Tag
The paper proposes an auditable, user-configurable rule-based method for argument selection in deliberative polling, demonstrating through simulations that it achieves competitive performance with opaque learned rankers while enabling greater transparency and personalization.