[STA] Information-based agency

John Debenham is going to explain this approach. This is the main reason I’m in Barcelona today. It’s going to be hard, so pay attention please.

Agents negotiate in a very uncertain environment about the world states, integrity of messages, enactment of commitments, completeness of contracts or validity of information, so we do know that we need a more predictable world for agents.

If uncertainty is the problem then information is the solution. Information-based agency is a framework on techniques to manage and exchange information. Basics
an agent \(\alpha\) interacts with negotiation and with information agents
a common ontology O is assumed organised as an is-a hierarchy
Sim(#) measures similarity between concepts
world model contains random variables \({X_1, X_n}\) For each one, ‘zero information’ is represented by a decay limit distribution Di, where when there is not information \(\lim_{t \to \infty}X_i=D_i\) That is: the information is useless after a period of time.

Singed contracts and arguments and information are all commitments \(\varphi\). All similar statements of over related concepts (for example, statement about a topic from the same user -weather forecasting-) are expected to have the same decay function.

Information can help us to establish Trust mechanism and model intimacy (keeping private information) that organises the type of information (motivations, goals…) and an abstraction of the ontology.

The big picture of information-based agents is something like this (look for it)

Blogged with the Flock Browser

Top