Evaluation of Alternative-Based Information Systems for Deliberative Polling using an Agentic Simulator

The paper presents the Agentic Bipolar Argumentation Simulator (ABAS), an LLM-based tool designed to evaluate deliberative polling solutions by simulating autonomous shareholder agents that engage in argumentation and voting. It utilizes a formal framework to address the coverage problem in decision-making contexts, implementing a ranking mechanism for justifications based on observable endorsement mass and assessing the impact of various parameters such as creativity rate and population size on coverage and diversity. This research is significant for practitioners as it offers a method to enhance the robustness of deliberative polling systems against strategic manipulation, particularly in adversarial environments.