### From OntologPSMW

Department of Systems Engineering and Operations Research
(1B)

Associate Director, Center of Excellence in Command, Control, Communications, Computing and Intelligence (C4I Center)
(1C)

Fairfax, Virginia, USA.
(1E)

e-mail: klaskey-at-gmu.edu
(1F)

see:
(1G)

I teach courses in Systems Engineering, Decision Support, Bayesian Statistics, and Computational Probabilistic Reasoning. My primary research interest is knowledge representation and reasoning in domains requiring decision making under uncertainty. I developed Multi-Entity Bayesian Networks (MEBN), a knowledge representation language based on first-order Bayesian logic (FOBL). First-order Bayesian logic integrates classical first-order logic and Bayesian probability theory. A theory in FOBL augments a classical first-order theory that defines a set of possible worlds with a coherent assignment of probabilities to subsets of possible worlds. FOBL is inherently open-world -- when we learn something new, we use Bayes Rule to update our knowledge and obtain new probabilities that incorporate the new knowledge.
(1J)

I believe we will ultimately find it necessary to augment the current generation of ontology languages with a way to represent uncertainty. Even if the determinists are right (which I doubt) that the entire future evolution of our Universe was determined at the time of the Big Bang by the Laws of Physics and the initial conditions, ontologies developed by humans and/or computers will always be partial and incomplete. Partial and incomplete knowledge is of limited practical use without a way to characterize the plausibility of assertions that can't be proven. Generic knowledge about plausibility (e.g., co-occurrence likelihoods of symptoms and diseases, as distinct from instance-specific likelihoods that particular patients have particular diseases) can and should be represented in domain ontologies. Just as physics was dragged kicking and screaming into embracing probability as a fundamental aspect of the foundational theory of modern physics, and just as the field of artificial intelligence first reluctantly and then enthusiastically embraced probabilistic knowledge representation formalisms, I believe computational ontology will soon find it necessary to build uncertainty into many kinds of ontologies.
(1K)

See also:
(1L)

- First and Second ISWC Workshops on Uncertainty in the Semantic Web ( URSW 2005 and URSW 2006), co-organized by myself and Ken Laskey, Paulo Costa, Mike Pool and Francis Fung. (1N)