Michelle O. Crosby, ASA Science Policy Fellow, and Chaitra H. Nagaraja, Chair of the Scientific and Public Affairs Advisory Committee
On August 20, the American Statistical Association submitted a formal comment to the Department of Education regarding its Proposed Priority and Definitions—Secretary’s Supplemental Priority and Definitions on Advancing Artificial Intelligence in Education, assisted by members of its Scientific and Public Affairs Advisory Committee, the chair of the Committee on Data Science and AI, and statistical education experts.
While the ASA leadership agree with the department’s goal of preparing students for a society in which AI plays a growing role, they argue that the proposed rule is incomplete because it fails to explicitly include statistics as a core pillar of AI literacy. They contend AI is built on data and understanding data fundamentally requires statistics.
The ASA’s submission highlights that AI is an evolution of data science—a field that relies on a three-pronged foundation of statistical science, computer science, and domain expertise—which emphasizes the idea that statistics is essential for AI literacy. Every AI model has inherent assumptions and limitations, is built using algorithms, and is evaluated using metrics. Without statistical and data literacy, students will not be able to critically evaluate an AI system’s accuracy, detect statistical bias, or improve its performance.
To remedy this, the comment urges the following changes to the proposed rule:
- Amend the background to state that students must understand both computer science and statistics to move from passive users to active creators of AI. It suggests including concepts such as data analysis and statistical modeling as foundational skills.
- Edit the proposed priority bullets to explicitly include statistics and statistical literacy in teaching practices, K–12 and higher education offerings, and teacher professional development programs.
- Incorporate a definition of statistics, “the science of learning from data, and of measuring and communicating uncertainty, often for the purpose of making better decisions.”
The ASA’s comments also provide a detailed justification for its recommendations, explaining the foundational role of statistical literacy in AI. It notes that AI systems, including machine learning and deep learning, are inherently statistical. A curriculum focused solely on the mechanics of algorithms, without the underlying statistical rationale, will produce practitioners ill-equipped to evaluate data quality, understand model limitations, diagnose failures, or innovate responsibly.
Furthermore, the comment states statistical literacy serves as a statistical bridge for effective collaboration among the diverse experts—computer scientists, domain specialists, and ethicists—who work on AI projects. It provides a common language for understanding the inherent uncertainties and limitations of models. This guarantees interdisciplinary teams can work together to ensure the integrity, reliability, and ethical deployment of AI systems.
The submission concludes by stating that ethical AI is at its core a statistically informed endeavor. Principles of accountability, transparency, and fairness depend on the ability to identify and quantify biases and uncertainties, which are fundamentally statistical problems. The document concludes urging the Department of Education to consider these points and explicitly incorporate the fundamental role of statistical literacy into its proposed priority for advancing AI in education.
For further insight into this topic, read an earlier statement and blog post about the role of statistics in data science.
Leave a Reply