The AI-R project develops neuro-symbolic methods for Natural Language Inference aimed at automatically extracting argumentative structures from text. By identifying claims and their inferential relationship (how propositions function as reasons for or against one another), the project seeks to model individuals’ webs of belief as networks of commitments linked by logical relations. Grounded in philosophical inferentialism, AI-R ultimately aims to build a dialectical interface that visualizes these belief networks, makes inconsistencies explicit, and supports a deeper understanding of argumentative reasoning within the space of reasons.