Implicit Culture for Multi-agent Interaction Support Paolo Giorgini Department of Mathematics University of Trento pgiorgini@science.unitn.it Joint work with: Enrico Blanzieri, Paolo Massa and Sabrina Recla Giorgini - CoopIS 2001 Outline • Motivations • Implicit Culture • Systems for Implicit Culture Support (SICS) – A SICS for Multi-agent interaction support • The eCulture Brokering System • Conclusion and future work Giorgini - CoopIS 2001 2 Motivations • Interaction among agents is crucial for the efficiency of MAS – new agents enter into the system without the necessary knowledge and skills – new agents are not able to learn from the others’ behavior – it is not possible to define and represent a priori the relevant knowledge the agents need for the interaction Giorgini - CoopIS 2001 3 Motivations • In order to improve its behavior, a new agent should act consistently with the knowledge and the behaviors (culture) of the other agents. • We propose a way for supporting multi-agent interaction based on the idea of Implicit Culture [Blanzieri, Giorgini and Giunchiglia: 2000] Giorgini - CoopIS 2001 4 Implicit Culture: basic definitions (1) Let P be a set of agents, O a set of objects, A a set of actions. We define: • environment eP O e • scene as the pair <B,A>, where B , and A A • situation as <a,s,t>, where aP and sis a scene • executed situated action as the action executed in given situation. • Fe : deterministic function that describes the evolution of the environment. Giorgini - CoopIS 2001 5 Environment Fe a g st s”t c a b s’t e b Giorgini - CoopIS 2001 6 Environment c a s”t+1 st+1 b Giorgini - CoopIS 2001 s’t+1 e 7 Implicit Culture: basic definitions (2) • Random variable ha,t that describes the action that the agent a executes at the time t • expected action as the expected value of ha,t , E(ha,t ) • situated expected action as the expected value of ha,t given a situation <a,s,t>; E(ha,t |<a,s,t>) • Cultural constraint theory for a group GP, as a theory on the situated expected actions of the agents of G • Cultural action w.r.t. G, as an executed action that satisfies a cultural constraint theory for G Giorgini - CoopIS 2001 8 Implicit Culture: basic definitions (3) Implicit Culture Relation between G e G’ such that the expected situated actions of G’ are cultural action for G Implicit Culture phenomenon G and G’ are in implicit culture relation Giorgini - CoopIS 2001 9 … the idea G G’ a a st g b s’t s”t c b e the agents of G’ perform actions that agents of G would perform in the same situations Giorgini - CoopIS 2001 10 Systems for Implicit Culture Support (SICS) Goal: establish an implicit culture phenomenon – acquisition of cultural constraint theory for G – proposing to G’ scenes such that the expected situated actions satisfy the cultural constraint theory for G. Giorgini - CoopIS 2001 11 SICS: architecture Observer stores in a data Inductive Module S S0 base the situated executed actions of the agents of G. Inductive Module that using the data of the DB and the a priori theory So, induces a cultural constraint theory S. G G’ a Giorgini - CoopIS 2001 a st s”t g Composer proposes to a group G’ a set of scenes such that the expected situated actions satisfies S. DB Observer Composer b s’t c b e 12 SICS: architecture Observer stores in a data S0 Inductive Module S base the situated execute actions of the agents of G. Inductive Module that using the data of the DB and the a priori theory So, it induces a cultural constraint theory S. G G’ a st+1 s”t+1 Composer proposes to a group G’ a set of scenes such that the expected situated actions satisfies S. Giorgini - CoopIS 2001 DB Observer Composer b e c s’t+1 13 The eCulture Brokering System • The system is the result of collaboration between University of Trento and ITC-irst. • Goal: Permit to a citizen to access, via web, to the information about cultural goods collected in the (Trentino) museums. – The user demands the system information about cultural goods related to a particular epoch. – The system queries the databases of the museums and answers. Giorgini - CoopIS 2001 14 Multi-agent Architecture Personal Agent (PA) permits a user to access the system Br1 Br2 … Brk Wrapper (Wr) is the interface Wr1 PA1 … Wr2 DB Wrh DF ARB Broker (Br) builds an “answer” with some grade of specialization in an area. Agent Resource Broker (ARB) … PAn DB between the system and a database. DB gives information about the external available resources Directory Facilitator (DF) knows the agents of the system and their services Giorgini - CoopIS 2001 15 User interface Giorgini - CoopIS 2001 16 Agents interaction Wr1 DB Wr2 DB PAn … Br2 … Brk … Br1 1) a user, by the PA, requests information PA1 Wrh DF ARB DB about a century; the PA asks the DF which Broker can satisfy the request. 2) The PA accepts or refuse the proposed Broker; it sends to the accepted Broker the request of the user. 3) the Broker asks the ARB which are the external resources that can be useful. 4) The Broker asks the DF which Wrappers are able to interface the resources. 5) The Broker queries the Wrappers and build the answer for the user. Giorgini - CoopIS 2001 17 Agents interaction Wr1 DB Wr2 DB PAn … Br2 … Br1 1) a user, by the PA, requests information … Brk PA1 Wrh DF ARB DB about a century; the PA asks the DF which Broker can satisfy the request. 2) The PA accepts or refuse the proposed Broker; it sends to the accepted Broker the request of the user. 3) the Broker asks the ARB which are the external resources that can be useful. 4) The Broker asks the DF which Wrappers are able to interface the resources. 5) The Broker queries the Wrappers and build the answer for the user. 6) The Broker send the answer to the PA that sends it to the user Giorgini - CoopIS 2001 18 DF and Implicit Culture • The DF provides a “yellow pages” service • The Brokers are specialized in a different thematic areas The SICS is used to support the activity of the DF with the goal of suggesting to each PA the most suitable Broker Giorgini - CoopIS 2001 19 DF and Implicit Culture Agents observed: Personal Agents (G = G’) Observed Actions: <requests,x,s,t> : The PA x request to the DF, at time t, a Broker for getting information about the century s <accepts,x,y,s,t> : at time t, PA x accepts the Broker y, proposed by the DF, about the century s <refuses,x,y,s,t> : at time t, PA x refuses the Broker y, proposed by the DF, about the century s Proposed Scenes: Brokers Cultural Constraint Theory: <request,x,s, Giorgini - CoopIS 2001 time > <accepts,x, Broker ,s> 20 Example PA1 asks for a Broker for the VI century. Observation stored by the SICS Br0 Br1 Br2 Br3 1) find the cultural actions Refuse(IV) Accept(XIII) Accept(XVII) Accept(VI) Accepts(VI) PA1 Refuse(IV) Accept(II) Accept(XVII) Accept(XI) Refuse(XVII) PA2 Refuse(IV) Accept(XVII) Accept(VI) Refuse(XVII) Refuse(IV) Accept(XI) PA0 PA3 2) find the scenes 1. Find the predictive agents Refuse(XVI) PA0, PA2 2. Select the similar agents Giorgini - CoopIS 2001 PA1 is more similar to PA2 than to PA0 3. Propose the scene with the maximum probability of facilitation Br1 21 The eCulture Brokering System • Developed using JACK Intelligent Agents, a commercial agent-oriented development environment built on top of and fully integrated with Java • It follows FIPA (Foundation for Intelligent Physical Agents) specifications for DF and ARB • Databeses: Oracle and Microsoft Access Giorgini - CoopIS 2001 22 Conclusions • We have presented – the idea of Implicit Culture and how to use it for supporting Multi-agent interaction – eCulture Brokering System • Implicit Culture Support allows us to improve the agents interaction without need to equip the agents with additional capabilities • Future work: – Extend the use of SICS to other agents, in particular to the ARB (Agent Resource Broker) – Implementing the inductive module for inducing cultural constraint theories for different groups of agents Giorgini - CoopIS 2001 23 … more • http://www.science.unitn.it/~pgiorgio/ic Giorgini - CoopIS 2001 24