Michael Rovatsos

Multiagent systems and social computation.

''For the past 15 years, my research has been in multiagent systems, where most of my contributions have been in agent communication languages, multiagent planning and learning, trust and reputation, social reasoning, argumentation, and norms. Most of this work has focused on reasoning about interaction, more specifically the development of adaptive and scalable knowledge-based methods for reaching agreement, resolving conflict, and sharing knowledge among agents. My work uses mainly methods from symbolic AI such as automated planning, deductive reasoning and declarative approaches to systems modelling, but it has also involved such techniques as reinforcement learning and statistical reasoning. While I am interested in building rigorous models of intelligent reasoning and decision making, my research emphasises the importance of producing systems that work in practice.

Recently, I have become more interested in hybrid man-machine collective intelligence, with two focal areas: social computation, and the evolution of meaning. In social computation, I am interested in how large collectives of collaborating humans and software agents can solve problems on the Web that are too complex for humans or machines to solve by themselves. In research on the evolution of meaning, I am investigating how artificial agents can benefit from the mechanisms humans use in negotiating the meaning of what they communicate with each other in order to improve the interoperability of distributed systems, but also to understand the social foundation of collective intelligence''

Pavlos Andreadis

Computational Nudging or How to alter a Human Agent's (HA) policy by controlling the information presented.

Utilizing a HA's learned preference model, we can guide his choice by adjusting the representation of alternatives in the feature space. Moreover, the presented alternatives can alter his perception of the feasibility of a future incentive, thus controlling its effectiveness.

Interests: Incentives, behavioural modelling, recommender systems, human Behaviour modelling (learning preferences, designing Incentives, and nudging)

Michael Anslow

Semantic Alignment of Heterogeneous Ontologies Induced from Sensor Data

My research is concerned with ontology alignment between heterogeneous ontologies where each ontology is induced independently from sensor measurements. We consider each disparate sensor/sensor cluster as an agent and all agents as a collaborative multi-agent system. As ontologies model a particular conceptualisation of a domain, for agents to exchange information about domain they must align ontologies. We are interested in the case that sharing domain knowledge improves a joint task performed by all agents and evidence for alignment takes the form of ontology samples exchanged between agents as messages in a coordinated communication game.

Interests: Language Games, Ontology Grounding, Ontology Matching, Ontology Negotiation

Dimitris Diochnos

Compositionality and social orchestration in the Smart Society project and the Smart Sharing application

Dimitris worked for his Ph D? on the analysis of algorithms in different variants of the Probably Approximately Correct model of learning (evolvability, multiple-instance learning, active learning), as well as on question answering with the aid of the commonsense knowledge base Concept Net? 4. Dimitris is currently working under the workpackage for compositionality and social orchestration in the Smart Society project and develops in parallel the relevant backend functionality of the server for the prototype of Smart Sharing.

Sergio Elizondo Gonzalez

Agent-based Computational Economics of ‘Prosumer’ Organisations in Microgrids.

''A Microgrid is a subsystem of a ‘Smarter’ electricity grid that integrates distributed energy resources (DER) and ICT in a localised distribution network. The dynamics of this system includes consumers who also produce electricity by their DER assets; these actors are known as ‘prosumers’ in the literature. Their assets can be distributed generation technologies (e.g., solar panels and wind turbines), domestic batteries and controllable loads, all of them managed by ‘smart’ controllers. This new dynamics of how electricity is to be produced and consumed poses some interesting challenges to balancing supply and demand in this complex dynamic system.

Therefore, organisations or clusters of prosumers can play an important role on coordinating many micro-resources in such a way it could be beneficial for a Microgrid so as to match supply with demand. The smart controllers are modelled as agents; they can compete or cooperate in a time-varying price electricity market. The main questions addressed are: (1) What are the determinants of organisational behaviour and performance given economic incentives? (2) How the prosumers’ skills matter to these organisations? And (3) how these organisations evolve? ''

Amy Guy

Content Creators and the Semantic Web

Behind the plethora of user-generated media content on the web are a multitude of content creators, actively pushing their creative works to a global audience. The activities, needs and desires of these content creators transcend the individual websites and communities they use. In order to answer questions about identity behaviours and social network dynamics on a large scale, as well as to build applications that can facilitate collaboration between content creators, I am working to incorporate these creators and their works into the linked data cloud.

Interests: semantic web, user-generated content, content creators, linked data, linked media, social network analysis, social machines.

Tania Marques

Communication Planning

Humans are not sole entities but social beings that can accomplish more working together than by themselves. This is only possible because humans are able to communicate and agree on what to do together. In the "Communication Planning" project (WP 3?/ESSENCE), we are interested in modelling this behaviour for autonomous and independent agents in a way that they will be able to communicate with a different agent or human without knowing their internal structure. In a few words, given a complex task environment, what kind of planning activity should an agent do to decide what should it say to whom in order to accomplish its goals.

Agents Group


edit SideBar

Blix theme adapted by David Gilbert, powered by PmWiki