Abstract interpretation has a reputation of being a kind of ``black art,'' and consequently difficult to work with. This paper describes a twenty-year quest by the first author to address this issue by raising the level of automation in abstract interpretation. The most recent leg of this journey is the subject of the second author's 2014 Ph.D. dissertation. The paper discusses several different approaches to creating correct-by-construction analyzers. Our research has allowed us to establish connections between this problem and several other areas of computer science, including automated reasoning/decision procedures, concept learning, and constraint programming.

(Click here to access the paper: PDF; (c) Springer-Verlag.)