Norms are shared expectations of behaviours that exist in human societies. Norms help societies by increasing the predictability of individual behaviours and by improving co-operation and collaboration among members. Norms have been of interest to multi-agent system researchers as software agents intend to follow certain norms. But, owing to their autonomy, agents sometimes violate norms which needs monitoring. There are two main branches of research in normative agent systems. One of the branches focuses on normative agent architectures, norm representations, norm adherence and the associated punitive or incentive measures. The other branch focuses on two main issues. The first issue is on the study of spreading and internalization of norms. The second issue that has not received much attention is the emergence of norms in agent societies. Our objective in this paper is to propose mechanisms for norm emergence in artificial agent societies and provide initial experimental results.
Download (PDF, 216 KB)
In this paper we propose a mechanism for norm emergence based on role models. The mechanism uses the concept of normative advice whereby the role models provide advice to the follower agents. Our mechanism is built using two layers of networks, the social link layer and the leadership layer. The social link network represents how agents are connected to each other. The leadership network represents the network that is formed based on the role played by each agent on the social link network. The two kinds of roles are leaders and followers. We present our findings on how norms emerge on the leadership network when the topology of the social link network changes. The three kinds of social link networks that we have experimented with are fully connected networks, random networks and scale-free networks.
Download (PDF, 488 KB)
This paper proposes to design a mechanism that will allow Maori users to specify their privacy preferences related to their culture when a software system asks for culturally sensitive information. We first identify various concepts associated with sensitive aspects of Maori culture, such as tapu. We propose to build an ontology that describes these concepts and the relations between them in a formal way. This ontology will help service providers integrate Maori cultural protocols in order to make Maori users more confident about the use of the sensitive information related to their culture.
Keywords: privacy, Maori culturally sensitive information
Download (PDF, 308 KB)
In tackling data mining and pattern recognition tasks, finding a compact but effective set of features has often been found to be a crucial step in the overall problem-solving process. In this paper we present an empirical study on feature analysis for classical instrument recognition, using machine learning techniques to select and evaluate features extracted from a number of different feature schemes. It is revealed that there is significant redundancy between and within feature schemes commonly used in practice. Our results suggest that further feature analysis research is necessary in order to optimize feature selection and achieve better results for the instrument recognition problem.
Download (PDF, 204 KB)
Referrals are used in multi-agent systems, network agents and peer-to-peer systems for the purpose of global or local information spreading to facilitate trust relationships and reciprocal interactions. Based on referral local interactions can be altered with a purpose to maximise the utility function of each of the participants, which in many cases requires mutual co-operation of participants. The referral system is often based on the global detailed or statistical behaviour of the overall society. Traditionally, referrals are collected by referring agents and the information is provided upon request to individuals. In this article, we provide a simple taxonomy of referral systems and on that basis we discuss three distinct ways information can be collected and aggregated. We analyse the effects of global vs. local information spreading, in terms of individual and global performance of a population based on the maximisation of a utility function of each of the agents. Our studies show that under certain conditions such as large number of non uniformly acting autonomous agents the spread of global information is undesirable. Collecting and providing local information only yields better overall results. In some experimental setups however, it might be necessary for global information to be available otherwise global stable optimal behaviour cannot be achieved. We analyse both of these extreme cases based on simple game-theoretic setup. We analyse and relate our results in the context of e-mail relying and spam filtering.
Download (PDF, 568 KB)
The concept of autonomy is a central concept in distributed computational systems and in multi-agent systems in particular. With diverse implications in philosophy and despite frequent use in social sciences and the theory of computation, autonomy remains somewhat a vague notion. Most researchers do not discuss the details of this concept, but rather assume a general, common-sense understanding of autonomy in the context of computational multi-agent systems. We will review the existing definitions and formalisms related to the notion of autonomy. We re-introduce two concepts: relative autonomy and absolute autonomy. We argue that even though the concept of absolute autonomy does not make sense in computational settings, it is useful if treated as an assumed property of computational units. For example, the concept of autonomous agents may facilitate more flexible and robust abstract architectures. We adopt and discuss a new formalism based on results from the study of massively parallel multi-agent systems in the context of evolvable virtual machines. We also present the architecture for building such architectures based on our multi-agent system KEA, where we use the extended notion of dynamic linking. We augment our work with theoretical results from cham algebra for concurrent and asynchronous information processing systems. We argue that for open distributed systems, entities must be connected by multiple computational dependencies and a system as a whole must be subjected to influence from external sources. However, the exact linkages are not directly known to the computational entities themselves. This provides a useful notion and the necessary means to establish an relative autonomy in such systems.
Download (PDF, 528 KB)
This paper presents a new approach of dynamic symmetric key distribution for encrypting the communication between two nodes in a Wireless Sensor Network (WSN). The distribution of a shared key can be performed by any sensor node and does not always require that it is performed by the base station (BS). Each node can be selected by one of its neighbor nodes in order to distribute a pair-wise key for a communication between two nodes. The selection is based on the local computation of a trust value granted by the requesting nodes. This scheme considerably reduces the cost of communication between the BS and the nodes when setting up pair-wise keys between neighboring nodes. This paper also describes a dynamic route selection mechanisms based on trust and cost, that each node performs to route data to neighbor nodes and to the BS.
Keywords: key distribution, trust, wireless sensor network, route selection,
Download (PDF, 448 KB)
One approach to moderating the behaviour of agents in open societies is the use of explicit languages for defining norms, conditional commitments and/or social expectations, together with infrastructure supporting conformance checking and the identification and possible punishment of anti-social agents. This paper presents a logical account of the creation, fulfilment and violation of social expectations modelled as conditional rules over a hybrid propositional temporal logic. The semantics are designed to allow model checking over finite histories to be used to check for fulfilment and violation of expectations in both online and offline modes. For online checking, expectations are always considered at the last state in the history, but in the offline mode expectations in previous states are also checked. At each past state, the then active expectations must be checked for fulfilment without recourse to information from later states: the truth of a future-oriented temporal proposition φ at state s over the full history does not imply the fulfilment at s of an expectation with content φ. This issue is addressed by defining fulfilment and violation in terms of an extension of Eisner et al.’s weak/strong semantics for LTL over truncated paths. The update of expectations from one state to the next is based on formula progression and the approach has been implemented by extending the MCLITE and MCFULL algorithms of the Hybrid Logic Model Checker.
Download (PDF, 240 KB)