the formal structure, and emergent lateral relationships driven by activity interdependencies—that connects the nodes. The simulator computes the total information-processing load on the organization from the project requirements for direct work and coordination (i.e., control) work. Organizational performance is determined by how closely the organization’s capacity to handle information aligns with the load that it is presented.” The “load” that Levitt et al. (1999) describe refers to the overall amount of information processing effort that agents working on a project within an computational organization collectively produce.

While others have described the core components of the VITE’s computational modeling platform in detail (Jin and Levitt 1996; Levitt et al. 1994, 1999), below we briefly describe the three basic components of the VITE platform: agents, tasks, and the organizational structure.

1.2 EC. 1.1. Agents

Agents within VITE’ work on “projects” consisting of linked tasks. How they perform project tasks is determined stochastically by their behavioral matrix which is developed from information processing principles. Consistency between components of agents’ behavioral matrices and actual human behavior have been tested and verified using both empirical research and extant organizational practice (Levitt et al. 1999). A computational agent’s behavioral matrix specifies their capacity to process and exchange the information necessary to complete their assigned tasks.

1.3 EC. 1.2. Tasks

Agents perform work by stochastically transferring activities from their computationally generated “in-tray” (i.e., work to do) to their “out-tray” (i.e., completed work). How quickly that transfer occurs depends on how efficiently agents can process information related to their own tasks (i.e., production work) and interdependent tasks performed by other agents (i.e., coordination work). The priority the organization places on particular activities as well as the order in which those activities stochastically arrive in an agent’s in-tray determines what activities an agent addresses at any particular point in time.²

To facilitate the effective completion of their tasks, individual agents may also exchange two important categories of information with other agents in the project. First, agents stochastically issue ad-hoc information requests to other agents who perform interdependent tasks. The purpose of these information exchanges is to coordinate and ensure that agents performing interdependent tasks make compatible choices in their respective activities. Second, agents often request that agents pursuing interdependent tasks “rework” failed outputs of task efforts. This most often happens when agents “downstream” in the production process, identify stochastically generated problems or “failures” with the products generated by “upstream” agents.

How efficiently individual agents exchange information with each other is a crucial component of an overall project. This is because when an individual agent submits a request for information or requests that other agents “rework” task products, the agent making the request will suspend work on their task and “wait” for their request to be answered. Each time that an agent waits for their information requests to be answered, the overall project is delayed, thereby compromising how efficiently the organization process information.

1.4 EC. 1.3 Organizational structure

How efficiently agents exchange information within a project is dependent both on agents’ behavioral matrices (described above) and the design of the organization within which agents perform work. Agents within a project are connected to each other within a hierarchy and assigned one of three roles in a decision-making hierarchy: (in order of descending authority) project managers, team managers, or team members. An agent’s hierarchical position determines the types of decisions they make and the amount of information they are required to process (Levitt et al. 1994; Jin and Levitt 1996).

The way in which information is transferred between agents is further determined by VITE’s four “organization” parameters: centralization, formalization, team experience, and matrix strength.

Centralization determines the level of the organizational hierarchy where decisions on “reworking” activity failures occur. In projects with higher centralization, agents higher in the organization (i.e., hierarchy) make rework decisions while in projects with lower centralization (i.e., decentralization), agents who perform tasks tend to make decisions on handling task failures.

Formalization describes the frequency with which agents transmit information requests to other agents. When formalization is high, agents focus on their particular functional duties and make fewer information requests of other agents. When formalization is low, agents seek to collaborate more and initiate a higher number of information requests of other agents.

Matrix strength specifies how often agents respond to the information requests of other agents. When matrix strength is low, agents focus primarily on their individual tasks and respond less readily to other agents’ requests for information. Agents in organizations with higher matrix strength are more collaborative and respond more to other agents’ information requests.

Team experience reflects the total amount of project-related experience that a team possesses. Agents possessing higher levels of team experience work more efficiently because they are familiar with each other and the demands of a particular project.

1.5 EC. 1.4. Project outcomes

Each run of the simulation calculates the collective agent-based information processing effort needed (in thousands of dollars) to generate a specified number production units. This cost is generated based on the collective amount of time it takes all the agents in a particular project organization to process the information necessary to complete their task work.