The Toulmin Argument Model in Artificial Intelligence Or: how semi-formal, defeasible argumentation schemes creep into logic

In 1958, Toulmin published The Uses of Argument. Although this anti-formalistic monograph initially received mixed reviews (see section 2 of [20] for Toulmin’s own recounting of the reception of his book), it has become a classical text on argumentation, and the number of references to the book (when writing these words1 — by a nice numerological coincidence — 1958) continues to grow (see [7] and the special issue of Argumentation 2005; Vol. 19, No. 3). Also the field of Artificial Intelligence has discovered Toulmin’s work. Especially four of Toulmin’s themes have found follow-up in Artificial Intelligence. First, argument analysis involves half a dozen distinct elements, not just two. Second, many, if not most, arguments are substantial, even defeasible. Third, standards of good reasoning and argument assessment are non-universal. Fourth, logic is to be regarded as generalised jurisprudence. Using these central themes as a starting point, this chapter provides an introduction to Toulmin’s argument model and its connections with Artificial Intelligence research. No attempt is made to give a comprehensive history of the reception of Toulmin’s ideas in Artificial Intelligence; instead a personal choice is made of representative steps in AI-oriented argumentation research. When Toulmin wrote his book, he was worried. He saw the influence of the successes of formal logic on the philosophical academia of the time, and was afraid that as a consequence seeing formal logic’s limitations would be inhibited. He wrote The Uses of Argument to fight the — in his opinion mistaken — idea of formal logic as a universal science of good reasoning. In the updated edition of The Uses of Argument [19], he describes his original aim as follows: