Doctoral Consortium Abstracts

The main aim of my thesis is the development of agents capable of reasoning about norms given that they are situated in an uncertain environment. The n-BDI agent architecture developed in my thesis is aimed at allowing agents to determine which and how norms will be obeyed and supporting agents when facing with norm violations. Reasoning About Norms Within Uncertain Environments Natalia Criadohas 2 papers

Agents usually encapsulate their principals' personal data attributes, which can be disclosed to other agents during agent interactions, producing a potential loss of privacy. We propose self-disclosure decision-making mechanisms for agents to decide whether disclosing personal data attributes to other agents is acceptable or not. Moreover, we also propose secure agent infrastructures to protect the information that agents decide to disclose from undesired accesses. Privacy and Self-disclosure in Multiagent Systems Jose M. Suchhas 2 papers

Software agents try to achieve the goals of roles that they have in an environment. It is supposed that the dynamic structure of role based agents can be connected with updatable domain ontologies of the environment. Ontology evolution can cause the update of agent behaviors or access restrictions to ontological elements. So regulation for the agent behaviors may be needed. Our motivation is to create a suitable policy model for agents, environments and organizations when ontologies in the environment can change. Policies for Role Based Agents in Environments with Changing Ontologies Fatih Tekbacak, Tugkan Tuglular, Oguz Dikenelli

This thesis investigates whether incorporating ideas from human decision-making in computer algorithms may help improve agents' decision-making performance, as either independent actors or in collaboration with humans. For independent actors, psychological cognitive appraisal theories of emotion are used to develop a lightweight algorithm that dynamically re-prioritizes their goals to direct their attention. In experiments in quickly changing and highly uncertain domains these agents are shown to perform as well as agents that compute expensive optimal solutions, and exhibit robustness with respect to the parameters of the environment. For agents interacting with humans, it is investigated whether expressing emotions has the ability to convey traits like trustworthiness and skill, and whether the appropriate emotional expression can help forge mutually beneficial relationships with the human. Finally, the theory of reasoning patterns is leveraged to analyze games and make it possible to answer questions about a system's strategic behavior without having to compute an expensive, precise solution. This theory is also employed to the generate advice for human decision-makers in complex games. This advice has been experimentally shown to improve their decision-making performance. Human Factors in Computer Decision-Making (PhD Thesis Extended Abstract) Dimitrios Antoshas 2 papers

Security of systems and information has always been a challenge to organisations and industries. Many technical solutions including firewalls, encryption and anti-virus software have been used, yet security still remains a problem. These security solutions failures are largely due to the fact that as systems become more complex, a lot of interaction is involved between various actors. Some of these interactions usually leave room for security vulnerabilities which are simply not accounted for by the technical security solutions: there are just too many possibilities. My research is focused on this aspect of organisational security. The proposed approach to this involves the monitoring of events for traces of behaviours that may eventually circumvent the security regulations of the organisation. The methodology includes organisational modeling and simulation of self monitoring agents using a normative framework. Security in the Context of Multi-Agent Systems Gideon D. Bibu

Agents have different interests and desires. Agents also hold different beliefs and assumptions. To accomplish tasks jointly, agents need to better convey information between each other and facilitate fair negotiations. In this thesis, we investigate agent dialogue systems developed with the Assumption-Based Argumentation (ABA) framework. Agents represent their beliefs and desires in ABA. Knowledge is exchanged via ABA arguments through dialogues. Main contributions include (1) understanding the connection between dialogues and argumentation frameworks and (2) applying argumentation dialogues in various agent applications. Agent Dialogues and Argumentation Xiuyi Fanhas 2 papers

Multi-agent path planning is a challenging problem with numerous real-life applications, including robotics, logistics, military operations planning, disaster rescue, and computer games. We look at navigating large numbers of mobile units to their targets on navigation graphs such as grid maps. The size of problems examined is significantly larger than can be handled using optimal multi-agent pathfinding algorithms in practice. We introduced Mapp, a tractable algorithm for multi-agent path planning on undirected graphs. Mapp and its extended versions are complete on well specified and tractably testable classes of problems. They have low-polynomial worst-case upper bounds for the running time, the memory requirements, and the length of solutions. Experiments on realistic game grid maps, with uniformly randomly generated start and target locations for each unit, show Mapp as a state-of-the-art multi-agent pathfinding algorithm in terms of scalability and success ratio (i.e., percentage of solved units). Even on challenging scenarios with 2000 units, Mapp solves 92% to 99.7% of units. Far and Whca, two fast but incomplete algorithms that were previously state-of-the-art in terms of scalability, solve as few as 17.5% and 12.3% of these problems. The quality of Mapp's solutions is empirically analyzed using multiple quality criteria: total travel distance, makespan, and sum of actions (including move and wait actions). Mapp is competitive in terms of solution quality and speed with Far and Whca*. Mapp further provides the formal characterizations that Far and Whca* lack, on problems it can solve as well as low-polynomial upper bounds on the resources required. As optimal algorithms have limited scalability, we evaluated the solution quality of suboptimal algorithms using lower bounds of optimal values. We showed that Mapp's solutions have a reasonable quality. For example, Mapp's total travel distance is on average 19% longer than a lower bound on the optimal value. Massively Multi-Agent Pathfinding made Tractable, Efficient, and with Completeness Guarantees Ko-Hsin Cindy Wang

Extensive transportation networks have become the economic backbone of the modern age. Thus, securing these networks against the increasing threat of terrorism is of vital importance. However, protecting critical infrastructure using limited security resources against intelligent adversaries in the presence of the uncertainty and complexities of the real-world is a major challenge. While game-theoretic approaches have been proposed for security domains, traditional methods cannot scale to realistic problem sizes (up to billions of action combinations), even in the absence of uncertainty. My thesis proposes new models and algorithms that have not only advanced the state of the art in game-theory, but have actually been successfully deployed in the real-world. For instance, IRIS has been in use by the Federal Air Marshal Service for scheduling officers on some international flights since October 2009. My thesis contributes to a very new area that uses insights from large-scale optimization for game-theoretic problems. It represents a successful transition from game-theoretic advancements to real-world applications that are already in use, and it has opened exciting new avenues to greatly expand the reach of game theory. Securing Networks Using Game Theory: Algorithms and Applications Manish Jainhas 3 papers

The aim of my PhD thesis is to propose a decentralized system for service management based on the social concept of homophily. The system provides self-organizing features, and it is established and maintained without supervision. Each agent manages autonomously events such as searching services, joining or leaving the system, which reduces the service management and the structure maintenance cost. Agents, considering only local information, carry out all these tasks. Decentralized Semantic Service Discovery based on Homophily for Self-Adaptive Service-Oriented MAS E. del Valhas 2 papers

Our main aim for this work is to provide a reorganization model which take into consideration the costs associated to the reorganization process. The reorganization model must provide the measurament of costs from the agent perspective (what does it cost the agent to play a new/other role) and from the organization perspective (what does it cost the organization to have an specific agent playing an specific role and how does it benefit from that). Furthermore, this measurement should be defined for static costs and also for dynamic costs. This reorganization model will allow us to reason about both reorganization dimensions: before and after reorganization. The first dimension is focused on measuring the effectiveness of the organization in the future and analyzing whether the organization will be able to cope with some changed circumstances. The second dimension is focused on measuring the impact of the problems appeared during the reorganization process in the cost of change. A Cost-Oriented Reorganization Reasoning for Multiagent Systems Organization Transitions Juan M. Alberolahas 2 papers

I introduce a graphical representation for modeling multi-agent systems based on different kinds of reasoning about agent behavior. I seek to investigate this graphical model's predictive and representative capabilities across various domains, and examine methods for learning the graphical structure from agent interaction data. I also propose to explore the framework's scalability in large real-world scenarios, such as social networks, and evaluate its prediction performance with existing network behavior models. Graphical Multiagent Models Quang Duong

Agents in Multi-Agent Systems depend on interactions with others to achieve their goals. Often, goals of agents conflict with each other, and agents can be unreliable or deceitful. Therefore, trust and reputation are key issues in this domain. As in human societies, software agents must interact with other agents in settings where there is the possibility that they can be exploited. This suggests the need for computational models of trust and reputation that can be used by software agents, and accordingly, much research has investigated this issue over the past decade. This thesis concentrates on two important questions, therefore it is divided in two parts. The first question is what sources agents can use to build their trust of others upon. The second question is how agents can use trust and reputation concepts to form stable coalitions. Extended Abstract of Elisabetta Erriquez Thesis Elisabetta Erriquezhas 2 papers

Game-tree search algorithms, such as the two-player Minimax algorithm and its multi-player counterpart, Max-n, are a fundamental component in the development of computer programs for playing extensive-form games. The success of these algorithms is limited by the underlying assumptions on which they are built. For example, it is traditionally assumed that deeper search always produces better decisions and also that search procedures can assume all players are selfish and ignore social orientations. Deviations from these assumptions can occur in real games and can affect the success of a traditional search algorithms. The goal of my thesis is to determine when such deviations occur and modify the search procedure to correct the errors that are introduced. Improving Game-tree Search by Incorporating Error Propagation and Social Orientations Brandon Wilsonhas 2 papers

In this paper, I present my ongoing research on agent-based negotiation teams. An agent-based negotiation team is a group of two or more agents with their own and possibly conflicting goals that join together as a single negotiation party because they share a common goal that is related to the negotiation. Our research goal is to provide agent-based solutions for problems which may need negotiation teams. Negotiation Teams in Multiagent Systems Víctor Sánchez-Anguixhas 2 papers

Game theory is a useful tool for reasoning about interactions between agents and in turn aiding in the decisions of those agents. In fact, Stackelberg games are natural models for many important applications such as oligopolistic markets and security domains. Indeed, Stackelberg games are at the heart of three deployed systems, ARMOR; IRIS; and GUARDS, for aiding security officials in making critical resource allocation decisions. In Stackelberg games, one player, the leader, commits to a strategy and follower makes her decision with knowledge of the leader's commitment. Existing algorithms for Stackelberg games efficiently find optimal solutions (leader strategy), however, they critically assume that the follower plays optimally. Unfortunately, in many applications, agents face human followers (adversaries) who - because of their bounded rationality and possibly limited information of the leader strategy - may deviate from their expected optimal response. Not considering these likely deviations when dealing with human adversaries may cause an unacceptable degradation in the leader's reward, particularly in security applications where these algorithms have seen deployment. To that end, I explore robust algorithms for agent interactions with human adversaries in security applications. I have developed a number of robust algorithms for a class of games known as “Security Games” and am working toward enhancing these approaches for a richer models of these games that I developed known as “Security Circumvention Games”. Real-World Security Games: Toward Addressing Human Decision-Making Uncertainty James Pitahas 2 papers

My research focuses on understanding and analyzing prediction markets using multi-agent system and game theory-based tools. Prediction markets, in which trading agents trade outcomes of events (like “Barack Obama will win the presidential election”), are similar to stock markets in that they aggregate the beliefs of many agents with different information about an event, and produce a single number, a market price. The main challenge in designing, deploying and operating prediction markets is the lack of understanding about what makes a successful prediction market, i.e.(1) how information affects traders and prediction market price, (2) how to incentivize traders to reveal their socially useful information, and (3) under what conditions do prediction markets perform the best. I address the first question by developing a multi-agent system that is used to analyze the effect of information on the prediction market performance and show that our model provides a better understanding and novel insights into the behavior of prediction markets and its participants. For the second question I propose a correlated equilibrium strategy for the trading agents within a partially observable stochastic game-based model that incentivizes traders to reveal their true beliefs and show that the market's participants following my strategy achieve higher profit by 35-127%. And for the third question I apply Boolean Network techniques to study the dynamics of prediction markets under various conditions and show under what parameter values the prediction market stabilizes and how its dynamics change with the introduction of noise and increased number of traders. A Multi-Agent System for Predicting Future Event Outcomes Janyl Jumadinovahas 2 papers

This thesis focuses on the design of autonomous agents which can negotiate with people using argumentation strategies. Argumentation is the ability to argue and to persuade another party to accept a desired agreement, to acquire or give information, to coordinate goals and actions and to find and verify evidence. Argumentation is endemic to human interaction. It facilitates knowledge about people's positions, and may improve the final outcome of negotiation. Despite the importance of argumentation within the general framework of negotiations, work on argumentation over the last few years has focused almost exclusively on the context of rational interactions between self-interested, automated agents. A Study of Computational and Human Strategies in Revelation Games Peled Noam

Modeling crowd behavior is an important challenge for agent-based simulation. My overall goal is to provide a single computational cognitive mechanism that, when executed by individual agents, would give rise to different crowd behaviors, depending on the perceptions and actions available to each individual. I propose a novel model of crowd behavior, based on Social Comparison Theory (SCT), a popular social psychology theory that has been continuously evolving. I am pursuing a concrete algorithmic framework for SCT and evaluating it on different social behaviors. Moreover, I have begun to explore the use of qualitative reasoning techniques to model global (macro-level) social phenomena in demonstrations. I believe that this is the first use of QR techniques for such purposes. Thesis Research Abstract: Modeling Crowd Behavior Based on Social Comparison Theory Natalie Fridmanhas 2 papers

We study how to achieve cooperation between two self-interested agents that play repeated randomly generated normal form games. We take inspiration from a model originally designed to identify cooperative actions by humans who play a game, but we use the model in a prescriptive rather than descriptive manner. To identify cooperative intent, agents use a particle filter to learn the parameters of the model. Cooperation between Self-Interested Agents in Normal Form Games Steven Damerhas 2 papers

In Multiagent Systems (MAS), various activities are related to decisions involving a group of agents such as negotiation, auctions and social choice. Group Decision Making (GDM) specializes in the situations where a group of agents need to pick one of possibly many options from a set and commit to it. We intend to provide a novel GDM framework in which the agents are able to employ abductive reasoning and discuss the options towards consensus. Group Decision Making in Multiagent Systems with Abduction Samy Sáhas 2 papers

To optimally secure large and complex infrastructures against crime activities, a scalable model for optimal defender allocation is needed. Game theory is successfully used to formalize the problem as a two-player game between an attacker and a defender. We consider both player to be mobile and we focus on proper path intersection modeling and we observe the trade-off between fidelity and computational complexity. We search for the a Nash Equlibirium of the game using oracle based algorithms and we evaluate the robustness of the solution in a multi-agent simulation where some assumptions made do not strictly hold. Security Games with Mobile Patrollers Ondřej Vaněkhas 4 papers

My research on normative systems is focused on the spreading and acceptance of social norms, defined as norm support. Our understanding of norm support deals with the problem of which norm is established as the dominant. Specifically, we deal with two different branches of the research on normative sytems: conventional norms and essential norms. On the one hand conventional norms fix one norm amongst a set of norms that are equally efficient as long as every member of the population uses the same (e.g. communication protocols, greetings, driving side of the road), and on the other hand, essential norms solve or ease collective action problems, where there is a conflict between the individual and the collective interests. The scientific question of this research is how to accelerate the establishment of a common norm in virtual societies: in the case of conventional norms, by dissolving the subconventions; and in the case of essential norms, by studying the effects of punishment and norm internalization. Self-Organization in Decentralized Agent Societies through Social Norms Daniel Villatorohas 2 papers

My thesis will contribute to the field of multi-agent systems by proposing a novel and formal trust-based decision model for supply chain management. A Trust Model for Supply Chain Management Yasaman Haghpanah