Introduction

Since the public release of ChatGPT 3 years ago, generative artificial intelligence (GenAI) has moved from novelty to ubiquity in higher education. Yet, discourse around it has often outpaced evidence. On one hand, we have seen sweeping proclamations that AI will displace writing, undermine academic integrity, or even render education worthless. On the other, we have heard equally expansive claims that GenAI will democratize access, personalize instruction, and solve long-standing pedagogical challenges. What has been striking is how little of this conversation has been grounded in systematic, empirical inquiry. Pessimism and optimism have both tended to run ahead of data, leaving us without reliable models for integration.