As far as I know, computers have not been applied in most of the major aspects of interpretation of fossil assemblages - the broad fields of paleoenvironment, evolution and paleobiology. In the long run, however, it is likely that expert systems will prove useful in this area, by providing the "conventional wisdom" against which new observations and hypotheses can be checked.
But computers have been widely applied in the areas of zonation and stratigraphic correlation - for example, the probabilistic procedure for ranking and scaling of paleontologic events, and for correlation, developed by F.P. Agterberg and F.M. Gradstein (see Agterberg and Byron, 1990 , and references therein), and the deterministic procedure for finding unitary associations devised by Guex and Davaud (1984).
One thing that has always bothered me about automated evaluation of the limits of stratigraphic ranges of species, and the constitution of unitary assemblages, is that no provision is made to take account of the relative reliability of each earliest or latest occurrence, or of species presences/absences in an assemblage. The dependability of these limits, or presences/absences, is affected by a number of factors such as the usual abundance of the taxon, the ease of distinguishing it from others in an assemblage, the preservation of the assemblage in relation to the robustness or delicacy of the taxon, the relation of the collecting locality to the area of distribution of the taxon, and so on. M.J. Westberg and I developed an "index of reliability " based on these factors, and applied it to the radiolarians in a few Deep Sea Drilling Project sequences, but we had no opportunity [= funding] to pursue the matter further.