The Estimation of Uncertain Gates: An Application to Educational Indicators.

Uncertain gates in possibility theory correspond to noisy gates in probability theory. They allow us to reduce the number of parameters which must be elicited to define conditional possibility tables in possibilistic networks. Usually the choice of the connector and its parameters is made by experts but sometimes if there is an existing CPT or if data are available, it can be interesting to perform an estimation of uncertain connectors. This estimation allows us to better understand how information is combined. Furthermore, it is possible to match one of three sorts of behaviour (indulgent, compromise, severe) with the corresponding information. This point is important for knowledge engineering. If the data are available, the estimation can also be useful to verify if the uncertain connector fits with the data, because expert knowledge is often imprecise and uncertain. In this paper, we will show how to perform the estimation of uncertain gates and we will illustrate our approach with several examples of results in the domain of education.