M.P. Singh - Agent Communication Languages: Rethinking the Principles Alessandro Giusti March, 28 2006.

Slides:



Advertisements
Similar presentations
Public versus Patient Representation Rosemary Chesson Health Services Research Group Faculty of Health & Social Care The Robert Gordon University.
Advertisements

Meditation IV God is not a Deceiver, Truth Criterion & Problem of Error.
Intelligent Architectures for Electronic Commerce Part 1.5: Symbolic Reasoning Agents.
Workpackage 2: Norms
Chapter 16: Multiagent Systems Service-Oriented Computing: Semantics, Processes, Agents – Munindar P. Singh and Michael N. Huhns, Wiley, 2005.
Please check. Announcements 1.Don't forget your plagiarism certificate next week. You must turn that in in order to stay enrolled in the class. 2.The.
Communities of Practice: An Introduction for Technical Communication Tracy Bridgeford, University of Omaha Communities of Practice Definition.
Chapter 18: Communication Service-Oriented Computing: Semantics, Processes, Agents – Munindar P. Singh and Michael N. Huhns, Wiley, 2005.
Week #7: Conversational Implicature and Explicature A Follow-up from Previous Presentation and Discussion by Students.
Understanding the Research Process
Argumentation-based negotiation Rahwan, Ramchurn, Jennings, McBurney, Parsons and Sonenberg, 2004 Presented by Jean-Paul Calbimonte.
Presented by: Thabet Kacem Spring Outline Contributions Introduction Proposed Approach Related Work Reconception of ADLs XTEAM Tool Chain Discussion.
OASIS Reference Model for Service Oriented Architecture 1.0
Specifying Agent Interaction Protocols with AUML and OCL COSC 6341 Project Presentation Alexei Lapouchnian November 29, 2000.
Adding Organizations and Roles as Primitives to the JADE Framework NORMAS’08 Normative Multi Agent Systems, Matteo Baldoni 1, Valerio Genovese 1, Roberto.
Mediation as a Source of Law Dale Dewhurst Athabasca University New York – IALMH, 2009.
Culture & Management Definitions of culture Theoretical frameworks of culture How culture affects management.
Multiagent Systems and Societies of Agents
Presentation on Formalising Speech Acts (Course: Formal Logic)
April 15, 2005Department of Computer Science, BYU Agent-Oriented Software Engineering Muhammed Al-Muhammed Brigham Young University Supported in part by.
Meaningful Learning in an Information Age
Ontology translation: two approaches Xiangkui Yao OntoMorph: A Translation System for Symbolic Knowledge By: Hans Chalupsky Ontology Translation on the.
ACOS 2010 Standards of Mathematical Practice
Zinovy Diskin and Juergen Dingel Queen’s University Kingston, Ontario, Canada Mappings, maps and tables: Towards formal semantics for associations in UML.
Part I: The Sociological Perspective.  Sociology is the scientific study of social structure, examining human social behavior from a group, rather than.
Chapter 6: Objections to the Physical Symbol System Hypothesis.
CS 478 – Tools for Machine Learning and Data Mining The Need for and Role of Bias.
General Theories of Classroom Management
Chapter 15: Agents Service-Oriented Computing: Semantics, Processes, Agents – Munindar P. Singh and Michael N. Huhns, Wiley, 2005.
WHAT IS QUALITATIVE INQUIRY? Martyn Hammersley The Open University NCRM Research Methods Festival, St Catherine’s College, Oxford, July 2010.
What is linguistics  It is the science of language.  Linguistics is the systematic study of language.  The field of linguistics is concerned with the.
Alignment of ATL and QVT © 2006 ATLAS Nantes Alignment of ATL and QVT Ivan Kurtev ATLAS group, INRIA & University of Nantes, France
Travis Steel. Objectives What is the Agent Paradigm? What is Agent-Oriented Design and how is it different than OO? When to apply AOD techniques? When.
Effective Requirements Management – an overview Kristian Persson Field Product Manager, Telelogic Asia/Pacific.
Rubrics Mean – 42 Some common problems ◦Presentation Errors ◦Relevancy – not changing the template to match assignment ◦Including too much into 1 criteria.
EEL 5937 Agent communication EEL 5937 Multi Agent Systems Lecture 10, Feb. 6, 2003 Lotzi Bölöni.
Chapter 8 Language & Thinking
Chapter 18: Communication Service-Oriented Computing: Semantics, Processes, Agents – Munindar P. Singh and Michael N. Huhns, Wiley, 2005.
EEL 5937 Models of agents based on intentional logic EEL 5937 Multi Agent Systems.
Intelligent agents, ontologies, simulation and environments for norm-regulated MAS Deliberative Normative Agents Ricardo Gralhoz Governance in Open Multi-Agent.
Thoughts Before Requirements Gathering. Requirements Gathering Functional Requirements – Functional requirements explain what has to be done by identifying.
“D.A.I. & S.M. for KM” a synergy of complementary domains and challenges  the semantic web addicted people “please, raise your hands !”
Managing multiple client systems and building a shared interoperability vision in the Health Sector Dennis Wollersheim Health Information Management.
Christoph F. Eick University of Houston Organization 1. What are Ontologies? 2. What are they good for? 3. Ontologies and.
Multiagent System Katia P. Sycara 일반대학원 GE 랩 성연식.
EEL 5937 Agent communication EEL 5937 Multi Agent Systems Lotzi Bölöni.
WHAT IS LANGUAGE? HUMAN AND ANIMAL LANGUAGES. HUMAN AND ANIMAL LANGUAGES The issue of continuity Are humans just a step further in practising an adapted.
Chapter Two: Ethical Relativism Ethical Relativism holds that there are no objective moral principles, but that such principles are human inventions.
What’s Ahead for Embedded Software? (Wed) Gilsoo Kim
Agent Communication Languages Speech act theory Speech act theory Semantics of languages Semantics of languages KQML KQML FIPA ACL FIPA ACL Comparison.
COM 340 Lecture 2 Perception, Meaning, Reality, and Axioms of ICC.
1 September 2003 Brokering Services A Transformation-Based Perspective Stephen Potter & Marco Schorlemmer
EEL 5937 Content languages EEL 5937 Multi Agent Systems Lecture 10, Feb. 6, 2003 Lotzi Bölöni.
Chapter 6 Guidelines for Modelling. 1. The Modelling Process 1. Modelling as a Transformation Process 2. Basic Modelling Activities 3. Types of Modelling.
A UML-Based Pattern Specification Technique Presented by Chin-Yi Tsai IEEE TRANSACTION ON SOFTWARE ENGINEERING, VOL. 30, NO. 3, MARCH 2004 Robert B. France,
English 100 Tuesday, and Wednesday, On a sheet of paper, write about the following prompt… you will keep this in your notebook: Education.
FACULTY OF LAW, UNIVERSITY OF OSLO The principle of integration and its dilemmas Hans Chr. Bugge Professor of Environmental Law University of Oslo.
Communication between cultures 8TH EDITION Chapter 8 Verbal Messages: Exchanging Ideas Through Language © Cengage 20121Chapter 8 Verbal Messages: Exchanging.
PRESUPPOSITION PRESENTED BY: SUHAEMI.
Service-Oriented Computing: Semantics, Processes, Agents
Service-Oriented Computing: Semantics, Processes, Agents
LECTURE 9: Agent Communication
Service-Oriented Computing: Semantics, Processes, Agents
Service-Oriented Computing: Semantics, Processes, Agents
Service-Oriented Computing: Semantics, Processes, Agents
A Semantics for the Interaction of Agents with Coordination Artifacts
Service-Oriented Computing: Semantics, Processes, Agents
CONSTRUCTIVISM Submitted To: Ma’am Misbah Yasmeen BPGCW (Air University)
Zinovy Diskin School of Computing, Queen’s University Kingston, Canada
Presentation transcript:

M.P. Singh - Agent Communication Languages: Rethinking the Principles Alessandro Giusti March,

Agent Communication Languages Allow agents to communicate Interoperability (key feature) Other key agent features Autonomy Heterogeneity Sony Philips Microsoft

Reality check (1998) Verbatim: “Theoretically, an ACL should let heterogeneous agents communicate. However, none currently do.” No interoperability Who to blame? Philips Microsoft

Thesis Blame current ACLs Knowledge Query Management Language (KQML): based on wrong principles France Telecom’s Arcol: based on wrong principles FIPA ACL: based on wrong principles  A paradigm shift is needed

What principles? Analysis of communication dimensions: Perspective Type of meaning Semantic / Pragmatic focus Context Coverage of communicative acts

1 - Perspective Private Sender’s perspective Receiver’s perspective Public Multiagent system’s perspective Private perspectives are approximations of the public perspective

1 - Perspective Public perspective is needed: ACLs must be normative Agents must be tested for compliance The ACL must have a public perspective (or compliance testing is not possible) KQML and Arcol: private perspective

2 - Type of meaning Personal Meaning: intent or interpretation of receiver or sender Conventional Meaning: usage conventions Language is a system of conventions Different conventions need different communicative acts

2 - Type of meaning Conventional meaning is needed KQML and Arcol: personal meaning Different communicative acts do not capture different conventions

Dialects KQML failed because many dialects arose; Blame private perspective and personal meaning: Idiolects "When I use a word," Humpty Dumpty said, in rather a scornful tone, "it means just what I choose it to mean -- neither more nor less.“ Lewis Carroll, “Through the Looking-Glass, and What Alice Found There” (1871)

3 - Semantics versus pragmatics Meaning = Semantics + Pragmatics Semantics what symbols denote Pragmatics how syntactic symbols are interpreted and used involves mental states and the environment constrain how agents interact

3 - Semantics versus pragmatics Semantics-focused language is needed Pragmatics require fully-cooperative agents Pragmatics fail where sincerity cannot be taken for granted KQML and Arcol: Pragmatics-focused languages

4 - Context Communication context: needed for understanding. Fixed context Flexible context Goal: flexible context

5 - Coverage of communicative acts Seven categories: Assertives Directives Commissives Permissives Prohibitives Declaratives Expressives Limited coverage vs Full coverage Full coverage is needed KQML and Arcol have limited coverage

Opposing paradigms Mental Agency Focus on mental state (e.g. BDI) Assumes intentional stance How to determine the mental state of agents? Introspection: unsatisfactory or impossible “Mental state” is an abstract concept: only the agent designer warrants compliance. Social Agency Focus on agent behavior (external) “Social creatures” (sic) Compliance : obey conventions in society (self-evident)

Autonomy Design autonomy: agent designer’s freedom: Promotes heterogeneity and applications KQML and Arcol require that agents have BDI-based mental states Execution autonomy: agent’s freedom Arcol assumes sincere, cooperative, benevolent agents KQML is less strict

Proposed solution Social agency Different from traditional ACLs Goals: Public perspective Conventional meaning Semantics over pragmatics Flexible context Full communicative acts coverage

Protocols Agents play different roles Roles Define commitments/obligations Restrictions on behavior and communication Agents can manipulate/cancel commitments  Metacommitments (avoid chaos) Protocol Set of commitments Testability without introspection; closed-source friendly. Autonomy Everything is allowed as soon as commitments are met Context is society (“Social context”) Context is better known and agreed on  better communication

Dialects in societies Agent societies are free from idiolects No private perspective nor personal meaning Dialects  good Allow “context sensitivity” and real- world applications Do not involve introspection No risk of Humpty Dumptyism

Instantiation How is this translated into practice? No clear answer A purely behavior-based approach is not viable – too limiting. The purely-mentalist approach has been criticized so far Combine both solutions: Define when a communicative act is satisfied Assertive: if the world matches what is described Directive: the receiver acts to ensure success Commissive: the sender acts to ensure success Coarse canonical set of objective definitions Do not ascribe beliefs and intentions to agents

Comments / critique Rewrite: BDI-based languages have drawbacks: Too strict Require introspection for compliance testing Limits autonomy Requires full cooperation... but many of the critiques are not adequately justified. Behavior-Commitments based agencies sound good Upon closer inspection, they have their limits as well: not powerful enough. Proposed solution is a not-better defined mix between the two

Conclusion FIPA ACL is based on wrong principles... every possible communication dimension is wrong... but after 8 years FIPA ACL is the standard. Some of the proposed concepts are intriguing, but they can not be easily translated into practice.