1. 1 - Strategic Voting and Strategic Candidacy in Multi-Agent Systems, 2h long

    Maria Polukarov AIC group, ECS, University of Southampton
    Svetlana Obraztsova, Tel Aviv University
    Zinovi Rabinovich, MobilEye Vision Technologies Ltd.

  2. Abstract:
    Multi-agent decision problems, in which independent agents have to agree on a joint plan of action or allocation of resources, are central to various applications. Often in such settings, agents' individual preferences over available alternatives may vary, and they may try to reconcile these differences by voting. Based on the fact that agents may have incentives to vote strategically and misreport their real preferences, much of the literature in computational social choice focuses on evaluating voting rules by their resistance to strategic behaviours and uses computational complexity as a barrier to them.

    In contrast, more recent works (counting from 2010) take another natural approach and analyse voting scenarios from a game-theoretic perspective, viewing strategic parties as players and examining possible stable outcomes of their interaction (i.e., equilibria).

    Finally, the candidates themselves may also have preferences about the outcome and try to affect it by strategically choosing whether to stand for election or not.
    This tutorial will begin by introducing the audience to basic notions of social choice and game theory, and will lead them to an understanding of strategic behaviours by voters and candidates, modelling such scenarios as voting/candidacy games, and analysing the existence of stable game outcomes (including those with predetermined properties), as well as their reachablity by natural iterative processes, such as best-response dynamics or its restricted variants. Convergence of such procedures is a highly desirable property of the game, since, from a system-wide perspective, it implies that a system has a deterministic stable state that can be reached by the agents without any centrallised control and/or communication.

    The tutorial will give an overview of the existing results and present a number of open problems in this - fairly new - line of research.

  3. 2 - Complex Event Recognition in Multi-Agent Systems, 2 h long 

    Alexander Artikis, University of Piraeus
    Matthias Weidlich, Imperial College London

    The concept of event processing is established as a generic computational paradigm in various application fields, ranging from data processing in Web environments, over maritime and transport, to finance and medicine. Events report on state changes of a system and its environment. Complex event recognition, in turn, refers to the detection of events that are considered relevant for processing, thereby providing the opportunity for reactive measures. Examples include the recognition of attacks in computer network nodes, and fraud in electronic marketplaces. In each scenario, complex event recognition allows to make sense of very large data streams and react accordingly. In this tutorial, we will present how complex event recognition techniques may be used for detecting activities of special significance, such as violations of the rules of the game, in multi-agent systems.

    Moreover, we will show how probabilistic event recognition supports robust inference in the presence of noisy agent communication channels and imperfect sensors, and how machine learning techniques optimise reasoning, allowing for event recognition in very large multi-agent systems.

  4. 3 - Decentralized Multiagent Systems, 4h long 

    Amit K. Chopra, Lancaster University
    Munindar P. Singh, NC State University


    Modern IT applications involve multiple people and organizations who interact to exchange information and services. For example, healthcare applications involve patients, physicians, nurses, hospitals, pharmacies, and so on. Today such applications are conceptualized and designed as a single entity system, e.g., as a set of Web services.

    This tutorial will introduce decentralized multiagent systems (MAS) as a way of realizing modern IT applications. Specifically, the tutorial will introduce interaction protocols as the basic building block for such applications. It will describe existing software engineering and artificial intelligence approaches and standards, including UML Sequence Diagrams, FIPA ACL, and multiagent systems programming.

     The tutorial will then introduce topics that specifically address decentralized MAS from the perspective of information modeling and asynchronous enactments, specifically, information-based protocols and commitment protocols. It will also introduce various correctness criteria relevant to such specifications.

     The tutorial will illustrate the concepts involved with a variety of examples, software demonstrations, and problem sets to be solved by the attendees during the tutorial.

     Attendees will appreciate the theoretical foundations of decentralized MAS and learn how to apply interactions protocols toward engineering decentralized MAS. Parts of this tutorial have already featured in courses for senior undergraduate and beginning graduate students.

    Interaction protocols and commitments have been a major theme of multiagent systems (MAS) research for the past decade and a half. This tutorial will cover the latest advances on these topics, especially from the perspective of information modeling (think databases) and decentralized enactments (think messaging and asynchrony). Understanding these perspectives is crucial to designing practical and correct MAS.

     In this tutorial, we will present the challenges of building practical and correct MAS with an emphasis on decentralized settings. We will provide a quick review of UML Sequence Diagrams (SDs) and show how it is easy to build erroneous SDs. We will describe the criteria for correct interactions and provide criteria for judging if an SD is acceptable. We will show why control flow is an inappropriate conception in decentralized systems.

    We will introduce an information-based conception of interactions. We will first introduce the Blindingly Simple Protocol Language (BSPL), and show how it addresses the above challenges. We will next introduce commitments and show the shortcomings of traditional formalizations of commitment protocols. We will show how commitments may be specified in Cupid, an information-based representation. We will describe how Cupid commitments can be layered on top of BSPL.

    We will illustrate the concepts involved with a variety of examples, software demonstrations, and problems sets to be solved by the attendees during the tutorial.
    After attending this tutorial, attendees will appreciate the theoretical foundations of interaction protocols and commitments, how they may be supported in software, and how one may apply them in their toward engineering decentralized MAS. Parts of this tutorial have already featured in courses for senior undergraduate and beginning graduate students.

  5. 4 - Norm Synthesis in Normative Multi-Agent System, 2h long

    Maite Lopez-Sanchez, University of Barcelona

    Multi-Agent Systems (MAS) design endows restricting agents' behaviour to meet system requirements whilst guaranteing agents' autonomy. Normative Multi-Agent Systems (NMAS) include norms as coordination mechanisms to accomplish such design requirements. Within NMAS, this tutorial is devoted to introduce the norm synthesis problem and different approaches that tackle it. Moreover, students will be able to follow a hands-on activity to see in detail the application of an on-line automatic norm synthesis process.

    In order to do so, we encourage attendees to bring their own laptop with SW at preinstalled.

  6. 5 - Principles of Automated Negotiation, 4h long

    Shaheen Fatima, Loughborough University
    Sarit Kraus, Bar-Ilan University and University of Maryland
    Michael Wooldridge, University of Oxford

    With an increasing number of applications in the context of multi-agent systems, automated negotiation is a rapidly growing area. In this tutorial, we will explore key issues involved in the design of negotiating agents, covering strategic, heuristic, and axiomatic approaches.

    We will also discuss the potential benefits of automated negotiation as well as the unique challenges it poses for computer scientists and for researchers in artificial intelligence.

    Finally, we will consider possible applications and give the audience a feel for the types of domains where automated negotiation is already being deployed.
    This tutorial will be presented by the authors of the new book titled Principles of Automated Negotiation. The book was published in November 2014 by Cambridge University Press.

  7. 6 - Multi-Agent Oriented Programming, 4h long

    Olivier BOISSIER G2I - Ecole Nationale Supérieure Mines
    Jomi HUBNER, Federal University of Santa Catarina
    Alessandro RICCI, University of Bologna
    Jaime Simão SICHMAN, University of São Paulo

    We propose an advanced tutorial for students with some familiarity on programming multi-agent systems, focusing on three complementary dimensions: agent-oriented programming, organisation-oriented programming, and environment-oriented programming. The tutorial will combine state-of-the-art technologies and platforms for each of these levels of multi-agent programming, stressing on their integration. In particular, we will illustrate our approach by using the JaCaMo framework, which integrates three separate technologies:

    Jason, Moise, and Cartago. These technologies are quite well known in the MAS community, and since they have already been developed for a certain number of years, they are fairly robust and fully-fledged. They will be used as examples of platforms for programming each MAS dimension. The tutorial will introduce the relevant techniques to use and integrate these technologies, providing an integrated scenario that will illustrate their usage and integration.

  8. 7 - Modeling and Simulation using Agents in Cell-Spaces,  2h long

    Gabriel A. Wainer, Carleton University

    In recent years, new methodologies have allowed building agent-based simulation software executing on grid-shaped cell spaces. There have been numerous efforts integrating agents and cellular models for simulation. The main issue with these methods is that they can require large amounts of compute time. We will discuss how to deal with this issue using a formal modeling technique that permits defining each cell in a cell space as individual independent entity, called Cell-DEVS. The goal of Cell-DEVS is to build discrete-event cell spaces, improving their definition by making the timing specification more expressive and the definition of complex models simpler. We will show how to improve model definition through Cell-DEVS, focusing on varied examples of application, and discussing open research issues in this area. We will start with an introduction to different agent models in urban traffic, sensor networks, and mobile communications. We will then show some examples of the current use of the methodology in natural systems: ant foraging systems, synapsin-vesicle interaction in nerve terminals, forest fire suppression by firefighters, etc. We will focus in showing how the application of these techniques can improve model definition. We will also focus in describing how to create models that can be executed automatically in a parallel environment without any modifications to the original models, or user intervention.

  9. 8 - Truth-Revealing Social Choice, 2h long

    Toby Walsh, University of New South Wales
    Lirong Xia, Rensselaer Polytechnic Institute
    Leandro Soriano Marcolino, University of Southern California

    Social choice theory studies representation and aggregation of individual preferences.

    It dates back to the 4th century B.C., and prospered since the 18th century. In axiomatic social choice theory, normative properties (called axioms) are used to evaluate and compare various social choice mechanisms.

    The focus of this tutorial is on another major branch of social choice theory, namely truth-revealing social choice theory, which was initiated by the celebrated Condorcet Jury Theorem. It justifies the wisdom of the crowds and plays an important role in political science and economics. In the past few decades, rapid developments in computer science and Internet technologies have brought fresh air to social choice by introducing not only many new applications, but also a computational point of view on preference representation and aggregation. In particular, truth-revealing social choice has found its place in many multi-agent system applications and e-commerce applications. Where the agents have conflicting preferences but they must make a joint decision to reveal the ground truth.

    In this tutorial, we will start with a brief introduction to social choice, including axiomatic social choice and commonly studied mechanisms. Then, we will discuss in depth the Condorcet Jury Theorem and its numerous extensions. Finally, we will give a brief overview of applications of social choice in the computer science world and make connections to machine learning especially ensemble algorithms and learning to rank.