Skip to main content
main-content

Über dieses Buch

This book constitutes the thoroughly reviewed post-proceeding of International Workshops on Coordination, Organization, Institutions and Norms in Agent Systems, COIN@AAMAS 2012, held in Valencia, Spain in June 2012. The 13 revised full papers presented together with 1 invited talk went through several rounds of reviewing and revision and were carefully selected for presentations. The papers are organized in topical sections on compliance and enforcement, norm emergence and social strategies, refinement, contextualisation and adaptation.

Inhaltsverzeichnis

Frontmatter

Invited Talk

Situating COIN in the Cloud

(Invited Paper)
Abstract
We start from the view that the central theme of the research at the core of coordination, organization, institutions and norms, is whether the social structures and mechanisms, that have emerged over time, can be adapted and applied to artificial societies of programs and perhaps more significantly, to mixed societies of humans and programs –and how The means by which the social constraints that guide and regulate behaviour are acquired and represented remains an open problem. If recent experiences in information retrieval and natural language processing are plausible indicators, the statistical may yet oust the logical. Technology aside, it is clear that for socio-technical systems, that integrate human and software components, we may expect the adoption of, or the illusion of observation of, and support for human social conventions. The growing migration to cloud computing of the services that make up current pervasive, social applications suggests near-term developments emerging from the same platform(s). Thus, the question considered here is, what pathways, opportunities and challenges exist for the development, wider use and validation of COIN technologies to help realize socio-technical systems that better meet human requirements. As examples of specific enabling technologies, we review current developments in resource-oriented architecture, complex event processing and stream reasoning and observe how COIN technologies might integrate with them.
Julian Padget

Compliance and Enforcement

Monitoring Interaction in Organisations

Abstract
In an organisational setting, such as an online marketplace, the organisation monitors agent interactions, and enforces norms by means of sanctions. This paper provides an operational semantics for agent interactions within such a setting, distinguishing constitutive norms for monitoring and sanction rules for enforcement of norms. Our contribution emphasizes a more detailed exploration of the processes of monitoring commitments created through agent interactions and imposition of sanctions when commitments are violated. We consider both agent–agent and agent–environment interactions, focusing on operationalizing enforcement of commitment-based norms. We provide a generic way to develop operational semantics from specific definitions of norm behaviour. For an example set of norm behaviours, we sketch some formal properties that follow from our semantics, such as continuity, (non-)interference, and (non-)redundancy.
Mehdi Dastani, Leendert van der Torre, Neil Yorke-Smith

Reasoning over Norm Compliance via Planning

Abstract
Norms are a way to provide some flexibility to the specification of acceptable actor behaviour in a shared context. Instead of viewing norms as static restrictions over an agent’s conduct (and autonomy), the full power of normative specifications comes when norms are seen as guidelines that agents can use in their decision-making. In literature there is a lot of work on norm theories, models and specifications on how agents might take norms into account when reasoning, but many of them focus on the goal or intention selection and few of them apply the norms into the agent’s plan generation. In this paper we present a norm-oriented agent that takes into consideration operationalised norms during the plan generation phase, using them as guidelines to decide the agent’s future action path. In our work norms can be obligations or prohibitions which can be violated, and are accompanied by repair norms in case they are breached. To make norms influence plan generation, our norm operational semantics is expressed as an extension/on top of STRIPS semantics, acting as a form of temporal restrictions over the trajectories (plans) computed by the planner. In combination with the agent’s utility functions over the actions, the norm-aware planner computes the most profitable trajectory concluding to a state of the world where the effects of all the active norms have been explored, including the repair norms. We use a simplified fire emergency scenario in order to demonstrate the usefulness of the framework, integrating the norm-aware planner to 2APL agent architecture. We depict possible outcomes depending on criteria such as time and danger.
Sofia Panagiotidi, Javier Vázquez-Salceda, Frank Dignum

An Agent-Based Simulation Approach to Comparative Analysis of Enforcement Mechanisms

Abstract
Incentive-based enforcement can be an effective mechanism for fostering cooperation in open distributed systems. The strength of such systems is the absence of a central controlling instance, but at the same time, they do depend upon (voluntary) regulation to achieve system goals, creating a potential “tragedy of the commons”. Many different mechanisms have been proposed, both in the multi-agent systems and the social science communities, to solve the commons problem by using incentive-based enforcement. This paper advocates the use of agent-based simulation to carry out detailed comparative analysis of competing enforcement mechanisms, by providing common settings, the environment and the basis for comprehensive statistical analysis. To advance this argument, we take the case study of wireless mobile grids, a future generation mobile phone concept, to ground our experiments and analyse three different enforcement approaches: police entities, image information and a well-known existing reputation mechanism. The contribution of this paper is not the enforcement mechanisms themselves, but their comparison in a common setting through which we demonstrate by simulation and statistical analysis that enforcement can improve cooperation and that a relatively small percentage (of the population as a whole) of police agents outperforms (under the chosen metrics) image- and reputation-based approaches. Hence, qualified conclusions may be drawn for the application of such mechanisms generally in open distributed systems.
Tina Balke, Marina De Vos, Julian Padget

Norm Emergence and Social Strategies

Shared Strategies in Artificial Agent Societies

Abstract
A shared strategy is a social concept that refers to a type of behavioural pattern that is followed by a significant number of individuals although it is, prima facie, not associated with an obligation or a prohibition. E. Ostrom has argued in favour of the pertinence of social strategies for institutional design and evolution and proposed a characterization suggestive of formal treatment. However, shared strategies as such have not been explicitly used in the context of regulated MAS in spite of their relevance and their affinity to more standard normative notions, of which a rich tradition exists in MAS research. In this paper, we discuss the notion of shared strategy, characterize its distinguishing features, propose its formalization using a temporal epistemic logic, and explore its potential use in regulated multi-agent systems.
Amineh Ghorbani, Huib Aldewereld, Virginia Dignum, Pablo Noriega

Goal-Directed Policy Conflict Detection and Prioritisation

Abstract
A policy (or norm) is a guideline stating what is allowed, forbidden or obligated for an entity, in a certain situation, so that acceptable outcomes are achieved. Policies occur in many types of scenarios, whether they are loose social networks of individuals or highly structured institutions. It is important, however, for policies to be consistent and to support their goals. This requires a thorough understanding of the implications of introducing specific policies and how they interact. It is difficult, even for experts, to write consistent, unambiguous and accurate policies, and conflicts are practically unavoidable. In this paper we address this challenge of providing automated support for identifying and resolving logical and functional conflicts.
Mukta S. Aphale, Timothy J. Norman, Murat Şensoy

Norms as Objectives: Revisiting Compliance Management in Multi-agent Systems

Abstract
This paper explores a hitherto largely ignored dimension to norms in multi-agent systems: the normative role played by optimization objectives. We introduce the notion of optimization norms which constrain agent behaviour in a manner that is significantly distinct from norms in the traditional sense. We argue that optimization norms underpin most other norms, and offer a richer representation of these. We outline a methodology for identifying the optimization norms that underpin other norms. We then define a notion of compliance for optimization norms, as well as a notion of consistency and inconsistency resolution. We offer an algebraic formalization of valued optimization norms which allows us to explicitly reason about degrees of compliance and graded sanctions. We then outline an approach to decomposing and distributing sanctions amongst multiple agents in settings where there is joint responsibility.
Aditya Ghose, Tony Bastin Roy Savarimuthu

Refinement, Contextualisation and Adaptation

Norm Emergence through Dynamic Policy Adaptation in Scale Free Networks

Abstract
As has been stated elsewhere, norms are a valuable means of establishing coherent cooperative behaviour in decentralised systems in which there is no central authority. Axelrod’s seminal model of norm establishment in populations of self-interested individuals provides some insight into the mechanisms needed to support this through the use of metanorms, but considers only limited scenarios and domains. While further developments of Axelrod’s model have addressed some of the limitations, in particular in considering its application to different topological structures, this too has been limited in not offering an effective means of bringing about norm compliance in scale-free networks, due to the problematic effects of hubs. This paper offers a solution, first by adjusting the model to more appropriately reflect the characteristics of the problem, and second by offering a new dynamic policy adaptation approach to learning the right behaviour. Experimental results demonstrate that this dynamic policy adaptation overcomes the difficulties posed by asymmetric distribution of links in scale-free networks, leading to an absence of norm violation, and instead norm emergence.
Samhar Mahmoud, Nathan Griffiths, Jeroen Keppens, Michael Luck

Norm Contextualization

Abstract
Agents interact with each other regulating by a set of norms which is expressed at different levels of abstraction that capture different contexts and operationalizations. Current normative frameworks deal with norm operationalization, yet few consider the contextual aspects of norms. Moreover, most frameworks are based on the independent evaluations of norms, which makes it difficult to evaluate interrelated effects of different norms and contexts. In this paper, we propose Norm Nets as a formalism to capture the structure of contextualized norm sets. This formalism will enable (1) the analysis of interrelations between norms, (2) the contextualization of normative statements, and (3) the verification of properties of interrelated norms. We apply this framework to a case study taken from the domain of international trade.
Jie Jiang, Huib Aldewereld, Virginia Dignum, Yao-Hua Tan

Programming Institutional Facts in Multi-Agent Systems

Abstract
In multi-agent systems with separate agents, environment, and institution dimensions, the institutional state can be affected by facts originating in any of those constituent dimensions. Most current approaches model the dynamics of the institution focusing on the agents and the institution itself as the main sources of facts that produce changes in the institutional state. In this paper, we investigate also the environment as an important source of facts that change the institution. We propose thus a model and a language to specify and program the institutional dynamics as consequence of events and state changes occurring in any of the three component dimensions of the system (agent, environment, and institution). Our approach was evaluated through a case study where we compare two solutions for an application: the original design and a new one based on our proposal. We observed a simplification of the agents’ reasoning, an increase in the functions performed by the environment and the institution, and greater independence of the agents within the system. This last result is specially important in open systems where we cannot take for granted that agents will take part in the system.
Maiquel de Brito, Jomi F. Hübner, Rafael H. Bordini

Towards a General Model for Adapting Structure while Maintaining Topology: Pipelines

Abstract
Large scale information systems are increasingly structured around flexible workflows of services providing a range of functionalities that are configured to suit particular needs, yet this flexibility can bring a lack of organisation in the ways in which services are combined. Particular system structures bring different benefits to an application in terms of efficacy and efficiency but sometimes need to reorganise as their circumstances change. In this context, this paper seeks to establish techniques for reorganisation that preserve particular topologies in support of their recognised benefit for the target applications. The contributions are twofold: first, a general vision of reorganisation of defined topologies, in which topology is preserved but efficiency and efficacy are optimised; and second, a specific solution for the case of pipelines, reorganising to optimise key application-specific metrics, while preserving topology. The paper is thus the starting point for a more ambitious general programme of research.
Matthew Shaw, Jeroen Keppens, Michael Luck, Simon Miles

Backmatter

Weitere Informationen

Premium Partner

    Bildnachweise