i-AGENTS: Modeling Organizational Problem Solving in Multi-Agent Teams

Organizations involving human and computer agents are constrained by a variety of factors including: task properties and arrangements; level of technology; knowledge held by, and distributed among, the agents; information and administrative structures; and organizational norms and policies. An important challenge to the scientific community is to develop, validate and apply theories and models to help managers re-engineer their organizations for higher levels of performance. Our research on organizational problem solving aims to develop a computational model of organizations to study interrelationships between agents' knowledge, task requirements, and organization structures and policies. This paper reports the first step of our research toward a computational organizational model—the i-AGENTS framework, a prototype computer system for modeling organizations of intelligent agents. i-AGENTS is composed of a number of high-level concepts: tasks, agents, organization and communication. A task is described in detail by task action, task object and task constraints; an agent is modeled to consist of cognitive attributes and expertise; role-based organizational structure is adopted for describing organizations. From an organization perspective, i-AGENTS extends traditional information processing models of organization (Galbraith 1977) by explicitly addressing the role of agents' knowledge of both the problem domain and the organization in problem solving. When viewed from an engineering perspective, our research is the first step toward an organizational problem-solving model that merges organization theory and distributed artificial intelligence and can be used to simulate and analyze organizational behavior of teams in engineering domains at a very specific level of detail.     

The combined effects of user schemas and degree of cognitive fit on data retrieval performance

Given the massive and accelerating amount of data modern organizations are collecting, it is imperative that employees possess the skills to navigate various data structures and develop sound data manipulation and retrieval strategies. This research investigates whether and how user schemas and the degree of cognitive fit combine to affect data retrieval task performance. We measured two types of user schemas associated with debit-credit-accounting (DCA) and resource-event-agent (REA) accounting systems. All participants completed tasks that were either facilitated by DCA or REA. Degree of cognitive fit was manipulated as high (when the task was facilitated by the system) or low (when the task was not facilitated by the system). Results show that the positive association between users' schemas and data retrieval performance is enhanced when the degree of cognitive fit is high, but is attenuated when the degree of cognitive fit is low. The findings that participants with equivalent amounts of training on the accounting models had varying schema strengths for those models provide prima facie evidence that one should not assume schemas' existence based on experience. Of particular importance is the finding that cognitive fit is even more important than schemas, as this provides guidance for companies interested in facilitating data retrieval to focus their decision support efforts first on providing interfaces that match the tasks to be performed and second on training their decision makers to develop schemas consistent with the interfaces. While the combination of strong user schemas and high cognitive fit will yield the best results, if a company must choose due to limited resources, the provision of high cognitive fit with the interface-task match is more important than developing strong user schemas.     

The effects of multimedia on cognitive aspects of decision-making

Businesses make multimedia presentations, combining film clips and text, which could be made at less cost using a single medium. Extending prior findings that film clips in multimedia presentations can impact affective states (e.g., [Rose J. Behavioral decision aid research: Decision aid use and effects. In Arnold and Sutton, Research Accounting as an Information Systems Discipline (ch 6). Sarasota: AAA; 2002]), our study examines how multimedia presentations affect decision-making cognitively (rather than affectively). Participants performed two types of tasks using information presented to them in two single medium modes (text-only and film-only) and a multimedia mode (combined text/film). Based on theories of presentation modes/modalities, media richness, and information overload research, we hypothesize the effects of different presentation modes (single medium versus multimedia), task familiarity (familiar versus unfamiliar), and information overload (above versus below overload) on decision-makers' consistency with, compliance with, and willingness to apply externally provided information processing rules concerning the solving of various types of tasks. Results provide full to partial support for the predicted effects of task familiarity and information overload on consistency, compliance and willingness. Specifically, multimedia presentations increase the willingness to apply externally provided rules but decrease the consistency with which rules are applied to tasks. Overall, our study suggests that decision-makers are reluctant to use externally provided instructions, instead of their own internally learned rules, especially as their understanding of the task at hand increases. Such understanding is enhanced by high task familiarity, clear presentation formats (i.e., single medium), and low information load. This study has implications for businesses trying to decide how much to invest in multimedia presentations, and also on organizations and society at large as they frequently direct individuals to follow various types of instructions.     

Organizational Structures and Exchange Processes

Formal organizations have a division of labor and some form of co-ordination. This paper explores how these social structures could be generated by an exchange process that governs work flow through an organization. Specifically, the exchange process controls how work tasks are assigned and carried out, and how members receive rewards for participating in a work group of the organization. A discrete-event simulation model is used to explore how the exchange process generates division of labor and co-ordination networks. The model contains members, called bureaucrats, with two kinds of skills: skill at referring work to others; and skill in completing work tasks. Both kinds of skills are distributed differentially across bureaucrats. Two types of rational exchange are explored: independent exchange and pooled exchange. For independent exchange, rewards go the bureaucrat who completes the task; in pooled exchange rewards are shared with the bureaucrat who referred the task. Both exchange processes generate a division of labor in a work group, with pooled exchange exhibiting the most pronounced division of labor. Also, pooled exchange leads to more efficient processing of work than independent exchange.     

A longitudinal study of applicable decision aids for detailed tasks in a financial audit

This paper reports on changes in the extent to which decision aids are perceived to be useful for performing detailed financial audit tasks. The decision aids investigated were automation (AU), decision support systems (DSS), and knowledge-based expert systems (KES). Data were collected from highly experienced mangers and partners from various international accounting firms in 1988 and 1996. The participants were asked to select the decision aid they viewed as applicable for each task in a comprehensive inventory of audit tasks. While the data indicate an increase in the choice of tasks for decision aids use between the two years, the fact remains that a majority of the responses (79% in 1988 and 69% in 1996) indicated human processing (HP) only. We present a list of detailed tasks for which at least 50% of the responses indicated some form of an applicable decision aid in 1996. We also discuss implications for practice and research. Copyright © 2001 John Wiley & Sons, Ltd.     

Early evidence of digital labor in accounting: Innovation with Robotic Process Automation

Robotic Process Automation (RPA) is an emerging technology that enables the automation of rules-based business processes and tasks through the use of software bots. Drawing upon the theory of Task-Technology Fit (TTF) and Technology-to-Performance Chain (TPC) (Goodhue and Thompson 1995) and research on expert systems (Messier and Hansen 1987; Sutton 1990), this study explores emerging themes surrounding bot implementation for accounting and finance tasks. We collect and analyze interview data from adopters of RPA and document task suitability, task-technology fit, implementation issues, and resulting performance outcomes. We find that securing technical capability is only a part of RPA implementation process. Organizations engage in standardization and optimization of processes, develop scorecard-like tools to rank tasks, adjust governance structures to include digital employees, and redefine internal controls. Organizations benefit from automating only certain processes, those that are structured, repeated, rules-based, and with digital inputs. Along with cost savings, organizations experience improved process documentation, lower error rates, more accurate measurement of process performance, and better report quality.     

Beyond brainstorming: The effectiveness of computer-mediated communication for convergence and negotiation tasks

Although a considerable body of research in information systems has established that computer-mediated communication (CMC) is beneficial for brainstorming (idea generation) tasks, less is known about its effectiveness for more complex decision-making tasks. This paper reports the results of two experiments comparing the performance of face-to-face and CMC teams in decision-making tasks that move beyond brainstorming. In the first experiment, the performance of face-to-face and computer-mediated teams was compared in two tasks: one requiring participants to engage in convergent thinking and a second brainstorming task requiring divergent thinking. Consistent with predictions derived from McGrath's task circumplex model, the results of experiment one reveal that participants using computer-mediated communication perform significantly better than those interacting face-to-face on the divergent (brainstorming) task. On the convergent task, computer-mediated and face-to-face teams performed equally well; i.e., there was not a significant difference in their performance. In the second experiment, the performance of face-to-face and computer-mediated teams was again compared in two tasks: an integrative negotiation task and an idea-generation task. The results of the second experiment were similar to those of experiment one, in that computer-mediated teams significantly outperformed face-to-face teams in the idea-generation task, while computer-mediated and face-to-face teams performed equally well on the integrative negotiation task. These experiments contribute to the literature by shedding additional light on the conditions under which computer-mediated communication is as effective as, and in some cases more effective than, face-to-face interaction.     

Teams versus individual accountability: solving multitask problems through job design

Many organizations are structured so that workers are jointly accountable for performance, even though there exist alternative organizational structures that align incentive compensation more closely with each worker's tasks. I develop a multitask agency model that demonstrates that such organizations may be optimal when multitask problems are severe or risk considerations are not too important. I also show that, in some circumstances, it may be optimal to share poorly measured tasks among several agents, contrary to the results of the existing multitask literature.     

Management: Agent of human cultural evolution

While the management discipline has been widely recognized as a culture, it is more instructive to view this culture as part of a larger system, to view it in fact as an integral part of—and an effective agent in—the process of human cultural evolution. This article argues that every management task and decision is influenced by—and in turn influences—the cultural environment and its evolution. Decision making in social organizations, mediated by the available ‘cultural pool’, is shaping our future on every level.     

Cooperative Knowledge Processing: The Key Technology for Future Organizations

Drawing from the challenges organizations are faced with today, there is a growing understanding that future market success and long-term survival of enterprises will increasingly be related to the effectiveness of information technology utilization. This, however, requires us to intertwine much more seriously organizational theory and research in information processing than previously. In this paper we approached this aim from the perspective of radically decentralized, computerized enterprises. We further assume that organizations are increasingly process-oriented, rather than applying to structuring organizations based on task decomposition and assignment. This scenario reveals that, due to the inherent autonomy of organizational units, the coordination of decentralized organizational activities (workflows, processes) necessitates a cooperative style of problem solving. On this basis, the paper introduces the research area of cooperative knowledge processing, with a particular focus on multi-agent decision support systems and human-computer cooperative work. Finally, several important organizational applications of cooperative knowledge processing are presented that demonstrate how future enterprises can take great advantage from these new technologies.     

A Computational Model of Coordination for the Design of Organizational Decision Support Systems

The Information Technology (IT) for realizing Organizational Decision Support Systems (ODSS) is in a nascent stage of development. This is particularly true in the area of coordination, which is a critical element of ODSS, and which distinguishes ODSS research from earlier research in Group DSS and individually oriented DSS. As a first step in ODSS coordination research, alternative representation schemes need to be examined in terms of both their match with the prevailing needs of organizations and of existing IT approaches that can be brought to bear. Matching ODSS needs with coordination representation requirements is examined by using several supporting reference disciplines including foundational DSS and recent ODSS research frameworks/architectures. Existing IT approaches are adapted from the reference disciplines of Active DSS, Distributed Artificial Intelligence (DAI), and Mathematical/Computational Organization Theory (MCOT) to operationalize a computational model of coordination that: (1) embodies the philosophies of Active DSS—including the idea that automated intelligent agents can play a significant role in supporting decision makers by independently carrying out rudimentary tasks to support the various phases of a decision making process; (2) adapts DAI and IT approaches to reflect practical human organizational realities including what we refer to as the ‘Open-Ended Knowledge World‘, and the evolutionary nature of organizations—whereby ODSS coordination representations will be subjected to almost constant revision due to both external environment disruptions and internal events that require adjustments to a preliminary plan; and (3) reflects the fact that organizational goals are often vague, which implies that a coordination representation should be sufficiently robust to reflect ad hoc analysis accommodating of strategy changes.     

Impacts of Additional Guidance Provided on International Financial Reporting Standards on the Judgments of Accountants

This pioneering study examines the impact of the provision of additional guidance on International Financial Reporting Standards (IFRS) as a “decision aid” on the accuracy of judgments of the accountants. To extend the prior research on accounting judgment and decision-making, we also examine the interactive effects of task complexity and additional guidance on the judgments of accountants. The results provide evidence that those accountants who are provided with decision aid in the form of additional guidance on IFRS make more accurate judgments than accountants who are not provided with such guidance. Furthermore, the study provides evidence that this additional guidance improves the judgments of accountants when they undertake tasks which they find complex. The results indicate that additional guidance on IFRS needs to be provided and suggests that accountants should exploit any guidance which is currently provided in IFRS and by the International Financial Reporting Standards Interpretations Committee.     

Specificity of reinforcement for risk behaviors of the Balloon Analog Risk Task using math models of performance

Risky decision-making has been studied using multitrial behavioral tasks. Concordance of such tasks to risky behaviors could be improved by: (1) mathematically modeling the components of decision change and (2) providing reinforcement specific to the risk behavior studied. Men completed two Balloon Analog Risk Tasks (BART). One provided financial reinforcement (money) and the other provided sexual reinforcement (seconds of erotic film viewing). Parameters of a mathematical model of BART performance were fit to each individual. Correlations between the model parameters and four risk categories (financial, sexual, antisociality, and substance use) demonstrated predictive utility for the same behaviors regardless of task reinforcement, providing little evidence of reinforcement specificity. A reward sensitivity parameter was uniquely related to sexual risk behavior. Additional analyses explored parameter stability fit to fewer trials.     

SPAN OF CONTROL, TASK INTERDEPENDENCE AND USEFULNESS OF MAS INFORMATION IN NOT-FOR-PROFIT GOVERNMENT ORGANIZATIONS

This study examines the relationships between subunit supervisors' span of control, their perceived task interdependence, and their perceived usefulness of MAS (management accounting systems) information in not-for-profit government organizations. One hundred and forty-nine subunit supervisors/managers working in twenty-one public hospitals in New Zealand participated in the study. The study used structured questionnaires for data collection. The results suggest that supervisors' perceived task interdependence intervenes in the relationship between their span of control and their perceived usefulness of MA information for decision making. The types of MAS information considered were broad in scope, timely available, aggregated and integrated. A path analytic technique was applied to test the model used in the study.     

Quantitative Modeling of Complex Environments

There are many formal approaches to specifying how the mental state of an agent entails the particular actions it will perform. These approaches put the agent at the center of analysis. For some questions and purposes, it is more realistic and convenient for the center of analysis to be the task environment, domain or society of which agents will be a part. This paper presents such a task environment-oriented modeling framework that can work hand in hand with more agent-centered approaches. Our approach features careful attention to the quantitative computational interrelationships between tasks, to what information is available (and when) to update an agent's mental state and to the general structure of the task environment rather than single-instance examples. A task environment model can be used for both analysis and simulation, it avoids the methodological problems of relying solely on single-instance examples and provides concrete, meaningful characterizations with which to state general theories. This paper is organized around an example model of co-operative problem solving in a distributed sensor network.     

Matching electronic communication media and audit tasks

Increasing marketplace demand for real-time reporting of business information is inevitably leading to the requirement for more frequent assurance from auditors over the accuracy and reliability of such information. The frequency and speed with which auditors must provide assurance, coupled with the global dispersion of business clients, place more demand on the use of electronic communication media in all phases of the assurance process. However, there is little theoretical guidance concerning how to best match various electronic communication media representations to audit tasks. Accordingly, the purpose of this study is to develop a media–task fit (METAFIT) model that can be used in future research endeavors, particularly for information inquiry tasks in judgment and decision-making. The METAFIT model specifies conditions and factors leading to an optimal bilateral (auditor–client) METAFIT, with the objective of maximizing task effectiveness.     

Cost Management in Sri Lanka: A Case Study on Volume,Activity and Time as Cost Drivers

Despite its theoretical superiority, the activity-based costing (ABC) model has had only moderate success in replacing the traditional volume-based absorption costing models in complex organizations worldwide. Even in organizations that have launched ABC projects, the implementations often do not sustain. In response to this general lack of enthusiasm worldwide for ABC, accountants developed the time-driven activity-based costing (TDABC) model as an alternative cost allocation model. This paper presents a comparison of the TDABC model with ABC, and considers if this alternative cost allocation model (1) is easier to implement from an international perspective and (2) provides comparable cost information for decision making. We use a case study in a country outside the model's country of origin to understand the similarities and differences in absorption costing systems that use ‘volume,’ ‘activities,’ and ‘time’ as the drivers of indirect cost allocations. We also use the case study to ascertain if any country‐specific factors impede ABC implementation. We conclude the following: the TDABC model has similar implementation complexities to ABC if modelling conditions are strictly adhered to; these complexities are independent of country‐specific factors; and in its simplest form, the model generates the same decision information errors of traditional costing.     

Electronic data interchange: an evaluation of alternative organizational forms

Electronic Data Interchange (EDI) has come to revolutionize in recent years the way in which businesses conduct their trading activities. This study examines alternative organizational forms for EDI processing within organizations and evaluates them on several task performance indices including independence, saturation, and satisfaction, and also relates them to EDI effectiveness as a function of information technology (IT) intensity and length of EDI use. Results from an extensive, two-phase survey of EDI users indicate that organizational form for EDI processing influenced task performance indices such that the more decentralized structures (i.e. star and circle structures) exhibited higher independence and lower saturation than the more centralized structures (i.e. wheel, kite, and chain structures). More decentralized forms also exhibited higher satisfaction with EDI than the more centralized forms on nonroutine EDI tasks. EDI task performance indices were also correlated with EDI effectiveness, such that higher independence, lower saturation, and higher satisfaction were associated with higher EDI effectiveness. IT intensity and length of EDI use also positively moderated the effects of decentralization on EDI effectiveness. These results are discussed in terms of the organizational implications of this strategic IT resource: EDI.     

A review of dashboards in performance management: Implications for design and research

Dashboards are expected to improve decision making by amplifying cognition and capitalizing on human perceptual capabilities. Hence, interest in dashboards has increased recently, which is also evident from the proliferation of dashboard solution providers in the market. Despite dashboards' popularity, little is known about the extent of their effectiveness, i.e. what types of dashboards work best for different users or tasks. In this paper, we conduct a comprehensive multidisciplinary literature review with an aim to identify the critical issues organizations might need to consider when implementing dashboards. Dashboards are likely to succeed and solve the problems of presentation format and information load when certain visualization principles and features are present (e.g. high data-ink ratio and drill down features). We recommend that dashboards come with some level of flexibility, i.e. allowing users to switch between alternative presentation formats. Also some theory driven guidance through pop-ups and warnings can help users to select an appropriate presentation format. Given the dearth of research on dashboards, we conclude the paper with a research agenda that could guide future studies in this area.     

A Comparative Analysis of Inductive-Learning Algorithms

Recently there has been an increasing interest in applying inductive learning algorithms to generate rules/patterns from a given example set. While such approaches serve as an efficient way of resolving the knowledge-acquisition bottleneck, their predictive accuracy, which is the popular measure of performance, varies widely. This paper contrasts major inductive-learning algorithms and examines their performance with two performance measures: the predictive accuracy and the representation language. Experiments involved three inductive-learning algorithms and five different managerial tasks in construction project assessment and bankruptcy-prediction domains. The test results indicate that the model performance is dependent on tasks with an exception of the neural network model and that there is a an effect of group proportion in the example set used to construct the model. The neural network approach presents relatively stable predictive power across different task domains, although it is difficult to interpret its representation.