Centre for Research in Reasoning, Argumentation & Rhetoric along with the PhD in Argumentation Studies at the University of Windsor invite you to a talk by
Oxana Pimenova, Ph.D., SSHRC Postdoctoral Fellow in Argumentation Studies
Developing a predictive machine learning model to detect and forecast Argument Continuities in government-led reasoning: the case of superficial Indigenous consultations
Abstract: In Canada, governments must consult Indigenous communities on resource projects. When a government agency believes in a project’s necessity, it has the institutional power to control the argument exchanges via imposing authority rules that define a reasoning capacity to argue for and against a project. For example, in limiting the evidence availability and putting more administrative burdens on Indigenous communities, rules can make it easier/less costly for the officials not to engage with epistemically diverse arguments but rebut them with an Argument Continuity. Argument Continuity is a set of arguments and counterarguments repeatedly produced and reproduced by the same dominant arguer through an adversarial reasoning process to dismiss unfavorable arguments without considering their merits.
In a distorted reasoning context of Indigenous consultations, Argument Continuity reconstructs motivated criticism in the sequential development of reasoning goals, practices, and outcomes by a government agency responsible for those consultations. The knowledge of Argument Continuity will increase the negotiation power of Indigenous communities and prevent the officials from rebutting outstanding Indigenous concerns with hollow responses. Following this practical exigency in predicting motivated criticism in Indigenous consultations, the project will detect Argument Continuities automatically with the help of a supervised ML algorithm.
Firstly, the Annotation Scheme will be developed to specify how to label entities (arguments, counterarguments, premises, conclusions) and categories (authority appeal, uncertainty appeal, irrelevance appeal, non-credibility appeal, incorrectness appeal, out-of-the-scope appeal, a lack of substantiation appeal) comprising a sequence of Argument Continuity. Secondly, the Scheme will be used to manually annotate text data of Indigenous consultation reports issued by a government agency. The outcome will be the training dataset with target labels/categories for Argument Continuity. Thirdly, the labeled data will be supplied to a supervised model for classifying and predicting Argument Continuities on new, unseen data (based on the patterns learned from the training data). In the classification task, each step of a sequence of Argument Continuity will be paired with its corresponding labels and categories. In the prediction task, the algorithm will predict labels and categories to which new text data belongs. The algorithm is not limited to the context of Indigenous consultations. However, it can be applied in any distorted context of public policy argumentation to combat fallacious moves by institutionally dominant arguers who are more likely to employ motivated criticism towards opponents.
Friday, November 10, 2023
3:00 pm
Chrysler Hall North, 1163
All Welcome