Do you have to be polite to AI? (www.bbc.com)

Recent research delves into the impact of politeness and creative interactions when communicating with AI chatbots, revealing that traditional beliefs about “prompt engineering” may be misguided. In a study, various chatbots were subjected to encouragement and role-playing scenarios, including a unique approach where one model acted as if it were on "Star Trek," leading to improved mathematical performance. However, other common strategies—like being polite or using threats—did not consistently enhance accuracy. Experts like Jules White emphasize that effective AI communication is less about word choice and more about clearly articulating objectives. The findings have significant implications for the AI/ML community, highlighting the need for refined prompting techniques as AI models evolve. As newer models become adept at discerning core instructions, tactics like politeness may not yield better results; instead, encouraging users to frame requests as tool-driven interactions could lead to improved outcomes. While habitual niceness may foster a more comfortable user experience, experts encourage focusing on traditional tools and techniques for optimal AI performance rather than relying on the anthropomorphism of chatbots.