Can Memory-Augmented Large Language Model Agents Help Journalism in News Interpretation and Cognitive Framing Across Diverse Audiences?

News media are critical in conveying information and shaping public perception across diverse domains, including technology, finance, and agriculture. The interpretation of news often varies significantly among different audiences due to their specific expertise, age, and standpoint, leading to comprehension gaps. In this work, we investigate how to identify these comprehension gaps and provide solutions to improve the audience's understanding of news content, particularly regarding the aspects of articles outside their primary domains of knowledge. We propose an agent-based framework using large language models (LLMs) to simulate society communication behaviours, where several agents can discuss news. These agents can be designed to be experts from various occupations, or from different age groups. Each agent has semantic (conceptual understanding), episodic (event abstraction), and procedural (analytical chain of thought) memory. This framework allows us to monitor and analyse agents' behaviour and cognitive processes. Our findings indicate that this framework can identify confusions or even misunderstandings of news through the iterative discussion process. Based on this accurate identification, the framework can design a supplement material according to the characteristics of the three memory types. Our findings show that agents exhibit significantly improved news understanding after receiving this material. We also show how content can influence the audience's perception. These findings highlight our framework's utility and efficiency in enhancing news comprehension for diverse audiences.