Be Nice or Be Correct?

Do you say "please" and "thank you" to AI? While it feels natural to treat AI like a human colleague, this "anthropomorphism trap" might actually be degrading your results. In this session, we’ll explore the empirical link between prompt psychology and AI output quality, introducing the concept of the "Correctness Tax"—the measurable drop in accuracy that occurs when we use polite, sarcastic, or frustrated tones.

Drawing on recent research evaluating coding models, we'll look at how adding emotional baggage to a prompt can turn an AI into a sycophantic "yes-man" that agrees with your mistakes instead of fixing them. You’ll walk away with practical, actionable strategies to strip the fluff from your prompts, treat your AI less like a coworker and more like a compiler, and immediately improve the quality of your daily workflows.

About the Speaker: Selva Jagannathan is the Global Lead for GTM Applied AI Design and Adoption at Google Cloud. Based in Dallas, TX, he is a hands-on architect currently building next-generation, real-time AI agents that augment enterprise workflows across Google's own global operations. With 20 years of experience spanning digital transformation, cloud architecture, and AI strategy, Selva has advised over 25 Fortune 500 enterprises on adopting cutting-edge technologies. He is a published researcher in the Discover Artificial Intelligence journal, focusing on human-AI interaction.