Argumentative Tension through Semantic Displacement. An Experiment Applied to the Debate on AI Development.

This article examines the generation of argumentative diversity from positions produced by artificial intelligence itself in debates about its development. It presents a semantic displacement experiment applied to a debate from the synthetic syncIALO corpus, using a language model, an external ontology represented in SKOS (Simple Knowledge Organization System), and a procedure of argumentative counterposition aimed at preserving genuine tensions. An original debate comprising twenty arguments and a transformed debate comprising twenty-one were compared through qualitative analysis and a size-adjusted composite diversity index. Eight transformations were identified, including the introduction of a concept of dual contestability, argumentative reformulations, and new relationships between distant regions of the debate. The overall diversity index increased by 5.87%, with increases in dimensional diversity (3.78%), SKOS richness (22.49%), and functional diversity (9.89%), while normalized semantic distance decreased by 2.98%. The findings provide exploratory evidence that AI-generated adversarial positions can be reconfigured to broaden ontological and functional variety without eliminating their tensions.