BELIEF is a package of Common Lisp functions which are designed to manipulate graphical belief models. It uses the Shenoy and Shafer version of the algorithm, so one of its unique features is that it supports both probability distributions and belief functions. It also has limited support for second order models (probability distributions on parameters). Keywords: Belief functions (Dempster-Shafer), Bayesian Probability, Graphical Models (Bayes Nets), 2nd order models. Two versions are currently available via anonymous ftp from ftp.stat.washington.edu (128.95.17.34). BELIEF 1.1 --- Old version ran under Lucid Lisp 3.0, with some incomplete attempts at ports to AKCL and VaxLisp. Written before Steele[1990] (CLtL-2) Some known bugs and portions of code which were written but not fully debuged. BELIEF 1.2 --- Attempt to track changes from X3J13 and be compatable with Allegro's multiprocessing stuff. Works well on Allegro 4.0, but I haven't tried it on any other version of Lisp. More thoroughly tested, no known bugs (but some unknown ones). There will probably be annoying little problems on other lisps, many of them having to do with the fact that ANSI Common LISP is still rather a moving target. I will be happy to give email support to people trying to get it up on other versions of LISP (I'm particularly interested in problems installing the new version on Lucid Lisp, which I don't have.) Currently, I'm working on a project called "GRAPHICAL-BELIEF." This will be a commercial version of BELIEF which will include interactive graphical displays and bundled knowledge acquisition tools. After GRAPHICAL-BELIEF becomes more solid, I hope to re-release the core inference engine back into the public domain as BELIEF 2.0. I continue to be interested in corresponding with anybody who is doing graphical modelling with BELIEF or belief functions. Russell Almond StatSci (a division of MathSoft, Inc.) 1700 Westlake Ave., N Suite 500 Seattle, WA 98109 (206) 283-8802 almond@statsci.com