Julien Saunier

Regulated multi-party communications and context awareness through the environment

Recent research in the multi-agent systems field has highlighted the relevance of complex interaction models such as multi-party communication and context awareness. Nevertheless, there are no generic interaction model and infrastructure that enable to apply them in a standardized way. Emerging as a first-order abstraction, the environment, in the sense of a common medium for the agents, is a suitable paradigm to support these new interaction models. We present an operational model called Environment as Active Support of Interaction, that enables each agent to actively modify the environment according to its interaction needs. This model provides a suitable framework for the regulation of MAS interactions, and priority policies are given to manage the rules. An algorithm is proposed and assessed with an example stemming from the ambient intelligence domain.

Environment as active support of interaction

Indirect interactions have been shown to be of interest in MultiAgent Systems (MAS), in the simulation area as well as in real applications. The environment is also emerging as a first-order abstraction. Intuitively, the environment being a common medium for the agents, it should be a suitable paradigm to provide an infrastructure for both direct and indirect interactions. However, it still lacks of a consensus on how the two relate to each other, and how the environment can support effectively notions as communication or awareness. We propose a general and operational model, Environment as Active Support of Interaction, that enables the agents to actively participate in the definition of their perceptions. Then, we show how the model provides a suitable framework for the regulation of the MAS interactions.

A Formal Model of Communication and Context Awareness in Multiagent Systems

Awareness is a concept that has been frequently studied in the context of Computer Supported Cooperative Work. However, other fields of computer science can benefit from this concept. Recent research in the multi-agent systems field has highlighted the relevance of complex interaction models such as multi-party communication and context awareness for simulation and adaptive systems. In this article, we present a generic interaction model that enables to use these different models in a standardized way. Emerging as a first-order abstraction, the environment, in the sense of a common medium for the agents, is a suitable paradigm to support the agents’ awareness. We present an operational model, called Environment as Active Support of Interaction, to take into account all the agents that can be interested in a communication. This model is then extended for the regulation of multiagent systems interactions. Priority policies are given to manage the rules governing the context (un-)awareness of the agents. We also present a new AUML connector to create protocols that take into account the agent awareness to implement proactive behaviour, and several communication scenarios are proposed to show practical applications of this model.

Detection of Multiple Identity Manipulation in Collaborative Projects

Various techniques are used to manipulate users in OSN environments such as social spam, identity theft, spear phishing and Sybil attacks... In this article, we are interested in analyzing the behavior of multiple fake accounts that try to bypass the OSN regulation. In the context of social media manipulation detection, we focus on the special case of multiple Identity accounts (Sockpuppet) created on English Wikipedia (EnWiki). We set up a complete methodology spanning from the data extraction from EnWiki to the training and testing of our selected data using several machine learning algorithms. In our methodology we propose a set of features that grows on previous literature to use in automatic data analysis in order to detect the Sockpuppets accounts created on EnWiki. We apply them on a database of 10.000 user accounts. The results compare several machine learning algorithms to show that our new features and training data enable to detect 99\% of fake accounts, improving previous results from the literature.

An Environment to Support Multi-Party Communications in Multi-Agent Systems

Two-party communication is the most-studied model to support interaction between two cognitive agents, whereas that is only one case of what an agent should be able to do. Multi-party communications enhance this model, by taking into account all the roles an agent can have in a communication. Nevertheless, there are no generic models and infrastructures that enable to apply multi-party communication in a standardized way. We emphasize that the environment, in the sense of a common medium for the agents, is a suitable paradigm to support multi-party communication. We propose a general and operational model called Environment as Active Support of Interaction (EASI), that enables each agent to actively modify the environment according to its communication needs. Algorithms are proposed and assessed with an example stemming from the ambient intelligence domain.

Agents Behavior Semi-automatic Analysis through Their Comparison to Human Behavior Clustering

This paper presents a generic method to evaluate virtual agents that aim at reproducing humans behaviors in an immersive virtual environment. We first use automated clustering of simulation logs to extract humans behaviors. We then propose an aggregation of the agents logs into those clusters to analyze the credibility of agents behaviors in terms of capacities, lacks, and errors by comparing them to humans ones. We complete this analysis with a subjective evaluation based on a questionnaire filled by human annotators to draw categories of users, making their behaviors explicit. We illustrate this method in the context of immersive driving simulation.

Fuzzy Rules for Events Perception and Emotions in an Agent Architecture

In complex simulations, multi-agents systems allow to model virtual humans with an explicit cognitive process representation. However, this cognitive process is hard to model and is therefore generally simplified in an application-dependant way. In order to improve the realism of individual and collective behavior of these agents, we propose to integrate the perception of events and the computation of agents emotions in a fuzzy framework. The modeling of the perception and its effect on emotions through fuzzy rules enables the agents to consider properly the virtual environment. We show how different kinds of fuzzy rules can help in the calculus of emotions. Computation of emotions is based on the evaluation of events’ occurrence. Once the events are perceived by the agents, our method uses the desirability of these events to compute emotions relevant to crisis situations. We illustrate this model with a traffic simulation example.

SocksCatch: Automatic detection and grouping of sockpuppets in social media

Abstract Since 2004, online social media (OSN) have evolved hugely. This fast development had interesting effects to increase the connection and information exchange of users, but some negative effects also appeared, including fake accounts number growing day after day. The sockpuppets are the multiple fake accounts created by the same user. They are the source of several types of manipulation such as those created to praise, defend or support a person or organization or to manipulate public opinion. In this article, we present SocksCatch, a complete process to detect and group sockpuppets which is composed of three main phases: first phase is the data collection and selection; second phase is the detection of the sockpuppet accounts using machine learning algorithms; third phase is the grouping of sockpuppet accounts created by the same user using graph theory. Experiments have been performed for the three phases using real data crawled from english Wikipedia. The results compare six machine learning algorithms for the detection phase and show that SocksCatch detects between 89% and 95% of the selected sockpuppets depending on the algorithms. We also compare five community detection algorithms for the grouping phase, and show that SocksCatch’s grouped sockpuppets and the real sockpuppet’s groups are similar between 80% and 88%, according to the cluster’s comparison measures: normalized variation of information (NVI) and normalized mutual information (NMI).

Designing adaptable virtual reality learning environments

The EAST (Scientific and technical learning environments) project aims at stimulating the interest of young people for science through virtual reality environments, based on industrial assets. Although training and learning environments are classical applications of virtual reality, the design of these environments is generally ad hoc, hence requiring the intervention of programmers whenever a modification of the pedagogical scenario is required. In this paper, we propose a methodology to design virtual environments which can be adapted by teachers to implement different scenarios according to the level of the trainees and to the pedagogical objectives. Current demonstrators include a windmill with three different learning situations: simulator, safety training and preventive maintenance training.

Agent Bodies: An Interface Between Agent and Environment

Interfacing the agents with their environment is a classical problem when designing multiagent systems. However, the models pertaining to this interface generally choose to either embed it in the agents, or in the environment. In this position paper, we propose to highlight the role of agent bodies as primary components of the multiagent system design. We propose a tentative definition of an agent body, and discuss its responsibilities in terms of MAS components. The agent body takes from both agent and environment: low-level agent mechanisms such as perception and influences are treated locally in the agent bodies. These mechanism participate in the cognitive process, but are not driven by symbol manipulation. Furthermore, it allows to define several bodies for one mind, either to simulate different capabilities, or to interact in the different environments - physical, social- the agent is immersed in. We also draw the main challenges to apply this concept effectively.