Wet Grass

Many belief network researchers have used networks similar to this (possibly not containing nodes 'Neighbor's Grass' or 'Wall') to demonstrate "explaining away", and how rule based or evidence support systems may do incorrect chaining with these types of situations. For instance, they may combine the rules "If the ground is wet then it is likely it rained" with "If the sprinkler was on then it is likely the ground is wet" to get "If the sprinkler is on then it is likely it rained" when in fact the opposite is true. For example see Judea Pearl (1988) Probabilistic Reasoning in Intelligent Systems, p. 7. The probabilities for this network were chosen by Norsys.

Paper link: <www.norsys.com...>

Brent Boerlage
Netica .dne format