To be useful in undergraduate education, such tools must be accessible to students willing to invest, say, 5 to 25 hours of study in their use. (With a nominal study-time of two hours outside class for each our in class, that investment comprises one to five weeks of course work effort.) In our experience over the past five years at the University of Oklahoma, the DrACuLa/ACL2 programming-environment/theorem-prover combination provides the necessary support for effective education on automated, predicate-based testing and the use of mechanical logic to formally verify software properties for senior-level computer science students. The use of these tools comprises about a third of the lecture material and project work associated with a required course in software engineering.

Ten to twelve hours of the 37 hours of lecture time in the course are devoted to the DrACuLa/ACL2 topic, and students succeed at normal rates. Almost all of them acquire the ability to effectively use predicate-based, automated testing provided in DrACuLa, the great majority have at least a few successes with ACL2, the top half acquire enough skills with ACL2 to seek experience similar tools in future projects, and the top decile becomes more effective with ACL2 than one could reasonable expect. A second, project-based software engineering course makes extensive use of the facilities with no significant additional lecture time devoted to the topic. The project-based course is also required for BS students in CS at OU, and virtually all students see a reasonable measure of success, with the best exceeding expectations.

Presently, the core programming courses focus on Java to better prepare students for the Advanced Placement exam. However, beginning in 2009 AP testing will no longer include an AB level, which is the level targeted by certain current programming courses. At that point, the faculty will reevaluate the languages chosen for advanced courses, we will look for a way to work Dracula/ACL2 or similar tools into the curriculum.

At this point, we are also looking into introducing some of these ideas into the symbolic logic course, since AP requirements do not constrain it to Java. We will consider both the Northeastern approach in this endeavor and an approach used at OU. We have had for many years agreements in place for concurrent registration between our courses and equivalent ones at Arkansas colleges, and would try to fit any new offering in symbolic logic into this framework.

To facilitate interchange in the Texas/Oklahoma area of ways to work concepts related to the material of this workshop into computer science curricula, we would like to begin a collaborative workgroup, with activities including in-person meetings at a central location.