As AI rapidly evolves, this question may be of particular interest to the statistical community: How are statistics, data science, and AI related, and how might their relationships evolve from here?
My recently published article in Harvard Data Science Review, Data Science at a Fork in the Age of AI: Recognizing Divergent Missions in Research and Education · Issue 8.3, Summer 2026, reflects on this question from the perspective of how AI is changing what we consider to be data.
The article starts from three shifts that I believe are becoming increasingly visible: from relation-based to expression-based data, from sample-based inference toward population-scale modeling, and from parsimonious models toward expansive representations. These shifts do not make traditional statistical analytics less important. Rather, they broaden the analytical landscape and raise a question about whether data science may increasingly serve partially divergent missions: one continuing its strong statistical and decision-oriented foundations, and another moving toward the generative, multimodal, and adaptive systems increasingly associated with AI.
The implications for data science education may be particularly important. As AI makes technical execution increasingly accessible, what should we continue to teach as foundational knowledge, and what new forms of analytical judgment and responsibility should we develop?
I share the article here not to suggest that these questions have settled answers, but in the hope of stimulating discussion within the statistical community. I would be very interested in hearing how others see the evolving relationships among statistics, data science, and AI.
Read the full article at: Panorama · Harvard Data Science Review
------------------------------
Zhiwei Zhu
Purdue University
------------------------------