In light of the growing enthusiasm but also skepticism surrounding advancements in generative artificial intelligence (AI), it is essential to systematically distinguish the diverse effects AI can have on learning. This chapter introduces and elaborates on the ISAR model, which distinguishes four types of effects of AI on cognitive learning processes and outcomes: inversion (I), substitution (S), augmentation (A), and redefinition (R). Drawing on foundational debates in educational technology research, the ISAR model synthesizes and extends existing conceptualizations by clarifying that AI can replicate existing instructional approaches (substitution), add value through enhanced support (augmentation), or help transform the nature of learning tasks (redefinition). However, the model also highlights the risk of reducing cognitive engagement, which can undermine learning effectiveness (inversion). Building on theoretical and empirical insights, this chapter uses examples of empirical studies on generative AI to explain each ISAR effect. This chapter aims to guide researchers, practitioners, and policymakers in understanding the varying pedagogical value of AI and the conditions under which it may support or hinder meaningful learning.