The integration of artificial intelligence (AI) into personalized learning environments is transforming educational paradigms by enabling adaptive, learner-centered support. This paper presents a hybrid human–AI collaborative cognitive tutoring system designed to address the limitations of conventional intelligent tutoring systems (ITSs), particularly their lack of metacognitive scaffolding and emotional sensitivity. Our approach leverages a modular architecture comprising three core components: a Cognitive Dialogue Module for Socratic questioning and content adaptation, a Metacognitive Regulation Module for promoting reflective learning, and an Affective Adaptation Module for emotional responsiveness. The system was developed using a Design-Based Research (DBR) methodology and evaluated through controlled simulations with diverse synthetic learner profiles. Results demonstrate the system’s capacity to deliver tailored cognitive support, foster metacognitive reflection, and partially adapt to learners’ emotional states. While affective modulation proved effective for some learner types, lower responsiveness was observed in high-stress scenarios, indicating areas for refinement. This study contributes to the advancement of intelligent educational systems by emphasizing co-agency, emotional engagement, and ethical oversight. It lays the groundwork for future empirical testing in real-world contexts and proposes a shift toward human–AI ecosystems in which AI acts not only as a tutor but as a cognitive and emotional partner in the learning process.