Technologies for the right problem are in a very early stage of development. A core problem is that we need a new synthesis of first-order (or high-order) logics and probabilistic reasoning, so that we can construct and use statistical models of relational data. In the past two decades, some work in inductive logic programming, social network analysis, and Bayesian networks has produced some good foundations, but years of additional work is needed before we will have tools that match the level of development of current commercial tools for propositional data.

Today, and particularly since late 2001, we have a small, loosely knit community of researchers in computer science that is moving forward with strong support from NSF, DARPA, and others. Because of this strong research support, we have made important strides in the past few years, and we expect even more substantial results in the next several years.