Artificial Psychological Engineering: A Taxonomy and Audit Framework for AI Systems That Shape Emotion and Behavior

AI systems increasingly sense users' affective states and adapt their behavior to change what users feel, believe, and do. The same techniques appear in mental-health chatbots, workplace tools, AI companions, tutors, and engagement-maximizing feeds, with very different consequences for users. We introduce Artificial Psychological Engineering (APE) as an umbrella term for the deliberate design of AI systems to influence psychological states, and propose a four-axis taxonomy classifying systems by mechanism, transparency, goal alignment with the user, and adaptivity. From it we derive an audit procedure with red-flag combinations, apply it to five system classes, map it to current regulation including the EU AI Act and India's DPDP Act 2023, and propose a validation plan. We argue that risk rises most sharply with the combination of high adaptivity, low transparency, and goals that diverge from the user's own.