Barbara Dunin-Keplicz, "Taming complex bieliefs"
<br>
<br>

Most logical approaches to modeling beliefs lead to a high complexity
of reasoning. As our aim is to bridge the gap between idealized models of  beliefs and their actual
implementations, I will show that a shift in perspective can
substantially reduce the complexity. This shift integrates stages of
belief acquisition, intermediate reasoning and final belief formation and permits a uniform modeling of individual and
group beliefs, where group is a generic concept consisting of
individual agents, groups of agents, groups of groups of agents, etc.
<br>
Importantly, the resulting layered architecture
allows one to avoid costly revisions of agents' beliefs when they join
a group. This is especially important when paradigmatic agent
interaction is considered. Cooperation, coordination and communication
is naturally modeled by creating a group and forming group beliefs to
achieve a common informational stance. What sort of structure it is
and how this influences agents' individual beliefs is a matter of
design decisions.
<br>
In order to ensure flexibility, individual characteristics of agents
are reflected in the diversity of epistemic profiles  characterizing
both their reasoning capabilities as well as the manner of dealing with conflicting or
lacking information. Naturally, agents reach conclusions by combining
various forms of reasoning, including belief fusion, disambiguation of
conflicting beliefs or completion of lacking information. This rich
repertoire of available methods enables for heterogeneity of agents'
reasoning characteristics. This way, presented approach ensures both the heterogeneity of agents involved and a
flexibility of group level reasoning patterns.
<br>
I will present a novel semantics suitable for
building complex belief structures in the context of incomplete and/or
inconsistent information. Namely, an agent starts with constituents,
i.e., sets of initial beliefs acquired by perception, expert supplied
knowledge, communication with other agents and perhaps other ways.
Next, the constituents are transformed into consequents according to
agents' epistemic profiles.
<br>
Importantly,  a recently proposed query language 4QL can serve  as a tool to implement all
epistemic profiles and belief structures constructible in
deterministic polynomial time. One can then query them in a tractable
manner, which provides a rich but still pragmatic reasoning machinery.
Also, nonmonotonic/defeasible
reasoning techniques are easily expressible in 4QL enhancing the agents' reasoning capabilities in real life applications.