Communication Scenarios

Personal Agent (A)

This scenario is the most important one and also builds the foundation for the other two communication scenarios. A user of an Instant Messaging network installs his own personal agent and configures information sources he owns or trusts. Information sources can be for example a FOAF profile of the user containing personal information about the user and about relationships to other people he knows and where to find further information about these. This information is represented in FOAF using the properties foaf:knows and rdfs:seeAlso.
The following listing shows an excerpt from a FOAF profile.
:me a foaf:Person ; foaf:knows [ a foaf:Person ; rdfs:seeAlso <http://eye48.com/foaf.rdf> ; foaf:name "Michael Haschke" ; foaf:nick "Haschek" ] .
Additionally this FOAF profile can link to other RDF documents which contain more information about the user and his activities. The RDF version of his calendar, for example, could be linked as follows:
:me rdfs:seeAlso <http://.../ical2rdf?u=http...> . <http://.../ical2rdf?u=http...> a ical:Vcalendar; rdfs:label "Haschek's Calendar" .
Such links span a network of information sources as depicted in theFigure above. Each user maintains his own information and links to information sources of his acquaintances. Depending on the query, the agent will access the respective resources.

Group Agent (B)

This communication scenario differs from the Personal Agent scenario in that multiple users get access the same agent. The agent should be able to communicate with multiple persons at the same time and to answer queries in parallel. As is also depicted in the above Figure the agent furthermore does not only access remote documents but can also use a triple store for answering queries.
When used within a corporate setting this triple store can for example contain a directory with information about employees or customers. The triple store can be also used to cache information obtained from other sources and thus facilitates faster query answering.
For agents themselves, however, the distinction between RDF sources on the Web and information contained in a local triple store is not relevant.

Agent Network (C)

This scenario extends the two previous ones by allowing communication and interaction between agents.
The rationale is to exploit the trust and provenance characteristics of the Instant Messaging network: Questions about or related to acquaintances in my network of trust can best answered by their respective agents. Hence, agents should be able to talk to other agents on the IM network. First of all, it is crucial that agents on the IM network recognize each other. A personal agent can use the IM account of its respective owner and can access the contact list (also called roster) and thus a part of its owner's social network. The agent should be able to recognise other personal agents of acquaintances in this contact list (auto discovery) and it should be possible for agents to communicate without interfering with the communication of their owners. After other agents are identified it should be possible to forward SPARQL queries (originating from a user question) to these agents, collect their answers and present them to the user.