REFLECT presents a human-centered framework for examining bias in large language models through perspectives from computer science, human–computer interaction, and cognitive psychology. The tutorial considers several ways bias can appear in model behavior, including selection effects, acquiescence, and stereotypical associations, and asks what these patterns reveal about the interaction among human data, model training, and generation. It also introduces interaction and design strategies for making potential biases visible and open to critical interpretation. The goal is to equip participants with practical and conceptual tools for evaluating LLM bias while supporting more transparent, accountable, and trustworthy human–AI interactions.