The evolution of rules for conflicts resolution in self-organizing teams

https://doi.org/10.1016/j.eswa.2011.07.014Get rights and content

Abstract

The purpose of the paper is to study the emergency and effects of conflict resolution rules in self-organizing teams. Intelligent agents are used to simulate team members of self-organizing teams. In the virtual self-organizing team, agents adapt the Q-learning algorithm to adjust their actions. Three sets of experiments are manipulated to study the evolution of rules. The results of few experiments show a new rule for conflict resolution emerged from the dynamic interactions of agents. For the other experiments, agents cannot resolve conflicts by themselves.

Highlights

► The emergency of the new rule can be proposed by the multi-agent model. ► The increase of conflicts’ amount at each period blocks the evolution of new rules. ► The increase of team scale can not promote the emergency of new rule for conflict resolution.

Introduction

In self-organizing teams, team members adopt knowledge to work for the team task. The team task consists of several dependent sub-tasks. Each sub-task has requirements for the member who want to accomplish it. Since there are no stronger leaders of the teams, each member chooses the sub-task whose requirements are fit with his abilities. The quality of team tasks depends on the worst quality of team members’ works. Since the characteristics of self-organizing teams, the fit rule between tasks and members is “Do the fittest task”. However, if the team members’ ability is not perfect for the tasks, the fit rule will lead to team problems. For example, two tasks requirements are 50 (t1) and 70 (t2). Two members have the ability of 70 (m1) and 90 (m2). Based on the rule of “Do the fittest task”, m1 chooses the task of t2. In order to accomplish the team task, m2 have to choose t1. Since the distance between m2 and t1, the quality of team task is 40 (90  50). This scenario is defined as assignment conflict in the paper. The paper studies the rules to resolve this kind of assignment conflicts in self-organizing teams.

For the target of high tasks’ quality, the optimal assignment of this example is m1 choose t1 and m2 do the task of t2. For the optimal assignment, the quality of team task is 20 (70  50 or 90  70). In the paper, this optimal assignment rule is defined as “Do a fitter task”. Team members following fixed behavioral rules can be limited in performance and efficiency. In order to emerge the rule of “Do a fitter task” from the dynamic interactions of team members, the paper uses multi-agent technology to simulate the self-organizing teams. Intelligent agents are used to simulated team members. Adaptability is key components of intelligent behavior which allow agents to improve performance in a given domain using prior experiences. The Q-learning algorithm is applied to improve the self-adaptive ability of agents. Three sets of experiments are manipulated to analyze the evolution of the rule in self-organizing teams. The emergency and effects of conflict resolution rules are analyzed by the experiments’ results.

The rest of the paper is organized as follows. The related literature is reviewed in Section 2 and then the multi-agent model is developed in Section 3. The experiments are conducted in Section 4. A detailed result analysis is presented in Section 4. Finally, the conclusions are summarized and future work is suggested in Section 5.

Section snippets

Review of the related research

For the management of self-organizing teams, Romme built a model of self-organizing processes in top management teams and described Boolean comparison as a rigorous method for testing process theories on the basis of qualitative evidence from case studies (Romme, 1995). Levi and Slem examined professional level teams in research and development facilities at three corporations. All of these corporations were attempting to introduce self-managing teams in their R&D projects. These teams

Conflicts resolution model of self-organizing team

The virtual self-organizing team (V = {v1, v2, v3, …, vN}) is a multi-agent model containing heterogeneous agents (vi, virtual members) which act in a virtual environment. All members cooperate to accomplish tasks with their knowledge. Each member is simulated by an agent in the model. The team task (M = {m1, m2, m3, …, mN}) consists of N sub-tasks. If the member vi has the ability which is required by sub-task mj, vi can do mj with high quality and vi can get maximal profits. If vi does not fit the ability

Experiments of the virtual team

The agent-based model is programmed by JAVA based on RePast. The program is run on WinXP. In order to show the validity of the model, the paper conducts three sets of experiments of the model. The parametric settings for the experiments are proposed in Table 1. The counts of team members (N) for different scenarios are 5, 10, and 20. Except the parameters of N, the parametric settings of three scenarios are same in Table 1.

Conclusions and future work

The motivation of this paper was to analyze the emergency and effects of conflict resolution rules in self-organizing teams. Multi-agent technology was used to simulate the self-organizing team and the Q-learning algorithm was proposed to adjust agents’ behavior. Three sets of experiments were performed to study the emergency and effects of the rule (“Do a fitter task”) in a virtual self-organizing team. The results indicate that the emergency of the new rule depends on the scale of the team

Acknowledgements

This research was supported in part by: (i) the NSFC (National Natural Science Foundation of China) under Grant 70771045 and (ii) Fund of Jiangsu Agricultural Machinery Bureau (Grant Number- GXS10012).

References (22)

  • M.R. Fuller et al.

    Does fit matter? The impact of TTF and appropriation on team performance in repeated tasks

    Information Systems Research

    (2009)
  • Cited by (0)

    View full text