Archive for December 19th, 2014

Cognitive agents platform to model complex social behaviour

Research Themes: Agent Programming / Programming languages / Agent-Based Modeling / Concurrent computation

PROBLEM DESCRIPTION

Current cognitive agents programming frameworks (ex. AgentSpeak/Jason, 2APL, Goal, etc.) mostly focus on the logical aspects associated to intentional entities, rather then the computational issues derived from concurrent execution. However, in principle, agents are entities that act concurrently in an environment (shared resources and infrastructures). Their internal cognitive processes may be concurrent as well. In the last months, few prototypes of these platforms have been built on top of functional programming languages, showing an enormous difference in efficiency, compared to the most known implementations, mostly implemented in Java.

On the other side, agent-based modelling frameworks like Repast, AgentScape, etc. refer to models of agents provided with basic rationality, usually not involving concepts like Beliefs-Desires-Intentions. Although there have been proposals in introducing these extensions, results seem to be not satisfactory, probably because of the high overload of these platforms. Furthermore, the configuration and exploitation of such applications still require skills that are not usual for non-IT researcher/practitioners, thus hampering specification by domain experts. Visual programming techniques may probably help in this respect.

Our project aims to develop a full framework for the acquisition, execution and exploitation of social institutional scenarios involving cognitive agents. The agents we are considering are fully non-reflective intentional agents. Their behaviour is deterministically defined according to a script, usually extrated from scenarios provided by experts. Conversely, the behaviour resulting from the interaction between several agents is not deterministic.

The underlying model we focus on is agent-role, which add intentionality to the role abstraction. The components of the agent-role model should be sufficient for a complete first-order theory of mind.

Given the complexity of the quest, various abstractions are possible, addressing different aspects of the problem. There is room for 3-4 students, with separate objectives and theses, but elaborating on the same conceptual framework. This is a not exhaustive list of topics concerning the research:

  • definition of agent-role model components:
    • intents, beliefs, reactive plans, maintenance goals, desires, etc.
    • speech acts: assertions, directives (questions, commands), commissives (not included in FIPA),
  • formal semantics and verification for an agent-role programming language,
  • transformation from visual models (e.g. Message Sequences Diagrams, adequately enriched with intentional components) to visual computational models (e.g. Petri Nets), and then in agent-role scripts,
  • development of an engine executing several concurrent agent-role scripts (e.g. to be run on a server),
  • development of an engine executing a defined scenario (e.g. to be run on a web-based client),
  • conception and development of user interface for an IDE tailored on agent and scenario modeling.

CONTACT PERSON(S)

Alexander Boer, a.w.f.boer@uva.nl, Leibniz Center for Law
Giovanni Sileno, g.sileno@uva.nl, Leibniz Center for Law

RELEVANT LITERATURE

  • Sileno, G., Boer, A., & van Engers, T. (2014). From Inter-Agent to Intra-Agent Representations: Mapping Social Scenarios to Agent-Role Descriptions. In Proceedings 6th International Conference on Agents and Artificial Intelligence (ICAART 2014).
  • Boer, A., & Engers, T. (2013). Agile: a problem-based model of regulatory policy making. Artificial Intelligence and Law, 21(4), 399–423.
  • D. Harel and R. Marelly. Specifying and executing behavioral requirements: The play-in/play-out approach. Software and Systems Modeling, 82–107, 2003.
  • Cohen, P. R., & Levesque, H. J. (1990). Intention is choice with commitment. Artificial Intelligence, 42(2-3), 213–261.
  • Rao, A. S. (1996). AgentSpeak (L): BDI agents speak out in a logical computable language. In Proc. of 7th European Workshop on Modelling Autonomous Agents in a Multi-Agent World.
  • Diaz, Á. F., Earle, C. B., & Fredlund, L. (2012). eJason: an implementation of Jason in Erlang. Tenth International Workshop on Programming Multi-Agent Systems – PROMAS 2012.
  • Harel, D., & Rumpe, B. (2000). Modeling Languages: Syntax, Semantics and All That Stuff, 1–28.
  • Best, E., Devillers, R., & Koutny, M. (1998). Petri nets, process algebras and concurrent programming languages. Lectures on Petri Nets II: Applications.

Add comment December 19th, 2014


Calendar

December 2014
M T W T F S S
« Jul   Apr »
1234567
891011121314
15161718192021
22232425262728
293031  

Posts by Month

Posts by Category