Large language models (LLMs) are increasingly used for zero-shot conversation summarization, but often exhibit positional bias—tending to overemphasize content from the beginning or end of a conversation while neglecting the middle. To address this issue, we introduce PoSum-Bench, a comprehensive benchmark for evaluating positional bias in conversational summarization, featuring diverse English and French conversational datasets spanning formal meetings, casual conversations, and customer service interactions. We propose a novel semantic similarity-based sentence-level metric to quantify the direction and magnitude of positional bias in model-generated summaries, enabling systematic and reference-free evaluation across conversation positions, languages, and conversational contexts.Our benchmark and methodology thus provide the first systematic, cross-lingual framework for reference-free evaluation of positional bias in conversational summarization, laying the groundwork for developing more balanced and unbiased summarization models.

Industry strives for trustworthy Artificial Intelligence (AI) systems through recognizing and implementing Responsible AI principles. Solutions supporting that goal are of the utmost interest in that context. Standardization is an essential element here, as it provides a platform for industry to discuss and facilitate not only the development of practical rules and requirements but also ways to implement AI based systems. One of Responsible AI principles is fairness, and bias is a serious obstacle against it. First, we explain the concept of Responsible AI and highlight results of our analysis on bias and fairness in ongoing international standardization works and AI Act (AIA). We identified a gap between the principles defined by high-level studies, including the AIA, and their practical implementations, and differences within standardization and research works. Second, we draw a standardization map for AI works. Finally, we state how international standardization bodies may fill this gap?