Abstract
Multi-agent organisations can use norms to regulate agent behaviour without strict regimentation. These norms are designed to encourage behaviour that facilitates cooperation and improves organisational performance, while discouraging undesirable behaviour. However, in a dynamic environment, behaviour which was once desirable may become undesirable and vice versa, thus changing the impact a norm has upon organisational performance. In order to ensure that norms remain congruent to organisational goals, their impact must be monitored over time, and, if found to be inadequate, new norms must be selected to replace them.We address the problem of quantifying and monitoring norm impact using a simulation approach: we simulate organisational activity both with and without the norm of interest in order to estimate its effect. In order to reduce the monitoring effort we also propose a mechanism for deciding which sets of norms to monitor based upon the impact magnitude and sensitivity to environmental change. We present a method for generating new norms via genetic programming in order to replace ineffective norms with more effective ones. Changing organisational norms during run-time can cause disruption as agents may need to abandon their ongoing work to accommodate new norms. We provide a mechanism to reduce this disruption by taking into account agent activity when choosing when to change norms. In addition, we propose a method to use the quantification of norm impact to decrease the risk of allowing external agents of unknown trustworthiness to play organisational roles.
| Date of Award | 2016 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Simon Miles (Supervisor) & Michael Luck (Supervisor) |