Abschlussarbeit

In abstract argumentation, self-attacking arguments and the unequal treatment of arguments in odd- and even-length attack cycles are some of the issues of classic semantics that are addressed by weak admissibility semantics. Most computational problems under weak admissibility semantics are PSPACE-complete. Our research presents a direct implementation to decide credulous acceptability under weak admissibility semantics. On the other hand, statistical analysis and machine learning have been widely used to solve complex problems with the aid of data. The question arises whether these tools could also provide insights about the credulous acceptability of arguments under weak admissibility semantics. This research answers this question by gathering data on argumentation frameworks and arguments, determining the credulous acceptability status of each argument, and statistically analyzing the data. Our findings, based on logistic regression, show that simple techniques are able to determine credulous acceptability under weak admissibility semantics better than chance. Furthermore, based on the results of the analysis, we assess the feasibility of using more complex models (e.g. Support Vector Machines, Graph Convolutional Networks) to solve the above mentioned problem, showing moderately better results.