Francesca Toni

Argumentation semantics could be the key to reliable and explainable debate judgement in AI, outperforming traditional LLM-based methods in formal guarantees.

LLMs may seem rational, but they operate on "superficial beliefs," revealing a gap between decision-making and self-reported reasoning.

LLMs can now offer globally contestable decision support by systematically mapping decision spaces into argumentation frameworks, allowing users to challenge the underlying logic, not just individual outputs.

Forget hand-tuning argument base scores – this new method automatically extracts them from user preferences, unlocking more transparent and contestable AI.