What if we studied artificial intelligence the same way psychologists study people? This episode explores machine psychology — a new discipline that treats large language models as participants in controlled behavioral experiments. Drawing on work by researchers at Google DeepMind, the University of Stuttgart, and TU Munich, we unpack how AI systems reveal patterns, biases, reasoning styles, and social tendencies that echo human cognition. Researchers adapt classic paradigms from cognitive and social psychology to probe model behavior. They test memory and generalization, measure inductive and deductive reasoning, and examine how prompts modulate cooperation or competition in multi‑agent settings. These experiments reveal signatures of learning and bias that are not visible from benchmark scores alone. The result is a framework for studying models as dynamic systems that respond to environments and incentives. Machine psychology complements mechanistic interpretability by focusing on externally observable behavior: not how the system is built, but how it acts under pressure. This perspective also spotlights ethical concerns — from stereotyping to goal misgeneralization — and invites standards for reproducible behavioral testing across labs and model providers. Produced by Cognivault — insight, intelligence, and innovation made clear.