DiBeRT - A Distributed Belief Revision Test bed
Purpose
The Distributed Belief Revision Tested - DiBeRT is intended to study and
model inherently distributed systems, with decentralised control, in which
the available information is incomplete and dynamic, and, the time factor
is relevant.
A companion project that has already finished, which tried out to apply
most of the topics related with those here described can be found
HERE
.
Methodologies
The decision of how and when to use external beliefs in an agent's knowledge
base is fundamental to the characterisation of a distributed belief revision
system. The DiBeRT agents act in "good faith" and exchange messages using
the direct message passing mechanism, thus guaranteeing, that the information
received by a recipient agent is, not only, relevant for its activity,
but also, truthful from the sender's perspective. A wide range of different
methodologies for the use of incoming beliefs can be adopted by the recipient
agents: unconditional use, conditional use, rejection, etc.. From this
spectra DiBeRT has chosen two policies for the local use of communicated
beliefs:
-
local evaluation of the shared propositions - the local beliefs
prevail over the communicated beliefs, i.e., the use of an external belief
is conditioned by the existence or absence of the belief in the agent's
local knowledge base. Once the local agent perceives or infers its own
local perspective about a shared proposition, the external points of view
regarding the shared proposition are ignored. In the absence of locally
deduced belief the existing external beliefs will be used. A shared
proposition is represented by as many nodes as there are agents with beliefs
concerning the shared proposition, regardless of whether the external beliefs
are or are not used locally;
-
global evaluation of the shared propositions - every communicated
belief is unconditionally used by the recipient agent. A shared proposition
owned by different agents has a multiple node representation: is represented
by as many nodes as there are agents with beliefs concerning the shared
proposition.
Upon accepting a set of external beliefs, an agent may find itself with
conflicting belief status for the same proposition. In such circumstances
which belief status to adopt? In DiBeRT three synthesis criteria were implemented
guaranteeing the attribution of an unique belief status to every shared
proposition:
-
the consensus synthesis criterion (CON) - a shared proposition is
believed, if and only if, it is believed by every agent that share the
proposition;
-
the majority synthesis criterion (MAJ) - a shared proposition is
believed, if and only if, it majority of the agent that share the proposition
believe in the proposition;
-
the at least one synthesis criterion (ALO) - a shared proposition
is believed as long as there is some agent where it is believed.
These synthesis criteria reflect different levels of demand: in the case
of the ALO synthesis, the belief in a shared proposition by one of the
involved agents is enough to make it believed by the system, in the case
of the MAJ synthesis, only if the majority of the agents believed in the
proposition will the shared proposition be believed by the system, while,
in the case of the CON synthesis, only the consensus among the involved
agents will make the shared proposition believed by the system.
Functionalities
A test bed for the evaluation and validation of different distributed consistency
maintenance methodologies has to be reusable, and user friendly. In DiBeRT,
the user is asked, at launch time, to select from the available set of
agents the sub-set to be run, the synthesis criterion to be applied, and
the level of consistency desired. One of the following four available distributed
consistency modes is selected for execution:
-
shared beliefs local evaluation and consensus belief status synthesis
criterion;
-
shared beliefs local evaluation and majority belief status synthesis
criterion;
-
shared beliefs local evaluation and at leat one belief status synthesis
criterion;
-
shared beliefs global evaluation and consensus belief status synthesis
criterion;
-
shared beliefs global evaluation and majority belief status synthesis
criterion;
-
shared beliefs global evaluation and at leat one belief status synthesis
criterion.
The private beliefs consistency level is unique: they are locally consistent.
After launching the community of multi-agents, the interaction between
DiBeRT and the User is performed by a specialised agent called User Interface
Agent. The User Interface Agent architecture is identical to the remaining
system agents, being the intelligent system role played by User. This interface
allows, during runtime: (i) the addition of new assumptions; (ii) the multiple
contexts management; (ii) the attribution of specific belief status; (iv)
the querying of the system about any beliefs.
For further information please contact:
Benedita Malheiro E-mail:
[email protected]
Eugenio Oliveira
E-mail: [email protected]