Social Abstractions for Information agents 1999년 11월 22일 전산과학과 이근섭

Introduction 문제 상황 : information-rich environment 문제 상황 : society of information agents 해결 : interaction modeling

Contents of presentation key concept interaction oriented programming architecture example conclusion & future work

Key concept 1 information-rich environments span boundaries heterogeneous loosely structured lack global control of contents intricate interdependency

Key concept 2 information agent social interaction find vs manipulate agent consistency vs coherence

consistency vs coherence autonomy vs consistency expectation & suitability consistency < coherence

Interaction oriented programming  until now  buried in operating procedures & manual  hard-coded  from now  modeling interactive aspects

Sphere of commitments (1/3) society role capability : required by role commitment : what agent must do authorities : what agent may do soCom : sphere of commitments

Sphere of commitments (2/3) C( x, y, p, G ) x : debtor => make a commitment y : creditor => receiver the commitment p : condition => what is being committed to G : context => soCom & real society

Sphere of commitments (3/3)  meta-commitment  create : consequence of adopting a role  discharge , cancel , debtor : by debtor  release , assign : by creditor or context  policy  cancel( x, C(x,y,p,G) ) => create ( x, C(x , y , q , G) )

Standard architecture (1/3) user agent application broker Execution agent ontology ontology agent Physical controller Execution agent DB Sensor

Standard architecture (2/3)  user agent  contain mechanism to select an ontology  support various interface  view of other agent  provide access to other resources  execution agent  manage software / hardware application  resource agent  represent resource  translate ( local vs global language )

Standard architecture(3/3)  broker agent  name service  message service  mediator agent  determine relevancy  decompose query & combine responses  ontology agent  provide ontology database  manage the evolution of ontology

Example 1 : MINDS Multiple Intelligent Node Documents Servers A B user agent B Document Document Metadata Metadata Search engine Metadata C Document Metadata repository Metadata Document base Network

Example 2 : Virtual Enterprise order matching hoses and valves promise Notify of update Confirm revised user agent VE agent order cancel ready order valves ready change order hose hoses valves

Conclusion & future work  consistency => coherence  interaction oriented  standard architecture  future work  develop metamodel  intuitive formal semantics  compositionality  temporal, causal reasoning  social learning