Issues of feasibility, coherence, and robustness in a premise-to-claim model of argumentation: Results from four experiments

The paper considers several issues. (1) The feasibility of a premise to claim model of dialogue: the results show that a rule-based system can generate claims in the same sequence as they appear in a real discussion. (2) The relationship between the increased knowledge created in the leap from premise to claim on the perceived coherence and comprehensibility of dialogue: the system's reaction to making available all premises before run time was to provide a rapid summary of the debate, whereas the result of randomly rearranging the order of introduction of premises showed that the majority of claims were still made, and that a very similar sequence of claims was followed; for the first half the observed sequence was taken by the simulation program, because initially the number of (randomly introduced) premises was so small that they did not match the available warrants. (3) Robustness of a rule-based argumentation system in the face of expert fallibility: the test condition here was a reduced set of rules (as a proxy for lack of expert knowledge) which, however, showed an overall convergence to the same claims in the same sequence as when all the rules were used.