As engineers, we know AI is built on algorithms and probabilities. But in practice, the hardest part is often not the system itself. It is how we as people respond to it. The problem is not just that AI makes mistakes, but that we react strongly to both its mistakes and its confidence.

We already trust AI with a lot. We use copilots to move faster, generate ideas, and reduce routine work. But while we rely on these systems every day, we often react differently to their mistakes than we do to human mistakes. We accept mistakes from colleagues because we understand that humans are imperfect. With AI, one wrong answer can make us dismiss it too quickly, while a confident answer can make us rely on it too easily.

This talk explores trust in AI through the lens of psychology. It argues that trust in AI is not a simple matter of trust or distrust, but a calibration problem. Through examples from coding and AI-assisted workflows, you will learn why people misjudge AI and how overtrust and undertrust show up in engineering work. You will leave with a clearer sense of when to rely on AI, when to verify it, and how to stay in control without losing its benefits.