370 likes | 385 Views
Explore challenges in arranging interactions among heterogeneous information components for effective and efficient integrations, considering structure, process, and policy.
E N D
Research Directions forAgents and Ontologies Michael N. Huhns Center for Information Technology University of South Carolina
The Fundamental Problem • We would like to arrange for effective and efficient interactions among large numbers of heterogeneous information components: databases, applications, and interfaces • Difficulties are • Components are incomprehensible, inconsistent, and often unknown in advance • We need to enable updates as well as retrievals • The information environment is open • We need to consider process and policy, as well as structure University of South Carolina
MCC Carnot Project:Ontologies and Articulation Axioms Common Ontology Application 1 Interface 1 TransportationDevice Articulation Axiom 3 Articulation Axiom 1 Train Vehicle Boat Truck Automobile Jeep Articulation Axiom 2 Articulation Axiom 4 DB1 DB2 Car Auto id make no model University of South Carolina
Three Problems • Where does the ontology come from? • The ontology must be greater than or equal to any local schema • How are the mappings created? • In general, they are arbitrarily complex and not one-to-one • Both schemas and data must be mapped • How and when are the mappings applied? • Agents University of South Carolina
Ontology Development • Bottom-Up from Schemas and Key Words • identify databases • identify names for all tables, fields, and enumerated values (e.g., if value is limited to a primary color “red”, “green”, or “blue”) • form groups of common concepts and assign name to covering concept for each group • iterate; or • Extensional View: form classes from instances University of South Carolina
Ontology Development • Top-Down from First Principles (intensional view): a class is defined by a set of membership conditions or properties • Restrictions on Class Formation: • a class must have instances • a class must contain all properties common to the instances in its extension • classification should obey cognitive economy--instances of a class must share some, but not all properties • classification should enable inference of properties based on class membership University of South Carolina
Ontology Development (cont.) • Restrictions on Class Structures: • Completeness--every property must be used in the definition of at least one class • Nonredundancy--a subclass must be defined by at least one property not in any of its superclasses (the result is that a subclass is always a specialization of any of its superclasses, i.e., it has more properties or restrictions, and has fewer instances) University of South Carolina
Alternative:Relate Ontology Pieces University of South Carolina
Tools for Developing Ontologies • Ontolingua and Chimaera (Stanford) • SHOE: Simple HTML Ontology Extension language (U. of Maryland) • JOE: Java Ontology Editor (U. of South Carolina) • IMTS (MCC) • Cyc Unit Editor • UML, ER, and Conceptual Modeling Tools University of South Carolina
Classification Is Difficult! From the ancient Chinese encyclopedia Celestial Emporium of Benevolent Knowledge, “It is written that animals are divided into • belonging to the emperor • embalmed • tame • sucking pigs • sirens • fabulous • stray dogs • included in the present classification • frenzied • innumerable • drawn with a very fine camel-hair brush • et cetera • having just broken the water pitcher, and • that from a long way off look like flies.”
Creating Semantic Mappings Interface Application Database use CASE Tool generate Conceptual Schema Cognition Universe of Discourse 1 MIST Universe of Discourse 2 Interface Application Database use CASE Tool generate Conceptual Schema Cognition University of South Carolina
Semantic Translation Semantic Translation by Mappings by Mappings Semantic Translation Semantic Translation by Mappings by Mappings Semantic Translation by Mappings DB1 DB1 DB1 Semantic Translation User Application 1 Application n Agent for Application Agent for Application Common Enterprise-Wide View Agent for Resource Agent for Resource Agent for Resource University of South Carolina
User Agent Resource Agent User Agent Resource Agent User Agent Resource Agent An Agent-Oriented Information System Middleware: Mediators, Brokers, Facilitators, Ontologies, and Registries
(de facto) Standard Agent Types and Architectures Application Program User Interface Agent MCC InfoSleuth CMU RETSINA SRI OAA USC-ISI SIMS & TeamCore Global InfoTek Grid Reply Reg/Unreg (KQML) Reply Query or Update (SQL) Ontology Agent Broker Agent Reg/Unreg (KQML) Mediator Agent Ontology (OKBC) Reg/Unreg (KQML) Registry Agent Mediated Query (SQL) Reg/Unreg (KQML) Schemas (CLIPS) 11179 Registry Mediated Query (SQL) Reply Reply Database Resource Agent Database Resource Agent SQL (JDBC) University of South Carolina
Agent Abstractions Agent abstractions are mentalistic beliefs: agent’s representation of the world knowledge: (usually) true beliefs desires: preferred states of the world goals: consistent desires intentions: goals adopted for action University of South Carolina
Multiagent Abstractions • Multiagent abstractions involve interactions • social: about collections of agents • organizational: about teams and groups • ethical: about right and wrong actions • legal: about contracts and compliance University of South Carolina
Why Do These Abstractions Matter? • Because modern applications go beyond traditional metaphors and models in terms of their dynamism, openness, and trustworthiness • virtual enterprises: manufacturing supply chains, autonomous logistics • electronic commerce: utility management • communityware: social user interfaces • problem-solving by teams University of South Carolina
Fundamentals ofSocial Abstractions • Commitments: social, joint, collective, … • Organizations and roles • Teams and teamwork • Joint intentions vs. individual rationality University of South Carolina
Kinds of Commitment • Psychological or mental: an agent’s state of being committed to a belief or an intention • Joint: agents’ commitments to the same intention or belief • Mutual: agents’ commitments to one another with respect to the same condition University of South Carolina
Social Commitments An agent’s commitment to another agent • unidirectional • arising within a well-defined scope or “context”, which is itself a MAS • revocable within limits University of South Carolina
Organizations Organizations help overcome the limitations of agents in various respects • reasoning • capabilities • perception • lifetime and persistence • shared context, essential for communicating University of South Carolina
Modeling Organizations • Abstractly, organizations • consist of roles • requiring certain capabilities • offering certain authorities • involve commitments among the roles • Concretely, organizations • consist of agents • acting coherently University of South Carolina
Teams Tightly knit organizations • shared goals, i.e., goals that all team members have • commitments to help team members • commitments to adopt additional roles and offer capabilities on behalf of a disabled member University of South Carolina
Teamwork When a team carries out some complex activity • negotiating what to do • monitoring actions jointly • supporting each other • repairing plans University of South Carolina
Joint Intentions Traditional accounts of teams are based on joint intentions and mutual beliefs • Team-members jointly intend the main goal of the team, which means that they • all intend it and mutually believe that they intend it • each will notify the others if it drops out and mutually believe this notification requirement University of South Carolina
Elements of Trust Ultimately, what we would like is to trust our agents. Trust involves • having the right capabilities • following legal contracts where specified • supporting one’s organization or society • being ethical • failing all else, being rational University of South Carolina
Programming Paradigms • 1950’s -- Machine and assembly language • 1960’s -- Procedural programming • 1970’s -- Structured programming • 1980’s -- Object-based and declarative programming • 1990’s -- Frameworks, design patterns, scenarios, protocols, and components (ActiveX/COM and Java Beans) The trend has been from elements that represent abstract computations to elements that represent the real world University of South Carolina
Interaction-Oriented Software Development • Modules are active • Modules are specified declaratively, in terms of “what”, not “how” • Modules hold beliefs about the world, especially about themselves and others • Modules more closely represent real-world objects • Modules volunteer University of South Carolina
Programming Paradigm • Effort is on assembling and coaching a team to achieve desired functionality • Requirements specified in terms of social commitments and team intentions • Components negotiate with each other, enter into social commitments to collaborate, and can change their mind about their results University of South Carolina
Features ofLanguages and Paradigms Concept Procedural Language Object Language Multiagent Language Abstraction Building Block Computation model Design Paradigm Architecture Modes of Behavior Terminology Type Instance, Data Procedure/Call Tree of procedures Functional decomposition Coding Implement Class Object Method/Message Interaction patterns Inheritance and Polymorphism Designing and using Engineer Society, Team Agent Perceive/Reason/Act Cooperative interaction Managers, Assistants, and Peers Enabling and enacting Activate University of South Carolina
Example: Children Forming a Circle (Most business software modules, which are passive, are meant to represent real objects, which are active) University of South Carolina
What is IOP? • A collection of abstractions and techniques for programming MAS • Classified into three layers of mechanisms • coordination: living in a shared environment • commitment: organizational or social coherence (adds stability over time) • collaboration: high-level interactions combining mental and social abstractions University of South Carolina
Benefits of IOP • Like conceptual modeling, IOP offers a higher-level starting point than traditionally available: • key concepts of coordination, commitment, collaboration as first-class concepts that can be applied directly • aspects of the underlying infrastructure are separated, leading to improved portability • ultimately, can lead to a many-fold increase in software productivity and performance by changing the way in which software is developed and deployed University of South Carolina
Open Problems in Agent Technology • Design rules for cooperative systems • Workflows • Ontologies • Security concerns in open environments • Scaling to millions of agents • User interfaces to heterogeneous resources • to acquire resource constraints • to access information • to control workflows • Shared ontology maintenance University of South Carolina
Ubiquitous Agents • Agent technology will affect • homes • transportation • commerce • business (supply chains) • education • Agents are popular because they provide a new way to think about systems --they are not simply a parallelization of a centralized system! University of South Carolina
A “Commuter” Approach to Logistics Instead of following a centralized plan for deploying supplies, imagine that each item of materiel is an intelligent agent whose sole objective is to reach its assigned destination. Just like a person commuting to work, this agent would dynamically • decide its means of conveyance • contend for storage and transportation resources • avoid or resolve conflicts with other agents • make local decisions as it wends its way through a distribution network University of South Carolina