Large language models (LLMs) have shown remarkable potential in various applications, but their decision-making processes often lack transparency and can be difficult to challenge. A new research paper proposes a novel approach to address these limitations: making LLMs argue with themselves. This innovative technique, called "argumentative LLMs," involves using LLMs to construct arguments for and against a given claim, creating a structured argumentation framework. Think of it like an internal debate where the LLM plays both sides, generating supporting and opposing arguments. Each argument is then assigned a strength score, also determined by the LLM, reflecting its relevance and persuasiveness. These arguments and their strengths are then evaluated using formal reasoning methods from the field of computational argumentation, leading to a final decision. This approach not only improves the LLM's reasoning abilities but also makes the decision-making process more transparent and contestable. By examining the arguments and their strengths, users can understand the rationale behind the LLM's decision and even challenge it by adding or modifying arguments. The researchers tested argumentative LLMs on claim verification tasks, achieving competitive results compared to existing methods. More importantly, this approach offers a significant step towards making AI decision-making more explainable and trustworthy, especially in complex, high-stakes scenarios. Imagine an AI system helping doctors make diagnoses, not by simply providing a result, but by presenting a reasoned argument based on available evidence. This allows doctors to understand the AI's logic and potentially identify biases or errors. While the current implementation is a starting point, future research could explore more sophisticated argumentation frameworks and methods for assigning argument strengths. This research opens exciting possibilities for building more transparent, reliable, and collaborative AI systems that can assist humans in making informed decisions.