Peer review, the cornerstone of academic publishing, is a human process, prone to biases. But what if we could make it fairer and more transparent with AI? Researchers are exploring how to use Large Language Models (LLMs), the tech behind ChatGPT, to help automate parts of peer review. One challenge? LLMs are often "black boxes," making it hard to understand their decisions. A new research project called PeerArg attempts to solve this using "argumentation frameworks." Imagine each review broken down into individual arguments, with pros and cons weighed against each other like a debate. PeerArg extracts these arguments using an LLM, classifying them by aspects like "Clarity" or "Novelty." It then builds a computational model of the arguments supporting or attacking a paper’s acceptance. By assigning strengths to these arguments and evaluating how they interact, PeerArg makes a final acceptance prediction. This approach offers a more transparent alternative to simply feeding entire reviews into a black box LLM. Early tests show PeerArg outperforming a simpler LLM-only approach in predicting paper acceptance across various datasets. The transparency of argumentation frameworks also provides insights into *why* a decision was reached. For instance, if a paper is rejected, we can see which aspects were its weakest points based on the argument analysis. This opens exciting doors for future research. One challenge is incorporating the nuances of human language into the argumentation model. Sarcasm, subtle disagreements, and complex reasoning are hard even for humans to process, let alone represent computationally. Future work could also combine multiple reviewers’ argumentation frameworks into a “multi-party debate,” leading to even richer insights. While a completely automated, bias-free peer review system is still a long way off, projects like PeerArg offer promising steps towards a more transparent and objective future for academic publishing.