Researchers propose Re2A, a framework for situated conversational recommendation. It uses rubric-based preference reasoning to guide response generation and optimize for user preference satisfaction and situation consistency.