Date: Tue, 10 Dec 1996 15:17:46 GMT Server: NCSA/1.4.2 Content-type: text/html Last-modified: Wed, 28 Feb 1996 00:57:32 GMT Content-length: 2189 Probabilistic Temporal Reasoning

Probabilistic Temporal Reasoning

Developing models of temporal reasoning (essentially, reasoning about dynamical systems) is just about as old as AI itself, and has been tackled using monotonic and non-monotonic logical frameworks, and more recently using probabilistic models.

We are working on a probabilistic model that particularly addresses issues concerning endogenous change---situations in which the system changes state but not as a result of exogenous forces applied to the system. In our domain---medical prediction, diagnosis, and treatment decisions---an exogenous event would be a planned treatment or test, and an endogenous event would be something like internal bleeding causing vital signs to destabilize. We are working on models that can be easily assessed from a human expert.

Another goal of the project is to understand the difference between the way Statisticians and Computer Scientists have approached the problem of temporal reasoning, and to say something interesting about causality in the process.

Other members of the team are:


Paper references:
hanks@cs.washington.edu